Thanks for the highquality content. Very well done. Designing and buying the pcbs then running the real word frequency analysis is what sets this channel apart.
Love digikey. They are 3 hrs away from me so with standard mail I always get the part with in 2 days. I want to get back into PCB building as a hobby. This is an awesome channel
Folks using Kicad will find the calculator for trace impedance in the calculators section. It is under high speed -> transmission lines Coplanar wave guide with ground plane is a good topology to use. It seems to give reasonable results Also to protect against "ooops I connected the power backwards" you can add a diode. If I was doing it, I would also add a little LED and resistor to indicate power is on to remind me.
Fantastic content! Thank you for breaking down those two amplifiers and sharing your design and test measurements! Greatly appreciate you sharing it with the world!
I used Mini0circuits back in the 1980's for a mass spectrometer. Amazing price-performance and so simple! Getting a PCB right at the time was the hard part!
Very good process steps to get started on the RF path - RF is an endless learning journey. 3 other figure of merits to measure which I think are quite useful are Noise Figure, Return Loss and Two-Tone Intermodulation (IP3 or linearity). Noise figure can be measured easily and reasonably accurate without a noise source since it has some gain - gain method measurement. A few things, I think, that may impact the roll-off at >1 GHz: (1) Transition from microstrip to larger SMD pads increases capacitances. Some tricks arounds are to either use smaller SMD packages and/or use a stackup so that the microstrip lines are around the same size as the pads. (2) Also, FR4's Dk changes with frequency as it varies with frequency. If the 50R transmission line was designed for Dk at 100MHz, we can expect it to be less 50R at higher frequencies and eventually becoming more inductive. Cheers!
Hi, thanks for the ideas. I don't have 2 signal generators so I cannot do IP3. For noise I was a bit worried about the accuracy. Matching would have been an easy one, I forgot about that one!
Nice video Hans, thanks for sharing it. Did you have a look at the harmonics and BW distortion? I mean driving them up to their P1dB point often causes a lot of it, at least for DATV. No worries, I use them as well in my driver amplifiers ;).
No, I would like to do an IP3 measurement but that would require 2 signal generators, that is why I left out the distortion. Near 1dB compression distortion should be really really bad :-)
@@HansRosenberg74 I would be interested in one that could be used for measurements, from DC to X MHz. The kind that e.g. the manufacturer of a differential voltage probe might use. I'm interested in how they are designed and what limits their bandwidth etc.
I have that on my list actually. It is a little bit of a pain in the but to design one of these. You need to create a very high common mode rejection. When you want DC it becomes really unpleasant. I've made one in the past though, I think it had 5Mhz bandwidth. It had a lot of gain and I used it to monitor small supply currents in low voltage applications.
That is a tough one! You'd need a hot/cold standard (I've made those myself in the past but you need a calibrated spectrum analyzer to calibrate those and mine isn't).
6:21 - I followed the design part, but I am confused by what you are saying in the measurement part. You said the gain is 46dB. Where is that coming from? I don't see this figure in either the S66+ or the 74+ datasheets. I'm seeing the S66+ with a gain of 21.6-16.4 dB, and the 74+ with a gain of 25.4-13.4 dB, depending on frequency. Am I reading the datasheet wrong? Or misunderstanding something fundamental?
You almost got it ;-) If you add the 25.4 and the 21.6, you get very close to the 46, its actually 47, I rounded that a bit too enthusiastically :-). I'm sweeping over a wide frequency range so I have to take the max gain at 100MHz into account as well. Sorry for the confusion :-)
Nice explanation. Thanks. My question : With such high gain, isn't there a high risk of loopback and oscillation ? Do you have advices about shielding / boxing this circuit ?
the ground plane and characteristic lines should minimize transmission of signals, so also reception. For boxing, I guess you can get standard shield cans at farnell, but they're quite expensive. In 'real' amplifiers they have milled aluminium boxes, even more expensive. So I don't have an easy fix.
hahaha, I thought this was an RF shielding manufacturer or something, but it's actually mint tins :-D I had a colleague who was also always collecting cans and tins for that purpose :-)
@@Soupie62 Another thing to have is some copper clad PCB material and some tinsnips you don't care about. You can solder PCBs to other PCBs to make a wall. Something grounded standing up from the PCB can block RF taking the path backwards. It is easy to make it fit inside your housing.
Thank you for the video! Have been enjoying watching your works! BTW, can you also address on the tuning stuff (like cut or paste the copper) to address the process variation and harmonic control?
@@HansRosenberg74 Thank you for your reply! Let's say the power amplifier uses a low-pass network as its output matching network(OMN). And after the fabrication, the performance of the power amplifier can be off because of the variation of the components, and usually people need to tune the OMN to get better performance. To tune the OMN, one can cut the copper or add copper foil to modify the OMN if it is realized by transmission line, or try different values of cap or inductor if it is realized by lumped elements.
Aha, now I get it. I would improve the components in that case. You can get pretty detailed models for inductors and capacitors from all the main suppliers. Murata for instance has something called 'simsurfing'. If you use those model parameters for your simulation, you can get pretty accurate results. I recently simulated and measured a combination of 2 capacitors in parallel and the results were remarkably close! A great way to get a very good bias circuit is to use 2 or 3 relatively large inductor values in series. That creates a very wide single parallel resonance instead of the sawtooth pattern you get with 1n/10n/100nH for instance.
Thank you for informative video. I wander about your measurement setup - I assume calculations are for dual stage setup, but how do you arrive to +15dBm P1dB value? 74+ P1dB point is 18.3 dBm, are you using +15 figure just to stay below that? Also, how amplifiers are designed in the realm of +30 - +33 dBm? I don't see much of integrated amplifiers at this power level. Is it discrete transistors only? (yes, I have my HAM licence ;)
The datasheets specs it at 100Mhz, I'm measuring at 1GHz. And I'm afraid the first stage is already beginning to compress at the same time as well (I did not check exactly). Hittite has apparently got a few that do 32.5dBm: HMC453ST89E Found them on the digikey website.
