This is gold, love the deep dives as this usually, for me at least, makes things simpler. I know people always think abstractions are better for that, I guess they are when you want to learn fast, but I always struggled with that in school. Deep diving for sure made my homework go slow, but I understood stuff way better. :)
The Miller effect is a multiplication of collector-base CAPACITANCE by the amplifier gain... totally different than base charge of a saturated transistor. A Schottkey diode from base to collector will prevent saturation. This trick is employed in 74Sxx and 74LSxx logic chips.
That definitely answers the transistor issue. Great to see this explained. Never heard about the Miller effect. Thanks for this ‘down the rabbits hole’ video.
This gets even crazier when you want to see devices' overload characteristics and stuff. I build guitar pedals and vacuum tube sims using discrete transistor circuits. There are so many undocumented "effects" that I've learned about and used. I can say that math is great for when you are trying to nail down a specific design that requires precision. The art of using them creatively is a whole different level and it has taken me over 10 years to learn about and document a whole slew of things that aren't documented.
Coffee in one hand, donut in the other and time to watch Noel's video. ps. wow. that was an interesting "eye opening" video. I didn't know about this charge build up effect on base.
The Z80 clock signal spec is more demanding than regular TTL and I suppose they could not integrate the driver on the gate array. The inversion is just a side effect and the gate array could have inverted it for sure, but the primary use of the transistor is to improve the signal. I forgot the details, but it was quite common to use a transistor to match the Z80 clock spec. Note that they used a 74HC04, not a 74HCT04 in the second revision, which is a hint in that direction as well. MOSFETs are a completely different technology than bipolar and while you can get lucky, the considerations to drive them are very different.
16:22 The small spike at the output right before the signal falls is also caused by the Miller effect: The input rises at that moment, and the rise goes directly through the parasistic capacitor to the output. The Miller effect amplifies the spike. After a very short delay, the MOSFET starts conducting and the voltage drops to zero.
@@airspeedmph Main reason is the retro computer renaissance, there is so much happening around the Commodore 64 and other retro computers these days that consumes a huge amount of hobby time. Some time ago SteamOS and gaming on Linux was brand new so that was where I was spending my spare time on. But currently... nothing less than wonders are happening on computers I love and I want to enjoy this.
@@danielmantione Totally get that, I spend my time alternating the two (Linux and retro) although I have to admit that the retro part I'm only experiencing vicariously through channels like this, since my only HW that can be called retro is a PS1 :) Lately I'm seriously consider getting real by acquiring some hardware, though I'm not sure where to begin. I'm somewhat inclined towards a Commodore 64, but a working one seem to be rather expensive. But yeah, is crazy what's going on in the retro space, new hardware+addons and even games, is impressive and alluring. And another thing that I especially find relaxing in the retro space is that everything is governed/motivated by passion, there's no in-fights, politics and ideology to be found here. Anyway, have fun.
I restore 2" broadcast video tape machines which are over 60 yrs old and work in the analog world. Compared to those beasts, digital is a piece of cake. I have to find transistors that match in many parameters, some even have to be matched as well because they are used in a balanced circuit. And in the analog circuits they must have a freq response from 30hz to 6mhz. When I was maintaining a complex digital device i thought as long as timing was correct and waveform was clean. I was good to go. Easy Peezy
Fantastic video, Noel! You're surpassing yourself. And I was delighted to see that you also had to battle the Bibl... err... the Sedra as an undergraduate. So did I! :-) Regarding the clock inverter, if the gate array uses the uninverted clock signal internally, it makes sense to invert it before passing it on to the CPU: the 180-degree phase shift helps account for, and even take advantage of, the propagation delay of the other signals shared with the CPU. The Atari line of 8-bit computers makes liberal use of this kind of phase shift to ensure proper synchronization between CPU and other active components, and it effectively doubles the speed of the system, since you get two clock tics per clock cycle, each in a different part of the system. Back in the day this kind of hack mattered. I also would be very careful about biasing any switching transistor without a base resistor. You got lucky (or perhaps you knew it) that the gate array had an internal pull-up. Otherwise the pin wouldn't be able to source the kind of current you get through a diode to ground, and the chip might have been ruined.
Thank you! Yes, the inverted clock signal makes a lot of sense. The thing that surprises me is that the Gate Array doesn't invert it before outputting it since nothing else on the board is using the uninverted one. I guess they either completely ran out of space in the GA or it was a "fix" they had to do afterwards.
@@NoelsRetroLab That was probably because of a limitation of the specific process technology used in the gate array, which may have prevented either enough buffering capacity (i.e. cannot drive or sink enough current) or fast-enough switching at the required current. It surely makes sense. An external transistor can use very different process technology than logic and can have larger die space than what could be done in the gate array.
My analogue electronics lecturers championed Sedra & Smith, and I was heading towards failing the course... Then one of my classmates showed me his copy of The Art Of Electronics, and suddenly everything was alright. =)
5 V digital is easy ! NPN is 2N3904 , PNP is 2N3906 just check Ic max …......... If the gate-array is open collector you need a pull up resistor on the transistor base . ( 1k - 10 k across Vcc & base should work ) . PS The 555 has a dc pedestal of around 0.7 - 0.9 V so off / low is not 0 .
For that particular transistor I have no problem: 1. I used to work for the manufacturer (Ferranti) 2. I have many ZTX312's and many other ZTX series devices from 45-50 years ago! 😱😳 3. I have several Towers' transistor selector books that detail the critical specs of the transistor and also give equivalents from that time (mid '70's!). I often find that modern equivalents quoted are actually NOT and finding an equivalent of the original equivalent works much better. 😎 Nice explanation of a really good approach though. 👍👍👍
Awesome video. Very interesting but also easy to follow those of us without a background in electronics. I'd love to see some similar content looking at different types of capacitors.
This is all switching stuff. Way easier than matching RF. The transistor is being used as a buffer. Not for current or voltage but to provide abstraction from capacitance. In most instances where it is a switching device Hfe is irrelevant. Frequency transition, Input/Output capacitance, t-rise, t-fall and t-storage are the parameters to look at. A 2n2222A would be unsuitable due to the storage time. A 2n3904 might be close enough but it has way more storage time than the ZTX. A FET is a poor choice for switching due to the insulated gate being a capacitor.
Those round bodied transistors are pretty common here in slovakia/czechia the local electronics manufacturer(Tesla) used that design a lot, even for OP, amps and some simple IC's
Looked at the datasheet for the ZTX312 and noticed that it specifies rise, fall and storage times - and the storage time is 13ns. The BC547 doesnt specify this, but a close cousin the 2N3904 does - 200ns! That explains why the speedup capacitor is needed. Interestingly I can't seem to find a common TO-92 transistor that has good switching times...
You have to be careful substituting mosfets because then the switching threshold becomes an issue - you have to raise the gate above the drain voltage by a certain amount to get the mosfet to saturate, whereas with a BJT you only need 0.6v, a MOSFET can often be up to 10V.
Yeah, I was really puzzled by the lack of timing info on the BC547. I wasn't sure if it was because it was fast enough for anything or just not intended for anything that you cared about timings. I guess it's because the answer wasn't a near number you can put on a table (and it would look really bad w/o the capacitor).
Try a 2N2222(A). It's a switching transistor, so its speed and transition times are optimized for switching. It is available in TO18 hermetic metal case with leads, a TO92 plastic case as a PN2222A, and SMD. Works for anything switching related below 300MHz. Check its spec sheet. Available everywhere and cheap (Digikey, Mouser).
Such a great hands-on guide to this issue I have been struggling with. Much of this is relevant with fet-based audio compressors from days gone by. Because the characteristics of the fet are what determine the some of the 'attitude' of the compressor, it is important to know how to find a modern match with similar properties. This now seems less daunting, thank Noel!
What sucks is when you are repairing motor drive circuits like an H bridge where one of the transistors is dead and made of unobtainium. Meanwhile a popular replacement is listed as close matching. In that case you get to rebias the circuit, or try to balance it. I fixed a 45 jukebox with a mobile player carriage. All I could do was one direction was normal speed the other was half speed. That is why I praise the use of common parts.
* 13:45 Huh, not quite a "perfect output", the falling edge is heavily non-monotonic which is pretty back for a clock signal ... it doesn't happen in the final circuit on the board, so not an issue but in that test setup it's quite messed up. * As noted, the output stage from the gate array is probably an internal pull up and a nmos driving the signal down, providing a high output impedance when the output is '1' and a low impedance '0'. But then that means on the falling edge it should have plenty of capability to sink the charge left over in the base junction of the transistor to turn it off quickly (since in the board there is no base resistor). So I'm not entirely sure why it had an issue.
Fun interesting stuff. I subscribed ☺️. I haven't practiced electronics in over 20 years. I'm just now getting back into it. Been building my Electronics Laboratory Bench. Going back to school this year. It's not about making money it's about having fun 😊!
Love to hear that! I also hadn't done electronics in many years (focused on the software side), so it's been a really fun journey getting back into it.
I'm afraid that's a gross over-simplification. In high-speed switching circuits, the current into and out of the gate capacitance of a MOSFET means that it cannot be considered as simply "voltage-controlled". Once you've added a current limiting resistor between input and gate and a pull-down resistor between gate and source, there won't be a lot of difference in the design between BJT and MOSFET.
That was a fantastic video! It had meat on it's bones and you explained it in a very clear manner. Coincidentally I also had to find some modern replacement transistors for a board I've been working on. I loved the experimentation, keep it up!
I don't plan to use transistors, even I don't have plans of working with electronics at this level but, oh man, I enjoyed this video A LOT! Your videos are always full of high quality material, presented in a way that everyone can understand. Great work!
Brilliant, just the right depth. I always wondered how to determine the correct replacement transistor and this provides a great starting point. More like this please Noel!
Very nice research on the transistors. Thank you for the effort showing the process! As for the 2nd edition of AoE, I recently said "What I learned from technical books: Contrary to popular belief they in many cases do not get better with revisions. It's often worth keeping earlier ones around." I'm glad I'm not the only one who found that. It's definitely worth it to keep older revisions.
The book 'Art of Electronics,' in Polish known as 'Sztuka elektroniki,' is excellent. It contains a lot of content and is easy to understand. I've been using this book throughout my entire high school, or as we say in Poland, 'W Technikum.' :)
Thank you for covering this topic. BC547 is just cheaply made transistor intended for low frequency amplifiers and similar applications. There are also specialized high frequency transistors available. A metal can package was actually pretty popular in the past. For example, KC507 transistor, a BC547 clone, was produced in metal can as well.
The follow up book to 3rd edition of "Art of Electronics", which is title "The Art of Electronics: The x-Chapters" has a section on the Miller effect. It can out in Jan 2020. You can read up on it at x.artofelectronics.net
a cynical view from the real world: 1. a difficult problem that takes a while to research and solve by an engineer. 2. the engineer documents it so that nobody has to re-do the effort. 3. nobody reads said document. 4. go to 1
Excellent work and thanks for sharing. I fully agree that older books are always good to have around. Was doing the same or similar experiments about a month ago.
There's usually several different transistors you can use as substitutes for most of the commonly used transistors, I've been pretty lucky with regards to transistor substitutions.
Years ago, when I still used transistors for prototypes, I considered that there were two types of BJT transistors: for amplification and for switching. Even when fast, the transistors for amplification were not suitable for switching (the 300MHz bandwidth for BC547 showed you that), what was needed was the switching time to be very low. I usually used 2N2222 for switching up to 1MHz and BC547 as amplifier. Never reversed them, because the results were bad, 2N2222 was not a good amplifier and BC547 was not a good switching transistor, (you saw it at 21:45). By looking at the datasheet, can see if a transistor is meant for switching applications, because there are few key parameters like switching time from low to high, from high to low, etc. You should have showed that in the video, not many people know how to read a datasheet.
I think Noels point was mostly to demonstrate that you need to do something extra to get the advertised 300MHz and this is valid for any transistor: A speedup capacitor will speed up a 2N2222 just as it speeds up a BC547.
@@danielmantione No, those 300MHz are one thing, they are valid in amplification. Make a amplifier, a oscillator and you will reach tenths of MHz domain, maybe hundreds, depending of the quality of the component. And he had this on his BC547. What he needed was the commutation speed that is give by the special construction of the transistor, (small capacities inside it, and others). And that was achieved with the transistor used in the original design without additional components. One can design a circuit, but to be replicate on industrial scale, the circuit must behave the same independent of the tolerances of the components. Noel said that, to fiddle with the components values in the BOM is not something that one will want to do when buy the kit to build it and enjoy it. The point he made (in my understanding) is that you must use the right component to achieve the desired goal. Only that he does not touched the subject I pointed out.
@@sebastian19745 GaAs FET's came out in the late 80s, often in a can w/ 4th gnd pin. Those things easily handled 600 MHz due to ultra low junction capacitance.
@@wiregold8930 Yes, but do you remember also theirs prices? One of these or a mosfet tetrode were quite expensive in 80s. Also, they were used mainly in RF, hence the extra ground pin. I got 3 BJT (BF200 or 214 I think it was, not sure) that had a 4th pin for ground, they were more expensive than a regular BC107/108, I did a 3 stage antenna amplifier with them. And they were pretty common by then, they were used in many TV tuners.
I would have thought like the TO220 package [heatsink tab] on the MOSFET's where the Gate is pin 1 to the left with the body facing you [heatsink tab to the back] pin 2 is Drain [also heatsink tab] and pin 3 is Source, that those small ones [TO-92] would also have the Gate be pin 1 with the flat side facing you and Drain be the middle pin [2] as well??? That would make for physical observation pinout identification. But I digress because on BJT's the middle pin 2 is the Base which is the same function as the Gate on a MOSFET.
In the past I once accidentally picked a BC308C PNP instead of a BC238C NPN for driving a ~20mA LED from an 74Nxx output. I wondered for several minutes why the LED lit up when the TTL output was low. Until I checked the printing again and noticed my mistake. But this wasn't just a PNP-NPN mistake, this also meant that the PNP had the C-E current running against its natural direction. The BC308C was still very happy and worked in other breadboard setups.