Yes sir, upload more n more. I wish you can do a vedio on common mode noise, filter design n remadies. One of my projects i have an long wire connection from a PIR sensor, for some reason the ground is affected in my PCB. I wish you can share a guide for such RF noises being absorbed such wire and affecting the PCB
I will make a series on EMC / EMI at some point, this will address it. For now I suggest you add a common mode transformer on the pcb where the 2 long wires enter the pcb. If you have more than 2 wires and the signal is LF than I'd use ferrite beads in series with the signals where the signals enter the pcb.
MELF is a pain both for hand assembly and automated pick/place assembly. I wouldn't choose them for cost reasons alone. But sometimes they are the best choice, especially for pulse loads.
I agree, after having used them for the first time, I'm on the fence on them..... I did look up assembly, apparently there are special nozzles that should solve that problem, but I did not talk to an actual assembly company.... I worked a lot with 0805, but they can be upside down, that problem is solved with the MELF parts, but they really like rolling around :-D (I did not anticipate that one :-) )
@@HansRosenberg74 Yeah, pick and place machine nozzles have trouble catching them, detecting cameras have trouble recognizing them, and eventually a lot of them end up in the dumpster. For manual placement, they are also a pain in the ass...embly :)
One interesting parameter to check would be return loss with different coupling capacitors. I wonder what you are actually seeing on 1GHz, the lower gain might be caused by worsening RL.
Thanks for going over this in detail! I am wondering the spec says the transistor can amplify from DC to 3GHz, but the input DC-blocking capacitor would make it not feasible to work at DC. Is there a way to make it truly functional from DC up to 3GHz?
Making an ultra wide band blocking capacitor as you suggest is extremely difficult. When you start to parallel them, you are inevitably going to get resonances which damage the flat response that you want. You would have to build an amplifier without capacitors. This is possible, but tricky.
@@HansRosenberg74 Thanks! I guess the difficulty for RF amplifier without coupling capacitor is that the DC level of the input needs to be at the bias point of the RF transistor, which is not necessarily the DC level of the input signal?
Apologize for my rookie question..when apply DC at the output of the amplifier ...why it is not damaging it although now it has two current opposing each other at input and output? Out of curiosity hope you can explain for me to understand
the price is too expansive I'm in Morocco and the course's price need a 3 month of saving , however I'm happy with the free content on your channel thanks for your time
I would love to make it cheaper, but costs here are quite high in my country. Up to this point I've invested around 4-5000 euro's to set up this channel, website, studio etc etc. And since this is full time, I've also not earned anything for 4 months.. So I have to get back that investment. However, if you keep watching my videos, you'll get most of my knowledge eventually, it is just not organized perfectly like it would be in a course, so you'd have to go through more trouble to get everything.
@HansRosenberg74 well I checked your website via the link and there were no prices, just another form to fill in. The webpage deserves to be more professional imo, and prices visible without a hunt.
I think you should design an LC filter and make sure the capacitors and inductors can handle the power. Now for caps this is easy, just use NP0/COG caps which are big enough, just check the datasheet, 1206 may already do it. The inductors are a bit more nasty. I think you'll have to wind your own air core inductors with silvered copper wire to increase the Q. Then you need to tune these a bit by bending the wires a bit closer or further apart. Making the design is easy, check out this site: markimicrowave.com/technical-resources/tools/lc-filter-design-tool/
Hah, nice idea, but I have no top 10. Everybody is recommending the book by Pozar. It takes some study but everything basic is in there. It's called microwave engineering. The paperback is around 60 USD.
That was really hard, those devices are unconditionally stable ;-) I think you may be able to get them to 'yell at you' if you annoy them enough, but you'd really have to do your best to achieve that.
hi! How comes it that 100pf was better than 33nf at high frequencies in your amplifier? If i compare two caps of the same size but different capacitances (example: GRM1885C1H101GA01 vs GRM1885C1H332JA01, 100pf vs 3300pf), their impedance at 1GHz and up match almost exactly, according to the manufacturer. But your measurements suggest they don't. Is this real? Are your caps of the same size? I was under an impression that there is a misconception that small-value caps are better at higher frequencies because they have a higher self-resonant frequency. But really, the frequency is higher simply because the capacitance is higher, while the actual impedance at high frequencies is the same, and the impedance is what actually matters, not that it "behaves like an inductor". And the only real reason to use small-value caps is that you can have them in smaller size, which 1) has smaller esl, 2) allows one to pack the circuit tighter, reducing other parasitic inductances too.
Your reasoning is sound, I also don't know why there is a difference. Apparently around 1GHz or so, something goes wrong but I don't know what. I also don't know what manufacturer both parts are from. The strange thing is that at 3 GHz the difference is almost gone. So I don't know what goes on here....
For signals above 1GHz it's necessary to use a more accurate model of a "capacitor". The big manufacturers will have SPICE models available for their parts. All passive components degenerate into a network of RLCs at a high enough frequency.
RF stuff is quite foreign to me, but I like very much the beauty of the simplicity of this design! What are the potential applications of such design - output - driving an antenna or input - boosting a very weak input signal?
Amplifiers are used just about everywhere. So there are soo many applications for this. The examples you mentioned, plus a lot more. I'd ask chatgpt for a list of RF amplifier applications. This came out :-) Wireless Communication Systems Cellular phones (boosting signal strength) Wi-Fi and Bluetooth devices Satellite communication Broadcasting AM/FM radio broadcasting TV transmitters Shortwave radio stations Military and Aerospace Radar systems Electronic warfare (jamming and countermeasures) Secure military communication Medical Equipment MRI (Magnetic Resonance Imaging) machines Medical telemetry systems Wireless medical devices Test and Measurement Spectrum analyzers RF signal generators Oscilloscopes for RF signal analysis Microwave Communication Microwave links for long-distance communication Satellite ground stations Wireless Infrastructure Base stations for cellular networks Repeaters and signal boosters in telecommunications Internet of Things (IoT) Devices RF communication modules in smart devices Low-power communication for sensor networks Amateur Radio (Ham Radio) Boosting signal strength for long-distance communication Signal amplification for higher output power Defense Systems Missile guidance and radar tracking systems Communication between military units over long distances
It is not a problem, as long as they're clean OR have a distance to sensitive signals. Sometimes you need to use power polygons on your signal layer because you are drawing a lot of current and voltage drop must be minimized. Usually for core voltages of processors or FPGAs.
Simple: Place decoupling caps regularly that connect the power plane to the ground plane. If you have really large current peaks, you may want to include a big capacitor here and there. Place smaller ones next to high speed loads like digital chips.
Excellent work... keep 'em coming. Also ... can I keep cascading to get a really high power output signal ? and can you somehow convert the input / output measurements in Watts ? coz It makes more sense to me.
The output power will always be limited by the last amplifier stage. That has to drive the final power. About the dBm values. 0dBm equals 1mW. 10dBm = 10mW and 20dBm = 100mW. It follows 0.001 * 10^(Pdbm/10) in Watts. And when cascading, you have to make sure that the driving amplifier does not clip before the final amplifier (I may have done that wrong in my design ;-) They clip at the same time I think but I did not check that carefully)
I am curious why you didn't use an LDO on the PCB. I would expect a good deal of noise added to the signal when powering from any switching supply, let alone a bench power supply. I don't know of an inexpensive way to measure noise figure (other than luck on eBay), but I wonder if your gain and compression measurements might even change if you repeated them using a lithium battery? Actually, if you have a variable attenuator, you could do a noise comparison (not measurement) by sticking this on a GPS receiver. See how high you can dial the attenuator before the receiver stops acquiring a cold fix. Then repeat, but using a battery as the power source.
I have a very good linear bench supply :-) I'm not sure what the transfer or supply noise to amplifier noise is. The bias resistor does take care of some isolation but some will definitely get through. If you want to supply it from an unspecified supply it would be great to add a low noise ldo. Do you know the LT3045?? That is an amazing device!
Can you explain why the quality of the resistors matters? What would be the impact of using simple & cheap thick film 0804 or 0603 ones? (Also, I see 22 Ohm thin film in 0603 for $0.022ea qty10, that's pretty cheap, isn't it, or are these the wrong ones?)
Ah, now I see, you need to dissipate a lot of power, that's why the simple thin film ones don't work. To be honest, I felt like this video was too compressed and left out a lot of the finer points and background info. Great overall, but would have preferred 2x the length. Thanks!
Thanks for the feedback. Video speed is a big topic on youtube. Hard to know what is best. What is your background and age if I may ask? Trying to find out a bit who my audience is.
@@HansRosenberg74 Speed is fine, I can stop, replay, slow-down when needed. Maybe instead of "compressed" I should have said "high-level". I thought that for too many things you just said "I need an X of value Y" without going into any details about Y. You are obviously making some rule-of-thumb / ballpark calculations in your head to arrive at an approx value for Y but you are going too fast to say that out loud. For example, for the DC block caps you probably have some rules of thumb: under 100Mhz around YnF, between 100Mhz and 1Ghz about YpF, etc. Or the ferrite beads, you just said "I can also put a 1 Ohm resistor" without saying a couple of sentences about the reasoning or the expected effect. I just thought "huh, a resistor and an inductor are not even the same kind of component, what is he doing??". Yes, each of these points could turn into a 20 minute calculation to get the exact perfect value, that's not what I'm looking for, I'm looking for the 3-4 sentence thinking that is going on in your head. I'm a software engineer >50 that has always done electronics on the side, mostly digital stuff, now doing quite some RF, mostly ultra-low power, ultra-light stuff (
Ahaa, thanks for the clarification, that makes a lot of sense. Basically the 1 Ohm vs ferrite bead is an impedance thing. The higher the RF impedance, the more gain you'll get (up to a limit of course, at some point the biasing does not affect the gain anymore). That biasing is like a load on the output, the higher the impedance of that bias, the less impact it has, less load. I'm explaining those ferrite beads with that impedance graph, that should explain it. For the capacitors: Pick the smallest value that will accomodate your minimum frequency and then pick a good dielectric (NP0 or COG). Actually, both capacitors should theoretically give the same gain 22nF/100pF but there is a difference. I'm not sure what causes it, I also don't know the manufacturers anymore since I've had those in my smd bins for years.
You mention calculation for a microstrip, yet PCB signal layer has groud pour. How do you decide on plane clearance for it not to be a coplanar wave guide?
Hi, I just fill in the values in a coplanar impedance calculator on the internet. Then I change the distance, if the impedance is hardly affected, it is far away enough....
Actually, no, it does not matter a lot. It does not make a lot of difference. You could argue that co-planar has less radiated emissions since there is more ground closer to the signal trace, but it is a minimal advantage as far as I know.