Ah yes. Good old electronics class. I think using a proper inverter would have made much more sense but they would obviously cost more. The revision was a wise route to take. The Miller effect is what deals with what is known as slew rate current ( the speed at which the amplifier can change it's output.)
Not quite. The slew rate is the maximum rate of change of output. In other words, it's the limit on rise time and fall time for edges in digital applications, and the frequency limit on large signal swings in analogue applications. In this circuit, the charge on the base capacitance is holding the transistor on (output low) in saturation for a period after the input signal goes low. This is effectively a propagation delay for the rising edge of the output, rather than a limit on its rise time. The fix for it when the input signal goes low is to find a way of sucking the charge out of the capacitance that the base of the transistor sees. If you have sufficient current sink from the input signal, then a capacitor across the current limit resistor will attempt to pull the base below ground when the input goes low and remove the charge on the base capacitance more rapidly, as shown with the BC547. If you have not much current sink, but more current source available from the input, then a low value resistor across the base-emitter (a "pull-down" resistor) will help to reduce the delay.
@@GORF_EMPIRE Don't patronise me; I'm perfectly aware of the Miller effect which is caused by the voltage gain of an inverting amplifier effectively multiplying the collector-base capacitance. The issue here is not the slew rate, which is clearly displayed on the rising edge of the output at 10:23 and can be seen to be around 5V in 200ns which equates to 25V/μs. The problem illustrated on that scope image is the 400ns that the BC547 takes to come out of saturation when the base current is shut off. That is entirely due to the charge on the capacitance as seen by the base having no low impedance path to discharge other than through the base-emitter junction. There is no Miller effect on a transistor _in saturation_ because a saturated transistor has almost zero voltage gain.
Excellent. This is probably going to be useful in my UHF rabbit trail, I'll start with the dedicated 555 buzz tester to sift a pile of Chinese mystery bulk stuff. And yes my DS212 ain't going to cut it. Myabe I can X-ray them.
Wow interesting info ! IMHO the ns Turn on / off time etc is important for pulse and digital signals Could a Gallium nitride (GaN) transistor be an interesting choice ? Nevertheless 2N2369 is an interesting one thanks for the info ! What could be complement to 2N2369 ? I found 2N4209 and 2N5771 but they have lower current ratings etc ? Take Care Marc
Great video Noel. I had a similar issue when building my Sixty Clone with transistor in Q2. I eventually found one with help from one of the component suppliers, but at the time I thought a lot about transistor replacement for these replica boards.
Different, i.e. reversed pinout on a transistor of the same type is not uncommon - it is intentional. See for example SMD-part BFS17A vs. BFS17AR (with reversed base/emitter pinouts; Vishay/Telefunken - datasheet). "R" means reversed. You have to pay attention.
I stopped watching the video at first, but then i coninued, Very interesting consept.. Transister for what you are using,, I have an old tube phono. I want to upgrade using transisters, to make it more stable.. I only need the phono inpust using s crystal pickup. i came to the conclusion either using a 23904 Transister or a BC547 Transister... I would avoid using the LM386 cause i dont like the way it sounds...However the tubes get really hot..
Many years ago while working on a Honeywell 200 that wouldn't multiply correctly I traced the problem to a shorted 2n2501 transistor (2501-7 in Honeywell terminology, one of several varieties used). I had a replacement but the leads were strangely bent so I re-bent them to the configuration I was accustomed to and soldered in the replacement. Now the thing wouldn't divide. Back to the same transistor which was now open. Closer examination revealed the reason for the strangely bent leads. Instead of EBC this version had the pinout ECB. So, yes, sometimes very similar transistors have different pinouts.
Maybe in the 2nd revision, they had a problem getting the clock inverter transistor as well, rather the use an alternate (which may be hard to find in the future as well), they decided to use an inverter from microcircuit. In aerospace and defense we have to deal with part obsolescence all the time. We often have to revist old designs and find alternate parts that are still available, it takes a design engineer and a component engineer to work together. The design engineer finds parts and analyzes/tests/simulates them to see how they will work as a replacement, while the component engineer determines where that part is in its own "life cycle", where it is made - China parts are out of the question, it's cost, and production lead times, among many other factors. Sometimes we have to do a "lifetime buy" and keep the parts in our own stock because the manufacturer has sent out notices that the part is going "obsolete" soon. Parts obsolescence is the least attractive part of engineering and many poor engineers palm it off to younger engineers, even if it was their own original design and know the circuitry better than anyone. I don't do that, I "own" my designs from concept to death and I insist that I handle the part obsolescence on my own designs... this gets it done faster and saves me from from doing it on others' designs. I get high marks for supporting ALL my prior designs no matter the age. Many engineers think it is beneath them to do such a "dubious" task and only want to work on new designs. That is sad, and some managers let them get away with that. The end result: the original design engineer gets a pass but the old design suffers in the end because someone not familiar with the design tries their best to fix it. This affects a company's reputation over time... but managers often worry only about short term solutions and to coddle some engineers that threaten to walk away if they are told to work on something not interesting. Me, I let them walk, I always find another to replace them, because the engineer I find is a "true" engineer and will not refuse to slove a problem.
NOELS RETRO LAB, can you make a video lesson about using a transistor biasing "inverse linear region" and why would you want to use this biasing and what applications would it work it?
Older books may have relevant information to restoring older electronics? Yeah that checks out to me and ironically maybe the best advice to people getting into this hobby that I've heard for a while.
Old now as I work through your videos. Most switching transistors have a Ton delay, Toff delay and storage time. High frequency transistors show the collector capacitance as well. The old CRT display base drive circuits for the line output stage were quite novel in trying to minimise storage time by reverse avalanching the base emitter junction to remove the storage charge as fast as they could. you would see a negative spike on the base of 15V.
I think that this is all wrong. The best way to replace a transistor, is to have somebody(who, I wonder...) do the dirty job and give us the best solution. By the way, if a group of hard-working old-schoolers is willing to gather all of this precious info in one place and make a table with them, according to each old application scenario and with all of the other components, would be even better! +1 by me as always, of course. EDIT: And I forgot to say that if the people who like to produce new boards for old computers consider putting some extra traces and holes for resistors or deringing capacitors etc, would be nice as well.
Hi, please forgive me for dropping in on this video's topic. I'm trying to repair an old QL with RAM, and or other, issues. Can I use an Analog oscilloscope to try to determine the failing chips? I cant afford a digital scope atm. I can afford some scopes offered second hand. So my question is, do I need a digital scope to try and educate myself and find some faulty chips on a digital, up to 8mhz cpu? Highest regards, Mac
If you only operate a transistor based on the model, you will miss the unusual application possibilities. In the 1980's I was wondering about a circuit diagram of a HIFI amplifier. To mute the sound, an NPN transistor was used the other way around. Collector to GND and emitter to the signal, base switched via a resistor. And so the NPN transistor was actually installed. Obviously the developer knew what he was doing and it works. Or how is it to use a transistor without any base at all? The transistor will turn off, right? But what happens when the voltage increases? Suddenly the collector emitter path becomes conductive without base current. A tiny current creeps from the collector to the base and when the voltage is high enough it turns on the transistor. Transistors can be operated in two areas. In the linear area, the tarnsistor worked up a sweat. Fully switched on it stays cooler but wants to stay there. Transistors are more complicated than stupid switches.