Yes, those connectors are a gods blessing. I was looking for something like that for years! And they're cheap to. And can carry insane currents! They're standard drone connectors, for drone batteries. xt-30, xt-60, xt-90, xt-120 and I believe they also have xt-150. The number indicates the current. I've seen youtubers try to destroy these with overcurrent only to find out that the solderings let go and the connectors are still fine :-). DON'T BUY THE ALIEXPRESS OR ALIBABA VERSION, THEY HAVE NO CLAMPING FORCE!! Only buy the real deal: AMASS. I use the xt90 or xt120 for connecting my audio power amplifier output, fantastic!
hahaha, ever worked with 0201?? They'll never let go of your tweezers at all :-D. It drove me totally nuts to the point where I wanted to release some serious anger :-D
Well, partially, you'll need a tx/rx switch. When transmitting, this amp should be used, but when receiving it should be turned off and the signal should be routed around the amp. So not so easy. You don' t have access to the tx rx signal
It's always tricky. You might get a parallel resonance with the inductance of the 22nF and the capacitance of the 100pF (I'd have to calculate how that turns out). Maybe it is not so bad here since both have a pretty low inductance.... however, the 22nF will turn inductive much sooner, so that might be a problem. When I was working on the Westerbork Telescope, I needed a very wideband (I believe 4kHz to 20MHz) coupling cap in a number of locations in the signal path so I had a tantalum elco with a 1206 capacitor. It made a really nice dip in the frequency response. Off course a tantalum cap is a disaster when it comes to inductance and this was my first year as a professional electronics engineer, so I had to find that out the hard way. I found out that putting multiple components of the same value in parallel works really well to make such a component more broadband. I plan to do some research on that for my channel and my course. I also would like to measure this so I have some real data. I did something similar last week for a very wideband RF bias tee and that worked out really nice using multiple of the same inductors in series.
@@HansRosenberg74 Hi Hans, you are correct about the added inductances. In simulations you can verify them well. In practice is is hard todo. In case of power pin decoupling one will often add more capacitors in parallel with different values and distance to the pin. See many application notes of FPGA's. With Ansys PI you can even tune this on PCB with advanced simulations.
I want to do some experiments with that at some point. I have a feeling there might be a good multiplication value for selecting the parallel components assuming their parasitic inductance is relatively equal. Should be easy to do parametrically in a simulator. If that gives some nice results I'll also put that in my course, I have a whole bunch of power supply filtering tactics I employ which would be very nice to document.
Hoi Klaas, wat leuk, ik ben heel slecht in namen en gezichten onthouden, wat is je achternaam en wat deed je bij Ericsson, dan kan ik het plaatje wel weer maken. Groetjes, Hans
Ah, dat dacht ik al. Ik moest er even over nadenken. Je had in bril en een snorretje, kon het 10 min na je berichtje weer voor de geest halen, Man dat is bijna 20 jaar geleden! Ben je al met pensioen? Waar heb je na SonyEricsson gezeten?
The datasheet for these seems very sparse. For starters, it seems like the voltage in the electrical specifications table doesn't match what's in the bias resistor section? The specifications table (for the 66+) says "Device operating voltage 3.0-3.5-4.0 V" (min, typ, max). But the bias section seems to require at least 7V. What's going on with this? Can the device be run off 3.3V, or does it really need 7 to operate?
great job can you pls make a video about how to measure and design wifi and ble pcb antennas with the nanovna or librevna and how to make sure that my antenna will work properly like i expecting
yes, I noticed, this was my first experience with MELF........ I'm on the fence now what is best. The 0805 components have the drawback that they're upside down half of the time (also annoying) but they do have the great advantage you can stack them when experimenting, which is tricky with a MELF........ so not sure what to do with my next power amplifier yet.....
Hi, thanks for the feedback. You are literally the first to mention this! This 'high information density' is a very conscious decision. I've had many comments of viewers who really appreciate that. According to research, talking fast is an advantage on youtube, Apparently the majority of viewers do not want to waste any time getting their information. What you could do is play the video at half of trhee-quarter speed. I had one viewer that was soo happy with no BS high speed approach, he said that he normally played clips at double speed but he had to slow mine down to normal speed :-D Best regards, Hans
![Gain block RF Amplifiers - Theory and Design [1/2]](http://i.ytimg.com/vi/8myFI-tmowo/mqdefault.jpg)
Love the straight to business videos, not bs or intro just facts
haha, that was the intention :-D
Thanks for the highquality content. Very well done. Designing and buying the pcbs then running the real word frequency analysis is what sets this channel apart.
Thanks for the thumbs up!
Love digikey. They are 3 hrs away from me so with standard mail I always get the part with in 2 days. I want to get back into PCB building as a hobby. This is an awesome channel
Thanks!
Folks using Kicad will find the calculator for trace impedance in the calculators section.
It is under high speed -> transmission lines
Coplanar wave guide with ground plane is a good topology to use.
It seems to give reasonable results
Also to protect against "ooops I connected the power backwards" you can add a diode.
If I was doing it, I would also add a little LED and resistor to indicate power is on to remind me.
The led is a nice idea indeed.
The Saturn PCB Toolkit is also a great calculator. But before I fabricated anything, I would get the actual Dk values from the PCB house.
That was super interesting. Thanks so much for teaching.
thanks
Fantastic content! Thank you for breaking down those two amplifiers and sharing your design and test measurements! Greatly appreciate you sharing it with the world!
You're welcome
I used Mini0circuits back in the 1980's for a mass spectrometer. Amazing price-performance and so simple! Getting a PCB right at the time was the hard part!
yeah that became a lot cheaper and easier these days :-)
Thank you so much 😊
You're welcome!
thanks homie. Learned more and faster about making an RF amp than hours spent googling stuff lol
thanks!
thanks hans interesting stuff leaning more every time i watch ❤
thanks!
Very good process steps to get started on the RF path - RF is an endless learning journey.
3 other figure of merits to measure which I think are quite useful are Noise Figure, Return Loss and Two-Tone Intermodulation (IP3 or linearity). Noise figure can be measured easily and reasonably accurate without a noise source since it has some gain - gain method measurement.