As you can see at the Oscilloscope the original Transistor does two things here: 1. stabilizing the square (so it is not overshooting etc). 2. amplifying it - as the gatearray does not output 5V (for whatever reason as well). So the transistor is just the cheapest version to implement it in a save way (to ensure that you always end up with a 5V square wave.
Playing background music while you're showing the experiments with the breadboard only made your words harder to follow and did not make the video better in any way. If I want to listen to chiptunes while watching videos, I can set that up myself, thank you very much.
I typically just find a compatible transistor by using the same Vbe and Ic specs. sometimes I'll try and stay within the same frequency limits. The hard part is finding the correct package type and pinout.
Hi Noel. I understand your method of trying out different transistors to replace an obsolete but working one but most of the time in the repair sector, the obsolete one has failed. Is the best way to find a suitable replacement by matching the hfe (gain) and VCEO as close as possible or higher ?. I am not talking about critical clock circuits but general purpose switching or amplification.
This seems to require more background than any of your other videos. I couldn't follow what was going on at all. What would you recommend as background for being able to understand this one? Would I have to get through the books on your Tools page first? :/
The switching time characteristics of BJT is a science on its own. It depends strongly on its manufacturing technology. Here is a very good video on it: ruclips.net/video/XgSKLsWAWGs/видео.html&si=EnSIkaIECMiOmarE . If you have a switching application you can either use a well specified switching transistor like the 2N2222 or an rf transistor. Then you can get sure to have a device with shorter delay and storage times, who strongly determine the switching behavior of the device.
OMG, why are they so complicated, i got no time for Science right know, and now after watching your video i got even more confused. Playing the Lottery is easier then finding the right Transistor : )
I *just* encountered this delay effect with a transistor circuit a few days ago. I solved it by replacing the transistor with a different part, but now I know I can add a capacitor or read TAoE. Thanks!
Big love for the deep dives here, never leave a problem unanswered, it's just too interesting having you track down the solution. Oh and your head planting the Art of Electronics book was a great shot!
Great video and I loved the details and your findings. Art of Electronics Edition 2 ,,, don't toss it out! And data sheets with Different pin outs!!! an Obstacle course. Congrats!
I'm more into microcontrollers etc than retro stuff (although I do have a great nostalgia for the 8-bit days)... so this video was really really useful... more useful than some of the electronics channels... thanks!
i almost never got a chance to replace a transistor. everything i see are full of square black boxes. except one power supply, where the small transistor blew after i changed the burned collector resistance. may be i mis-imagined or miscalculated the value of the burned resistance and it overloaded it. never actually took up that repair again. but, imo base should always have some resistance, since they are supposed to have small base current for small signal ones. and chip pins, even with passive pullups, often deliver more current. without a limiter, transistor may work fast, but they are destined to fail, like that one failed in the video. seems someone unknowingly made some fast & cheap solution & when it failed, brought out a new version with inverting chip.
Current is not multiploed. You cant get blood from a stone. Small changes in base current affect large changes in emitter current 5:34 .in other words you are using a small power supply current to affect a larger power supply at the rate on input change. Go on keep making it more complicated than it really is. However breadboard experiments are a great idea, providing you dont exceed yhe transistors rating.
Just to let you know a MOSFET is a voltage controlled device and a transistor is a current controlled device. I know you can have the issue of the charge staying on the gate of the MOSFET as well and unless it is rated for it operating it in it;s nonlinear region can damage it if you do not drive the gate with enough voltage. I think a logic level MOSFET would have fared better in the circuit you were trying to use it in.
Well explained, I only want to remark a few things: The problem is the charge strorage in the base due to the transistor being biased in saturation, the Miller effect is another different thing related to analog circuits, not digital ones like this "too" simple inverter. The CPC is a nice computer but its circuit design is quite poor, and at the end this transistor inverter was replaced by a logic gate as it should have been from the begining. (The other thing the Amstrad people never did was connecting the AY-3-8912 to the Z80 bus ;)
Indeed, the Miller effect is that the parasitic capacitance is much larger than expected. Noel explains very well why parasitic capacitance makes the transistor slow to turn off, but that doesn't explain such a low capacitance (a few pF) has such a significant effect. The effect is so significant because of the Miller effect. The Miller effect says that an inverting amplifier doesn't just amplify the signal, but also the parasitic capacitance between input and output, and this is why the parasitic capacitance is more more significant than expected. This explanation is missing in the video.
This is gold, love the deep dives as this usually, for me at least, makes things simpler. I know people always think abstractions are better for that, I guess they are when you want to learn fast, but I always struggled with that in school. Deep diving for sure made my homework go slow, but I understood stuff way better. :)
The Miller effect is a multiplication of collector-base CAPACITANCE by the amplifier gain... totally different than base charge of a saturated transistor. A Schottkey diode from base to collector will prevent saturation. This trick is employed in 74Sxx and 74LSxx logic chips.
That definitely answers the transistor issue. Great to see this explained. Never heard about the Miller effect. Thanks for this ‘down the rabbits hole’ video.
Hope he cooked the rabbit while he was there. :)
My first thought when I saw the slow turn off time was to clamp the Base to GND with a resistor, but then again I am not an electronics engineer.
This gets even crazier when you want to see devices' overload characteristics and stuff. I build guitar pedals and vacuum tube sims using discrete transistor circuits. There are so many undocumented "effects" that I've learned about and used. I can say that math is great for when you are trying to nail down a specific design that requires precision. The art of using them creatively is a whole different level and it has taken me over 10 years to learn about and document a whole slew of things that aren't documented.
Coffee in one hand, donut in the other and time to watch Noel's video.
ps. wow. that was an interesting "eye opening" video. I didn't know about this charge build up effect on base.
Glad you enjoyed it! I didn't know that either. I love it when I learn new stuff!
The Z80 clock signal spec is more demanding than regular TTL and I suppose they could not integrate the driver on the gate array. The inversion is just a side effect and the gate array could have inverted it for sure, but the primary use of the transistor is to improve the signal. I forgot the details, but it was quite common to use a transistor to match the Z80 clock spec. Note that they used a 74HC04, not a 74HCT04 in the second revision, which is a hint in that direction as well. MOSFETs are a completely different technology than bipolar and while you can get lucky, the considerations to drive them are very different.
I agree. I came to understand the Z80 clock input requirements designing a computer. Z80 is generally TTL level but clock is special.
16:22 The small spike at the output right before the signal falls is also caused by the Miller effect: The input rises at that moment, and the rise goes directly through the parasistic capacitor to the output. The Miller effect amplifies the spike. After a very short delay, the MOSFET starts conducting and the voltage drops to zero.