A few things, I think, that may impact the roll-off at >1 GHz: (1) Transition from microstrip to larger SMD pads increases capacitances. Some tricks arounds are to either use smaller SMD packages and/or use a stackup so that the microstrip lines are around the same size as the pads. (2) Also, FR4's Dk changes with frequency as it varies with frequency. If the 50R transmission line was designed for Dk at 100MHz, we can expect it to be less 50R at higher frequencies and eventually becoming more inductive.
Cheers!
Hi, thanks for the ideas. I don't have 2 signal generators so I cannot do IP3. For noise I was a bit worried about the accuracy. Matching would have been an easy one, I forgot about that one!
Theoretical question: can the same circuit diagram be used with KT805b transistors?
Great content
Thanks!
Nice video Hans, thanks for sharing it. Did you have a look at the harmonics and BW distortion? I mean driving them up to their P1dB point often causes a lot of it, at least for DATV. No worries, I use them as well in my driver amplifiers ;).
No, I would like to do an IP3 measurement but that would require 2 signal generators, that is why I left out the distortion. Near 1dB compression distortion should be really really bad :-)
😊👍 Thank you.
You're welcome 😊
Hello, thanks for the content. I would also be interested in a guide for differential amplifiers.
what kind of differential amplifiers exactly? Just a differential pair, or some differential opamp amplifier?
@@HansRosenberg74 I would be interested in one that could be used for measurements, from DC to X MHz. The kind that e.g. the manufacturer of a differential voltage probe might use. I'm interested in how they are designed and what limits their bandwidth etc.
I have that on my list actually. It is a little bit of a pain in the but to design one of these. You need to create a very high common mode rejection. When you want DC it becomes really unpleasant. I've made one in the past though, I think it had 5Mhz bandwidth. It had a lot of gain and I used it to monitor small supply currents in low voltage applications.
Great content on your channel! Did you try a measurement of noise figure to see how close that was to spec?
That is a tough one! You'd need a hot/cold standard (I've made those myself in the past but you need a calibrated spectrum analyzer to calibrate those and mine isn't).
6:21 - I followed the design part, but I am confused by what you are saying in the measurement part. You said the gain is 46dB. Where is that coming from? I don't see this figure in either the S66+ or the 74+ datasheets. I'm seeing the S66+ with a gain of 21.6-16.4 dB, and the 74+ with a gain of 25.4-13.4 dB, depending on frequency. Am I reading the datasheet wrong? Or misunderstanding something fundamental?
You almost got it ;-) If you add the 25.4 and the 21.6, you get very close to the 46, its actually 47, I rounded that a bit too enthusiastically :-). I'm sweeping over a wide frequency range so I have to take the max gain at 100MHz into account as well. Sorry for the confusion :-)
Nice explanation. Thanks.
My question : With such high gain, isn't there a high risk of loopback and oscillation ?
Do you have advices about shielding / boxing this circuit ?
the ground plane and characteristic lines should minimize transmission of signals, so also reception. For boxing, I guess you can get standard shield cans at farnell, but they're quite expensive. In 'real' amplifiers they have milled aluminium boxes, even more expensive. So I don't have an easy fix.
When it comes to shielding of (small) RF circuits, I use Eclipse Mint tins. 2.2 x 8.5 x 4.7 cm.
hahaha, I thought this was an RF shielding manufacturer or something, but it's actually mint tins :-D
I had a colleague who was also always collecting cans and tins for that purpose :-)
@@Soupie62 Another thing to have is some copper clad PCB material and some tinsnips you don't care about. You can solder PCBs to other PCBs to make a wall. Something grounded standing up from the PCB can block RF taking the path backwards. It is easy to make it fit inside your housing.
Is it microstrip, or coplanar transmission line?
Technically, it's a coplanar, but the ground on the top is soo far away compared to the ground plane that it behaves as a microstrip.....
simplicity of this design is impressive.great job thanks sir.
do we need impedance controlled traces in this design?
they're quite short, so probably you can get away with the wrong impedance, but the traces are very close to 50 ohms
Thank you for the video! Have been enjoying watching your works! BTW, can you also address on the tuning stuff (like cut or paste the copper) to address the process variation and harmonic control?
Thanks for the thumbs up! I'm not sure what you mean by cut or paste copper, can you clarify?
@@HansRosenberg74 Thank you for your reply! Let's say the power amplifier uses a low-pass network as its output matching network(OMN). And after the fabrication, the performance of the power amplifier can be off because of the variation of the components, and usually people need to tune the OMN to get better performance. To tune the OMN, one can cut the copper or add copper foil to modify the OMN if it is realized by transmission line, or try different values of cap or inductor if it is realized by lumped elements.
Aha, now I get it. I would improve the components in that case. You can get pretty detailed models for inductors and capacitors from all the main suppliers. Murata for instance has something called 'simsurfing'. If you use those model parameters for your simulation, you can get pretty accurate results. I recently simulated and measured a combination of 2 capacitors in parallel and the results were remarkably close! A great way to get a very good bias circuit is to use 2 or 3 relatively large inductor values in series. That creates a very wide single parallel resonance instead of the sawtooth pattern you get with 1n/10n/100nH for instance.
@@HansRosenberg74 Thank you! Very informative!
Thank you for informative video. I wander about your measurement setup - I assume calculations are for dual stage setup, but how do you arrive to +15dBm P1dB value? 74+ P1dB point is 18.3 dBm, are you using +15 figure just to stay below that?