@Daniël Mantione I used to see you frequently on various Linux forums, not much lately, what happened?
@@airspeedmph Main reason is the retro computer renaissance, there is so much happening around the Commodore 64 and other retro computers these days that consumes a huge amount of hobby time. Some time ago SteamOS and gaming on Linux was brand new so that was where I was spending my spare time on. But currently... nothing less than wonders are happening on computers I love and I want to enjoy this.
I was noticing that spike and it looked severe enough that it could possibly re-trigger some circuitry at the wrong time
@@danielmantione Totally get that, I spend my time alternating the two (Linux and retro) although I have to admit that the retro part I'm only experiencing vicariously through channels like this, since my only HW that can be called retro is a PS1 :)
Lately I'm seriously consider getting real by acquiring some hardware, though I'm not sure where to begin. I'm somewhat inclined towards a Commodore 64, but a working one seem to be rather expensive.
But yeah, is crazy what's going on in the retro space, new hardware+addons and even games, is impressive and alluring. And another thing that I especially find relaxing in the retro space is that everything is governed/motivated by passion, there's no in-fights, politics and ideology to be found here.
Anyway, have fun.
I restore 2" broadcast video tape machines which are over 60 yrs old and work in the analog world. Compared to those beasts, digital is a piece of cake.
I have to find transistors that match in many parameters, some even have to be matched as well because they are used in a balanced circuit. And in the analog circuits they must have a freq response from 30hz to 6mhz.
When I was maintaining a complex digital device i thought as long as timing was correct and waveform was clean. I was good to go. Easy Peezy
Wow, I learned quite a bit here. Transistors are still something I get the basics of but not beyond, so this was really insightful.
Fantastic video, Noel! You're surpassing yourself. And I was delighted to see that you also had to battle the Bibl... err... the Sedra as an undergraduate. So did I! :-) Regarding the clock inverter, if the gate array uses the uninverted clock signal internally, it makes sense to invert it before passing it on to the CPU: the 180-degree phase shift helps account for, and even take advantage of, the propagation delay of the other signals shared with the CPU. The Atari line of 8-bit computers makes liberal use of this kind of phase shift to ensure proper synchronization between CPU and other active components, and it effectively doubles the speed of the system, since you get two clock tics per clock cycle, each in a different part of the system. Back in the day this kind of hack mattered. I also would be very careful about biasing any switching transistor without a base resistor. You got lucky (or perhaps you knew it) that the gate array had an internal pull-up. Otherwise the pin wouldn't be able to source the kind of current you get through a diode to ground, and the chip might have been ruined.
Thank you! Yes, the inverted clock signal makes a lot of sense. The thing that surprises me is that the Gate Array doesn't invert it before outputting it since nothing else on the board is using the uninverted one. I guess they either completely ran out of space in the GA or it was a "fix" they had to do afterwards.
@@NoelsRetroLab That was probably because of a limitation of the specific process technology used in the gate array, which may have prevented either enough buffering capacity (i.e. cannot drive or sink enough current) or fast-enough switching at the required current. It surely makes sense. An external transistor can use very different process technology than logic and can have larger die space than what could be done in the gate array.
My analogue electronics lecturers championed Sedra & Smith, and I was heading towards failing the course...
Then one of my classmates showed me his copy of The Art Of Electronics, and suddenly everything was alright. =)
Transistor in the clock circuit is to adjust the high voltage level to what Z80 needs.
👍 👍 👍 👍 👍 👍 👍 🍀 Great video! And very usefull for geeks who like to repair old devices... like me! 😜
5 V digital is easy !
NPN is 2N3904 , PNP is 2N3906 just check Ic max ….........
If the gate-array is open collector you need a pull up resistor on the transistor base .
( 1k - 10 k across Vcc & base should work ) .
PS
The 555 has a dc pedestal of around 0.7 - 0.9 V so off / low is not 0 .
For that particular transistor I have no problem:
1. I used to work for the manufacturer (Ferranti)
2. I have many ZTX312's and many other ZTX series devices from 45-50 years ago! 😱😳
3. I have several Towers' transistor selector books that detail the critical specs of the transistor and also give equivalents from that time (mid '70's!). I often find that modern equivalents quoted are actually NOT and finding an equivalent of the original equivalent works much better. 😎
Nice explanation of a really good approach though. 👍👍👍
It’s this kind of content that sets your channel apart Noel, love it. Thanks for putting the time into this video.
Awesome video. Very interesting but also easy to follow those of us without a background in electronics. I'd love to see some similar content looking at different types of capacitors.
This is all switching stuff. Way easier than matching RF. The transistor is being used as a buffer. Not for current or voltage but to provide abstraction from capacitance. In most instances where it is a switching device Hfe is irrelevant. Frequency transition, Input/Output capacitance, t-rise, t-fall and t-storage are the parameters to look at. A 2n2222A would be unsuitable due to the storage time. A 2n3904 might be close enough but it has way more storage time than the ZTX. A FET is a poor choice for switching due to the insulated gate being a capacitor.
ON Semi: Producer of quality Semiconductors and Semi-correct datasheets
Those round bodied transistors are pretty common here in slovakia/czechia the local electronics manufacturer(Tesla) used that design a lot, even for OP, amps and some simple IC's
Looked at the datasheet for the ZTX312 and noticed that it specifies rise, fall and storage times - and the storage time is 13ns. The BC547 doesnt specify this, but a close cousin the 2N3904 does - 200ns! That explains why the speedup capacitor is needed.
Interestingly I can't seem to find a common TO-92 transistor that has good switching times...
You have to be careful substituting mosfets because then the switching threshold becomes an issue - you have to raise the gate above the drain voltage by a certain amount to get the mosfet to saturate, whereas with a BJT you only need 0.6v, a MOSFET can often be up to 10V.
Yeah, I was really puzzled by the lack of timing info on the BC547. I wasn't sure if it was because it was fast enough for anything or just not intended for anything that you cared about timings. I guess it's because the answer wasn't a near number you can put on a table (and it would look really bad w/o the capacitor).
Try a 2N2222(A). It's a switching transistor, so its speed and transition times are optimized for switching. It is available in TO18 hermetic metal case with leads, a TO92 plastic case as a PN2222A, and SMD. Works for anything switching related below 300MHz. Check its spec sheet. Available everywhere and cheap (Digikey, Mouser).
@@cjay2 I thought this too, but its storage time is still a rather large 225ns according to the Fairchild data sheet.
Such a great hands-on guide to this issue I have been struggling with. Much of this is relevant with fet-based audio compressors from days gone by. Because the characteristics of the fet are what determine the some of the 'attitude' of the compressor, it is important to know how to find a modern match with similar properties.
This now seems less daunting, thank Noel!
Awesome video! Thank you for taking the time and fully deep dive into subjects like this!
This was a really great and interesting piece of information, Noel! Well done!
3rd Edition: The Art of Electronics
2nd Edition: The Lost Art of Electronics
What sucks is when you are repairing motor drive circuits like an H bridge where one of the transistors is dead and made of unobtainium. Meanwhile a popular replacement is listed as close matching. In that case you get to rebias the circuit, or try to balance it. I fixed a 45 jukebox with a mobile player carriage. All I could do was one direction was normal speed the other was half speed. That is why I praise the use of common parts.
I wish all electronics lectures were this good and this polished! Excellent work Noel ... thank you! 👏👍
Wow. What a comprehensive video. Great information
its always fascinating to remember all computers run on frequencies, just like electrons protons and neutrons spinning inside atoms.
* 13:45 Huh, not quite a "perfect output", the falling edge is heavily non-monotonic which is pretty back for a clock signal ... it doesn't happen in the final circuit on the board, so not an issue but in that test setup it's quite messed up.
* As noted, the output stage from the gate array is probably an internal pull up and a nmos driving the signal down, providing a high output impedance when the output is '1' and a low impedance '0'. But then that means on the falling edge it should have plenty of capability to sink the charge left over in the base junction of the transistor to turn it off quickly (since in the board there is no base resistor). So I'm not entirely sure why it had an issue.
I think the not perfect output is because of the input signal itself of the 555 which is totally not square, that's why I didn't give it much thought.
Fun interesting stuff. I subscribed ☺️. I haven't practiced electronics in over 20 years. I'm just now getting back into it. Been building my Electronics Laboratory Bench. Going back to school this year. It's not about making money it's about having fun 😊!
Love to hear that! I also hadn't done electronics in many years (focused on the software side), so it's been a really fun journey getting back into it.
MOSFETS are voltage controlled, BJT are current controlled. BJT have a current gain or Beta (HFE). They are NOT interchangeable.
I'm afraid that's a gross over-simplification. In high-speed switching circuits, the current into and out of the gate capacitance of a MOSFET means that it cannot be considered as simply "voltage-controlled". Once you've added a current limiting resistor between input and gate and a pull-down resistor between gate and source, there won't be a lot of difference in the design between BJT and MOSFET.
That was a fantastic video! It had meat on it's bones and you explained it in a very clear manner.
Coincidentally I also had to find some modern replacement transistors for a board I've been working on.
I loved the experimentation, keep it up!
This was amazing. Your deep dive was a great story
I don't plan to use transistors, even I don't have plans of working with electronics at this level but, oh man, I enjoyed this video A LOT! Your videos are always full of high quality material, presented in a way that everyone can understand. Great work!
Brilliant, just the right depth. I always wondered how to determine the correct replacement transistor and this provides a great starting point. More like this please Noel!
I've really wondered about this topic myself. Great practical problem to explore this!
Very nice research on the transistors. Thank you for the effort showing the process!
As for the 2nd edition of AoE, I recently said "What I learned from technical books: Contrary to popular belief they in many cases do not get better with revisions. It's often worth keeping earlier ones around."
I'm glad I'm not the only one who found that. It's definitely worth it to keep older revisions.
The book 'Art of Electronics,' in Polish known as 'Sztuka elektroniki,' is excellent. It contains a lot of content and is easy to understand. I've been using this book throughout my entire high school, or as we say in Poland, 'W Technikum.' :)
This was fascinating. Thanks for this.
@Noel Do you have a link to the online simulator? I'd like to have a play with that!
There you go: tinyurl.com/y6vfkjbw I added it to the description as well.
Thank you for covering this topic. BC547 is just cheaply made transistor intended for low frequency amplifiers and similar applications. There are also specialized high frequency transistors available. A metal can package was actually pretty popular in the past. For example, KC507 transistor, a BC547 clone, was produced in metal can as well.
Never have I been more happy to have an OLDER version of a book. (My "Horowitz & Hill" is the Second Edition.) Thanks!
The follow up book to 3rd edition of "Art of Electronics", which is title "The Art of Electronics: The x-Chapters" has a section on the Miller effect. It can out in Jan 2020. You can read up on it at x.artofelectronics.net
a cynical view from the real world: 1. a difficult problem that takes a while to research and solve by an engineer. 2. the engineer documents it so that nobody has to re-do the effort. 3. nobody reads said document. 4. go to 1
I have a few editions of Towers equivalent guides, If you see these books they are very useful. EDIT: International Transistor Selector
Fantastic video, thanks a lot. I checked the AoE 3rd ed book and found the Miller effect in Sect. 2.4.5
Excellent work and thanks for sharing. I fully agree that older books are always good to have around. Was doing the same or similar experiments about a month ago.
I didn't watch the video yet, but I know it's going to be good and help me in my repairs. +1 Thumbs up !
There's usually several different transistors you can use as substitutes for most of the commonly used transistors, I've been pretty lucky with regards to transistor substitutions.
Years ago, when I still used transistors for prototypes, I considered that there were two types of BJT transistors: for amplification and for switching. Even when fast, the transistors for amplification were not suitable for switching (the 300MHz bandwidth for BC547 showed you that), what was needed was the switching time to be very low. I usually used 2N2222 for switching up to 1MHz and BC547 as amplifier. Never reversed them, because the results were bad, 2N2222 was not a good amplifier and BC547 was not a good switching transistor, (you saw it at 21:45).
By looking at the datasheet, can see if a transistor is meant for switching applications, because there are few key parameters like switching time from low to high, from high to low, etc. You should have showed that in the video, not many people know how to read a datasheet.
I think Noels point was mostly to demonstrate that you need to do something extra to get the advertised 300MHz and this is valid for any transistor: A speedup capacitor will speed up a 2N2222 just as it speeds up a BC547.
@@danielmantione No, those 300MHz are one thing, they are valid in amplification. Make a amplifier, a oscillator and you will reach tenths of MHz domain, maybe hundreds, depending of the quality of the component. And he had this on his BC547.
What he needed was the commutation speed that is give by the special construction of the transistor, (small capacities inside it, and others). And that was achieved with the transistor used in the original design without additional components.
One can design a circuit, but to be replicate on industrial scale, the circuit must behave the same independent of the tolerances of the components. Noel said that, to fiddle with the components values in the BOM is not something that one will want to do when buy the kit to build it and enjoy it.
The point he made (in my understanding) is that you must use the right component to achieve the desired goal. Only that he does not touched the subject I pointed out.
@@sebastian19745 GaAs FET's came out in the late 80s, often in a can w/ 4th gnd pin. Those things easily handled 600 MHz due to ultra low junction capacitance.
@@wiregold8930 Yes, but do you remember also theirs prices? One of these or a mosfet tetrode were quite expensive in 80s. Also, they were used mainly in RF, hence the extra ground pin.
I got 3 BJT (BF200 or 214 I think it was, not sure) that had a 4th pin for ground, they were more expensive than a regular BC107/108, I did a 3 stage antenna amplifier with them. And they were pretty common by then, they were used in many TV tuners.