Also, how amplifiers are designed in the realm of +30 - +33 dBm? I don't see much of integrated amplifiers at this power level. Is it discrete transistors only? (yes, I have my HAM licence ;)
The datasheets specs it at 100Mhz, I'm measuring at 1GHz. And I'm afraid the first stage is already beginning to compress at the same time as well (I did not check exactly).
Hittite has apparently got a few that do 32.5dBm: HMC453ST89E Found them on the digikey website.
@@HansRosenberg74 thanks!
Yes sir, upload more n more.
I wish you can do a vedio on common mode noise, filter design n remadies.
One of my projects i have an long wire connection from a PIR sensor, for some reason the ground is affected in my PCB.
I wish you can share a guide for such RF noises being absorbed such wire and affecting the PCB
I will make a series on EMC / EMI at some point, this will address it. For now I suggest you add a common mode transformer on the pcb where the 2 long wires enter the pcb. If you have more than 2 wires and the signal is LF than I'd use ferrite beads in series with the signals where the signals enter the pcb.
check out Fesz Electronics channel
MELF is a pain both for hand assembly and automated pick/place assembly. I wouldn't choose them for cost reasons alone. But sometimes they are the best choice, especially for pulse loads.
I agree, after having used them for the first time, I'm on the fence on them..... I did look up assembly, apparently there are special nozzles that should solve that problem, but I did not talk to an actual assembly company.... I worked a lot with 0805, but they can be upside down, that problem is solved with the MELF parts, but they really like rolling around :-D (I did not anticipate that one :-) )
@@HansRosenberg74Even with the special nozzles they can be problematic. They can roll off the paste in between placement and reflow.
@@HansRosenberg74 Yeah, pick and place machine nozzles have trouble catching them, detecting cameras have trouble recognizing them, and eventually a lot of them end up in the dumpster. For manual placement, they are also a pain in the ass...embly :)
Ok, you convinced me.......... Darn, expensive 0805 thin film resistors it is then......
I also noticed they were quite 'mobile' even when using rework flux. So it's really not a nice idea....
One interesting parameter to check would be return loss with different coupling capacitors. I wonder what you are actually seeing on 1GHz, the lower gain might be caused by worsening RL.
Based on past experience, I don't think the capacitors are the problem. The bandwidth of the amplifier is just not high enough.
Thanks for going over this in detail! I am wondering the spec says the transistor can amplify from DC to 3GHz, but the input DC-blocking capacitor would make it not feasible to work at DC. Is there a way to make it truly functional from DC up to 3GHz?
Making an ultra wide band blocking capacitor as you suggest is extremely difficult. When you start to parallel them, you are inevitably going to get resonances which damage the flat response that you want. You would have to build an amplifier without capacitors. This is possible, but tricky.
@@HansRosenberg74 Thanks! I guess the difficulty for RF amplifier without coupling capacitor is that the DC level of the input needs to be at the bias point of the RF transistor, which is not necessarily the DC level of the input signal?
@bluesunflower2007 yes!
Apologize for my rookie question..when apply DC at the output of the amplifier ...why it is not damaging it although now it has two current opposing each other at input and output? Out of curiosity hope you can explain for me to understand
The amplifier gets its power through the output pin. Hope that clears it up?
The inductor in the output separates the DC side from the RF side and allows current to sink through the device when/as needed.
the price is too expansive I'm in Morocco and the course's price need a 3 month of saving , however I'm happy with the free content on your channel thanks for your time
It's not that bad, I'm from Argentina and I would need to save my entire salary for about 8 months to get that money 😢
I would love to make it cheaper, but costs here are quite high in my country. Up to this point I've invested around 4-5000 euro's to set up this channel, website, studio etc etc. And since this is full time, I've also not earned anything for 4 months.. So I have to get back that investment. However, if you keep watching my videos, you'll get most of my knowledge eventually, it is just not organized perfectly like it would be in a course, so you'd have to go through more trouble to get everything.
@HansRosenberg74 well I checked your website via the link and there were no prices, just another form to fill in.
The webpage deserves to be more professional imo, and prices visible without a hunt.
hello sir ,
is it possible do one video on power amplifier ?
Will try
I really hope you can make a simple filter for a 78 - 108 MHz FM transmitter, with RF power from 5 - 15W. Thanks a lot!
I think you should design an LC filter and make sure the capacitors and inductors can handle the power. Now for caps this is easy, just use NP0/COG caps which are big enough, just check the datasheet, 1206 may already do it. The inductors are a bit more nasty. I think you'll have to wind your own air core inductors with silvered copper wire to increase the Q. Then you need to tune these a bit by bending the wires a bit closer or further apart. Making the design is easy, check out this site: markimicrowave.com/technical-resources/tools/lc-filter-design-tool/
Could you make a video about Top 10 essential books for RF electronics?
Hah, nice idea, but I have no top 10. Everybody is recommending the book by Pozar. It takes some study but everything basic is in there. It's called microwave engineering. The paperback is around 60 USD.
How is input and output stability accounted for in this design?
That was really hard, those devices are unconditionally stable ;-) I think you may be able to get them to 'yell at you' if you annoy them enough, but you'd really have to do your best to achieve that.
any thoughts on smaller mmcx connectors for hobbyists?
I never used those.... so I can't comment. I do use smb's a lot, easy to connect and readily available.
hi! How comes it that 100pf was better than 33nf at high frequencies in your amplifier? If i compare two caps of the same size but different capacitances (example: GRM1885C1H101GA01 vs GRM1885C1H332JA01, 100pf vs 3300pf), their impedance at 1GHz and up match almost exactly, according to the manufacturer. But your measurements suggest they don't. Is this real? Are your caps of the same size?