I would have thought like the TO220 package [heatsink tab] on the MOSFET's where the Gate is pin 1 to the left with the body facing you [heatsink tab to the back] pin 2 is Drain [also heatsink tab] and pin 3 is Source, that those small ones [TO-92] would also have the Gate be pin 1 with the flat side facing you and Drain be the middle pin [2] as well??? That would make for physical observation pinout identification. But I digress because on BJT's the middle pin 2 is the Base which is the same function as the Gate on a MOSFET.
really really Nice, thank you for the great video!!
S9018 is a fast BJT with low base and collector capacitance, a so its probably ok for fast switching, and widely available.
In the past I once accidentally picked a BC308C PNP instead of a BC238C NPN for driving a ~20mA LED from an 74Nxx output. I wondered for several minutes why the LED lit up when the TTL output was low. Until I checked the printing again and noticed my mistake. But this wasn't just a PNP-NPN mistake, this also meant that the PNP had the C-E current running against its natural direction. The BC308C was still very happy and worked in other breadboard setups.
Ah yes. Good old electronics class. I think using a proper inverter would have made much more sense but they would obviously cost more. The revision was a wise route to take. The Miller effect is what deals with what is known as slew rate current ( the speed at which the amplifier can change it's output.)
Not quite. The slew rate is the maximum rate of change of output. In other words, it's the limit on rise time and fall time for edges in digital applications, and the frequency limit on large signal swings in analogue applications. In this circuit, the charge on the base capacitance is holding the transistor on (output low) in saturation for a period after the input signal goes low. This is effectively a propagation delay for the rising edge of the output, rather than a limit on its rise time.
The fix for it when the input signal goes low is to find a way of sucking the charge out of the capacitance that the base of the transistor sees. If you have sufficient current sink from the input signal, then a capacitor across the current limit resistor will attempt to pull the base below ground when the input goes low and remove the charge on the base capacitance more rapidly, as shown with the BC547. If you have not much current sink, but more current source available from the input, then a low value resistor across the base-emitter (a "pull-down" resistor) will help to reduce the delay.
@@RexxSchneider It's still the slew rate. The fix for that is irrelevant. Look up the Miller Effect.
@@GORF_EMPIRE Don't patronise me; I'm perfectly aware of the Miller effect which is caused by the voltage gain of an inverting amplifier effectively multiplying the collector-base capacitance. The issue here is not the slew rate, which is clearly displayed on the rising edge of the output at 10:23 and can be seen to be around 5V in 200ns which equates to 25V/μs. The problem illustrated on that scope image is the 400ns that the BC547 takes to come out of saturation when the base current is shut off. That is entirely due to the charge on the capacitance as seen by the base having no low impedance path to discharge other than through the base-emitter junction. There is no Miller effect on a transistor _in saturation_ because a saturated transistor has almost zero voltage gain.
@@RexxSchneider Sure whatever pal.
@@GORF_EMPIRE Very mature.
You can't just replace a BJT transistor with a FET transistor.
The BJT is a current-controlled device, while
FET is a voltage-controlled device.
Excellent. This is probably going to be useful in my UHF rabbit trail, I'll start with the dedicated 555 buzz tester to sift a pile of Chinese mystery bulk stuff. And yes my DS212 ain't going to cut it. Myabe I can X-ray them.
Wow interesting info !
IMHO the ns Turn on / off time etc is important for pulse and digital signals
Could a Gallium nitride (GaN) transistor be an interesting choice ?
Nevertheless 2N2369 is an interesting one thanks for the info !
What could be complement to 2N2369 ?
I found 2N4209 and 2N5771 but they have lower current ratings etc ?
Take Care
Marc
Great video Noel. I had a similar issue when building my Sixty Clone with transistor in Q2. I eventually found one with help from one of the component suppliers, but at the time I thought a lot about transistor replacement for these replica boards.
Different, i.e. reversed pinout on a transistor of the same type is not uncommon - it is intentional. See for example SMD-part BFS17A vs. BFS17AR (with reversed base/emitter pinouts; Vishay/Telefunken - datasheet). "R" means reversed. You have to pay attention.
I stopped watching the video at first, but then i coninued, Very interesting consept.. Transister for what you are using,, I have an old tube phono. I want to upgrade using transisters, to make it more stable.. I only need the phono inpust using s crystal pickup. i came to the conclusion either using a 23904 Transister or a BC547 Transister... I would avoid using the LM386 cause i dont like the way it sounds...However the tubes get really hot..
My takeaway from all of this is if you're not using an oscilloscope then you're driving blind.
Great video! Thanks for taking the time to follow this one right through and then share it with us. Very enlightening!
Many years ago while working on a Honeywell 200 that wouldn't multiply correctly I traced the problem to a shorted 2n2501 transistor (2501-7 in Honeywell terminology, one of several varieties used). I had a replacement but the leads were strangely bent so I re-bent them to the configuration I was accustomed to and soldered in the replacement. Now the thing wouldn't divide. Back to the same transistor which was now open. Closer examination revealed the reason for the strangely bent leads. Instead of EBC this version had the pinout ECB. So, yes, sometimes very similar transistors have different pinouts.
Wow this is an informative video! Could you use a waveform generator instead of building up the 555?
JFETs are completely symmetrical. MOSFETs not quite. BJTs not at all.
Maybe in the 2nd revision, they had a problem getting the clock inverter transistor as well, rather the use an alternate (which may be hard to find in the future as well), they decided to use an inverter from microcircuit. In aerospace and defense we have to deal with part obsolescence all the time. We often have to revist old designs and find alternate parts that are still available, it takes a design engineer and a component engineer to work together. The design engineer finds parts and analyzes/tests/simulates them to see how they will work as a replacement, while the component engineer determines where that part is in its own "life cycle", where it is made - China parts are out of the question, it's cost, and production lead times, among many other factors. Sometimes we have to do a "lifetime buy" and keep the parts in our own stock because the manufacturer has sent out notices that the part is going "obsolete" soon. Parts obsolescence is the least attractive part of engineering and many poor engineers palm it off to younger engineers, even if it was their own original design and know the circuitry better than anyone. I don't do that, I "own" my designs from concept to death and I insist that I handle the part obsolescence on my own designs... this gets it done faster and saves me from from doing it on others' designs. I get high marks for supporting ALL my prior designs no matter the age. Many engineers think it is beneath them to do such a "dubious" task and only want to work on new designs. That is sad, and some managers let them get away with that. The end result: the original design engineer gets a pass but the old design suffers in the end because someone not familiar with the design tries their best to fix it. This affects a company's reputation over time... but managers often worry only about short term solutions and to coddle some engineers that threaten to walk away if they are told to work on something not interesting. Me, I let them walk, I always find another to replace them, because the engineer I find is a "true" engineer and will not refuse to slove a problem.
Oh k? Why do I think a common 2n2222 would have been sufficient? Anyway, very good explanation. Well done. 👌🏻
NOELS RETRO LAB, can you make a video lesson about using a transistor biasing "inverse linear region" and why would you want to use this biasing and what applications would it work it?
Older books may have relevant information to restoring older electronics? Yeah that checks out to me and ironically maybe the best advice to people getting into this hobby that I've heard for a while.
Hooray! I
Old now as I work through your videos.