I was under an impression that there is a misconception that small-value caps are better at higher frequencies because they have a higher self-resonant frequency. But really, the frequency is higher simply because the capacitance is higher, while the actual impedance at high frequencies is the same, and the impedance is what actually matters, not that it "behaves like an inductor". And the only real reason to use small-value caps is that you can have them in smaller size, which 1) has smaller esl, 2) allows one to pack the circuit tighter, reducing other parasitic inductances too.
Your reasoning is sound, I also don't know why there is a difference. Apparently around 1GHz or so, something goes wrong but I don't know what. I also don't know what manufacturer both parts are from. The strange thing is that at 3 GHz the difference is almost gone. So I don't know what goes on here....
For signals above 1GHz it's necessary to use a more accurate model of a "capacitor". The big manufacturers will have SPICE models available for their parts. All passive components degenerate into a network of RLCs at a high enough frequency.
yes, would be interesting to look at the murata models for a 22nF and a 100pF and see if that explains it....
RF stuff is quite foreign to me, but I like very much the beauty of the simplicity of this design! What are the potential applications of such design - output - driving an antenna or input - boosting a very weak input signal?
Amplifiers are used just about everywhere. So there are soo many applications for this. The examples you mentioned, plus a lot more. I'd ask chatgpt for a list of RF amplifier applications. This came out :-)
Wireless Communication Systems
Cellular phones (boosting signal strength)
Wi-Fi and Bluetooth devices
Satellite communication
Broadcasting
AM/FM radio broadcasting
TV transmitters
Shortwave radio stations
Military and Aerospace
Radar systems
Electronic warfare (jamming and countermeasures)
Secure military communication
Medical Equipment
MRI (Magnetic Resonance Imaging) machines
Medical telemetry systems
Wireless medical devices
Test and Measurement
Spectrum analyzers
RF signal generators
Oscilloscopes for RF signal analysis
Microwave Communication
Microwave links for long-distance communication
Satellite ground stations
Wireless Infrastructure
Base stations for cellular networks
Repeaters and signal boosters in telecommunications
Internet of Things (IoT) Devices
RF communication modules in smart devices
Low-power communication for sensor networks
Amateur Radio (Ham Radio)
Boosting signal strength for long-distance communication
Signal amplification for higher output power
Defense Systems
Missile guidance and radar tracking systems
Communication between military units over long distances
Nice presentation. Why not the innotion yg602020?
Yes, looks like a nice one!
What are your views on power pours on signal layer?
It is not a problem, as long as they're clean OR have a distance to sensitive signals. Sometimes you need to use power polygons on your signal layer because you are drawing a lot of current and voltage drop must be minimized. Usually for core voltages of processors or FPGAs.
@@HansRosenberg74 is there some good rules to follow here to keep the power "clean "?
Simple: Place decoupling caps regularly that connect the power plane to the ground plane. If you have really large current peaks, you may want to include a big capacitor here and there. Place smaller ones next to high speed loads like digital chips.
Excellent work... keep 'em coming.
Also ... can I keep cascading to get a really high power output signal ?
and can you somehow convert the input / output measurements in Watts ? coz It makes more sense to me.
The output power will always be limited by the last amplifier stage. That has to drive the final power. About the dBm values. 0dBm equals 1mW. 10dBm = 10mW and 20dBm = 100mW. It follows 0.001 * 10^(Pdbm/10) in Watts. And when cascading, you have to make sure that the driving amplifier does not clip before the final amplifier (I may have done that wrong in my design ;-) They clip at the same time I think but I did not check that carefully)
I am curious why you didn't use an LDO on the PCB. I would expect a good deal of noise added to the signal when powering from any switching supply, let alone a bench power supply. I don't know of an inexpensive way to measure noise figure (other than luck on eBay), but I wonder if your gain and compression measurements might even change if you repeated them using a lithium battery?
Actually, if you have a variable attenuator, you could do a noise comparison (not measurement) by sticking this on a GPS receiver. See how high you can dial the attenuator before the receiver stops acquiring a cold fix. Then repeat, but using a battery as the power source.
I have a very good linear bench supply :-) I'm not sure what the transfer or supply noise to amplifier noise is. The bias resistor does take care of some isolation but some will definitely get through. If you want to supply it from an unspecified supply it would be great to add a low noise ldo. Do you know the LT3045?? That is an amazing device!
Can you explain why the quality of the resistors matters? What would be the impact of using simple & cheap thick film 0804 or 0603 ones? (Also, I see 22 Ohm thin film in 0603 for $0.022ea qty10, that's pretty cheap, isn't it, or are these the wrong ones?)
Ah, now I see, you need to dissipate a lot of power, that's why the simple thin film ones don't work. To be honest, I felt like this video was too compressed and left out a lot of the finer points and background info. Great overall, but would have preferred 2x the length. Thanks!
Thanks for the feedback. Video speed is a big topic on youtube. Hard to know what is best. What is your background and age if I may ask? Trying to find out a bit who my audience is.
@@HansRosenberg74 Speed is fine, I can stop, replay, slow-down when needed. Maybe instead of "compressed" I should have said "high-level". I thought that for too many things you just said "I need an X of value Y" without going into any details about Y. You are obviously making some rule-of-thumb / ballpark calculations in your head to arrive at an approx value for Y but you are going too fast to say that out loud. For example, for the DC block caps you probably have some rules of thumb: under 100Mhz around YnF, between 100Mhz and 1Ghz about YpF, etc. Or the ferrite beads, you just said "I can also put a 1 Ohm resistor" without saying a couple of sentences about the reasoning or the expected effect. I just thought "huh, a resistor and an inductor are not even the same kind of component, what is he doing??". Yes, each of these points could turn into a 20 minute calculation to get the exact perfect value, that's not what I'm looking for, I'm looking for the 3-4 sentence thinking that is going on in your head.