Most switching transistors have a Ton delay, Toff delay and storage time. High frequency transistors show the collector capacitance as well.
The old CRT display base drive circuits for the line output stage were quite novel in trying to minimise storage time by reverse avalanching the base emitter junction to remove the storage charge as fast as they could. you would see a negative spike on the base of 15V.
I think that this is all wrong. The best way to replace a transistor, is to have somebody(who, I wonder...) do the dirty job and give us the best solution.
By the way, if a group of hard-working old-schoolers is willing to gather all of this precious info in one place and make a table with them, according to each old application scenario and with all of the other components, would be even better!
+1 by me as always, of course.
EDIT: And I forgot to say that if the people who like to produce new boards for old computers consider putting some extra traces and holes for resistors or deringing capacitors etc, would be nice as well.
Hi, please forgive me for dropping in on this video's topic. I'm trying to repair an old QL with RAM, and or other, issues. Can I use an Analog oscilloscope to try to determine the failing chips? I cant afford a digital scope atm. I can afford some scopes offered second hand. So my question is, do I need a digital scope to try and educate myself and find some faulty chips on a digital, up to 8mhz cpu?
Highest regards, Mac
This is a great idea for a video. Would be nice when you do a "How To Find ... Replacement" part with diodes (f.e. z-diodes).
If you only operate a transistor based on the model, you will miss the unusual application possibilities.
In the 1980's I was wondering about a circuit diagram of a HIFI amplifier. To mute the sound, an NPN transistor was used the other way around.
Collector to GND and emitter to the signal, base switched via a resistor. And so the NPN transistor was actually installed.
Obviously the developer knew what he was doing and it works.
Or how is it to use a transistor without any base at all?
The transistor will turn off, right?
But what happens when the voltage increases?
Suddenly the collector emitter path becomes conductive without base current.
A tiny current creeps from the collector to the base and when the voltage is high enough it turns on the transistor.
Transistors can be operated in two areas. In the linear area, the tarnsistor worked up a sweat.
Fully switched on it stays cooler but wants to stay there.
Transistors are more complicated than stupid switches.
As you can see at the Oscilloscope the original Transistor does two things here:
1. stabilizing the square (so it is not overshooting etc).
2. amplifying it - as the gatearray does not output 5V (for whatever reason as well).
So the transistor is just the cheapest version to implement it in a save way (to ensure that you always end up with a 5V square wave.
This video is filling the gap left by the previous one! It's very nice of you to delve that deep in this subject!
Playing background music while you're showing the experiments with the breadboard only made your words harder to follow and did not make the video better in any way. If I want to listen to chiptunes while watching videos, I can set that up myself, thank you very much.
I typically just find a compatible transistor by using the same Vbe and Ic specs. sometimes I'll try and stay within the same frequency limits. The hard part is finding the correct package type and pinout.
You can actually buy 3Mhz 555 timers these days :)
Hi Noel. I understand your method of trying out different transistors to replace an obsolete but working one but most of the time in the repair sector, the obsolete one has failed.
Is the best way to find a suitable replacement by matching the hfe (gain) and VCEO as close as possible or higher ?. I am not talking about critical clock circuits but general purpose switching or amplification.
This seems to require more background than any of your other videos. I couldn't follow what was going on at all. What would you recommend as background for being able to understand this one? Would I have to get through the books on your Tools page first? :/
I scrolled past this, only barely knowing what a transistor is, and I still scrolled back up and clicked it.
The switching time characteristics of BJT is a science on its own. It depends strongly on its manufacturing technology. Here is a very good video on it: ruclips.net/video/XgSKLsWAWGs/видео.html&si=EnSIkaIECMiOmarE . If you have a switching application you can either use a well specified switching transistor like the 2N2222 or an rf transistor. Then you can get sure to have a device with shorter delay and storage times, who strongly determine the switching behavior of the device.
OMG, why are they so complicated, i got no time for Science right know, and now after watching your video i got even more confused.
Playing the Lottery is easier then finding the right Transistor : )
I *just* encountered this delay effect with a transistor circuit a few days ago. I solved it by replacing the transistor with a different part, but now I know I can add a capacitor or read TAoE. Thanks!
can you make the same video for zx spectrum's TZX 650 (aka tr4)
Big love for the deep dives here, never leave a problem unanswered, it's just too interesting having you track down the solution. Oh and your head planting the Art of Electronics book was a great shot!
Wow! I'm need this part knowlege.. (YT alghoritms are better)
you found the correct transistor by posting to youtube... you got it nearly immediately. so... it IS in fact, trivial.
Great video and I loved the details and your findings. Art of Electronics Edition 2 ,,, don't toss it out! And data sheets with Different pin outs!!! an Obstacle course. Congrats!
I'm more into microcontrollers etc than retro stuff (although I do have a great nostalgia for the 8-bit days)... so this video was really really useful... more useful than some of the electronics channels... thanks!
Glad you enjoyed it!
i almost never got a chance to replace a transistor.
everything i see are full of square black boxes. except one power supply, where the small transistor blew after i changed the burned collector resistance. may be i mis-imagined or miscalculated the value of the burned resistance and it overloaded it. never actually took up that repair again.
but, imo base should always have some resistance, since they are supposed to have small base current for small signal ones. and chip pins, even with passive pullups, often deliver more current. without a limiter, transistor may work fast, but they are destined to fail, like that one failed in the video.
seems someone unknowingly made some fast & cheap solution & when it failed, brought out a new version with inverting chip.
Current is not multiploed. You cant get blood from a stone. Small changes in base current affect large changes in emitter current 5:34 .in other words you are using a small power supply current to affect a larger power supply at the rate on input change. Go on keep making it more complicated than it really is. However breadboard experiments are a great idea, providing you dont exceed yhe transistors rating.
Hi!
Just to let you know a MOSFET is a voltage controlled device and a transistor is a current controlled device. I know you can have the issue of the charge staying on the gate of the MOSFET as well and unless it is rated for it operating it in it;s nonlinear region can damage it if you do not drive the gate with enough voltage. I think a logic level MOSFET would have fared better in the circuit you were trying to use it in.
2N7000 small mosfets are wery ESD sensitive, be warned
Transistors are only as complicated as you are. Basically they are switches used as amplifiers.
Well explained, I only want to remark a few things: The problem is the charge strorage in the base due to the transistor being biased in saturation, the Miller effect is another different thing related to analog circuits, not digital ones like this "too" simple inverter. The CPC is a nice computer but its circuit design is quite poor, and at the end this transistor inverter was replaced by a logic gate as it should have been from the begining. (The other thing the Amstrad people never did was connecting the AY-3-8912 to the Z80 bus ;)
Indeed, the Miller effect is that the parasitic capacitance is much larger than expected. Noel explains very well why parasitic capacitance makes the transistor slow to turn off, but that doesn't explain such a low capacitance (a few pF) has such a significant effect. The effect is so significant because of the Miller effect. The Miller effect says that an inverting amplifier doesn't just amplify the signal, but also the parasitic capacitance between input and output, and this is why the parasitic capacitance is more more significant than expected. This explanation is missing in the video.