I'm a software engineer >50 that has always done electronics on the side, mostly digital stuff, now doing quite some RF, mostly ultra-low power, ultra-light stuff (
Ahaa, thanks for the clarification, that makes a lot of sense. Basically the 1 Ohm vs ferrite bead is an impedance thing. The higher the RF impedance, the more gain you'll get (up to a limit of course, at some point the biasing does not affect the gain anymore). That biasing is like a load on the output, the higher the impedance of that bias, the less impact it has, less load. I'm explaining those ferrite beads with that impedance graph, that should explain it. For the capacitors: Pick the smallest value that will accomodate your minimum frequency and then pick a good dielectric (NP0 or COG). Actually, both capacitors should theoretically give the same gain 22nF/100pF but there is a difference. I'm not sure what causes it, I also don't know the manufacturers anymore since I've had those in my smd bins for years.
You mention calculation for a microstrip, yet PCB signal layer has groud pour.
How do you decide on plane clearance for it not to be a coplanar wave guide?
Hi, I just fill in the values in a coplanar impedance calculator on the internet. Then I change the distance, if the impedance is hardly affected, it is far away enough....
@@HansRosenberg74 Thanks for reply. Got it, but am I missing something and microstrip is somehow superior here?
Actually, no, it does not matter a lot. It does not make a lot of difference. You could argue that co-planar has less radiated emissions since there is more ground closer to the signal trace, but it is a minimal advantage as far as I know.
Hey Hans, thanks for the great video! I learned a lot here. By the way: What are these yellow power connectors called? These look neat.
Yes, those connectors are a gods blessing. I was looking for something like that for years! And they're cheap to. And can carry insane currents! They're standard drone connectors, for drone batteries. xt-30, xt-60, xt-90, xt-120 and I believe they also have xt-150. The number indicates the current. I've seen youtubers try to destroy these with overcurrent only to find out that the solderings let go and the connectors are still fine :-). DON'T BUY THE ALIEXPRESS OR ALIBABA VERSION, THEY HAVE NO CLAMPING FORCE!! Only buy the real deal: AMASS. I use the xt90 or xt120 for connecting my audio power amplifier output, fantastic!
Drone shops will have these off the shelf, max 2 usd for a pair (male female) I believe
@@HansRosenberg74 Awesome, thank you so much! Ordered some of these right away! :)
If you solder them, do it with 2 connectors attached to each other, so the contacts don't move due to melting plastic.
Are these PCB schematics available to download/purchase?
Hi, no, this is part of a larger testboard designed with eagle, however it is a piece of cake to make this board in kicad if you want to try that.
I'd forgotten what MELF stands for, so I looked it up: "Most End up Lying on the Floor"
hahahahaha
Can confirm that :D
0402 (US sizes) resistors evaporate into thin air off my tweezers.
hahaha, ever worked with 0201?? They'll never let go of your tweezers at all :-D. It drove me totally nuts to the point where I wanted to release some serious anger :-D
@@HansRosenberg74 0201's are so tiny, they only exist in the quantum realm.
Can I use this amplifier for a wifi antenna that receives AND sends data?
Well, partially, you'll need a tx/rx switch. When transmitting, this amp should be used, but when receiving it should be turned off and the signal should be routed around the amp. So not so easy. You don' t have access to the tx rx signal
That was a great hint! I'm using an ESP32 and there is indeed a "WiFi Multiple Antennas" section in the documentation 👍😁
Thank you!
Course price is 'ex-vat'. How much % is vat?
I think that may depend on your country. Here we have 21%
What about 22nF parallel to 100pF in de series capacitance?
It's always tricky. You might get a parallel resonance with the inductance of the 22nF and the capacitance of the 100pF (I'd have to calculate how that turns out). Maybe it is not so bad here since both have a pretty low inductance.... however, the 22nF will turn inductive much sooner, so that might be a problem. When I was working on the Westerbork Telescope, I needed a very wideband (I believe 4kHz to 20MHz) coupling cap in a number of locations in the signal path so I had a tantalum elco with a 1206 capacitor. It made a really nice dip in the frequency response. Off course a tantalum cap is a disaster when it comes to inductance and this was my first year as a professional electronics engineer, so I had to find that out the hard way. I found out that putting multiple components of the same value in parallel works really well to make such a component more broadband. I plan to do some research on that for my channel and my course. I also would like to measure this so I have some real data. I did something similar last week for a very wideband RF bias tee and that worked out really nice using multiple of the same inductors in series.
@@HansRosenberg74 Hi Hans, you are correct about the added inductances. In simulations you can verify them well. In practice is is hard todo. In case of power pin decoupling one will often add more capacitors in parallel with different values and distance to the pin. See many application notes of FPGA's. With Ansys PI you can even tune this on PCB with advanced simulations.
I want to do some experiments with that at some point. I have a feeling there might be a good multiplication value for selecting the parallel components assuming their parasitic inductance is relatively equal. Should be easy to do parametrically in a simulator. If that gives some nice results I'll also put that in my course, I have a whole bunch of power supply filtering tactics I employ which would be very nice to document.
Hé Hans. Groeten van Klaas. Van Ericsson vroeger
Hoi Klaas, wat leuk, ik ben heel slecht in namen en gezichten onthouden, wat is je achternaam en wat deed je bij Ericsson, dan kan ik het plaatje wel weer maken. Groetjes, Hans
Klaas Schepers. Ik was PCB layout engineer . @@HansRosenberg74
Ah, dat dacht ik al. Ik moest er even over nadenken. Je had in bril en een snorretje, kon het 10 min na je berichtje weer voor de geest halen, Man dat is bijna 20 jaar geleden! Ben je al met pensioen? Waar heb je na SonyEricsson gezeten?
The datasheet for these seems very sparse. For starters, it seems like the voltage in the electrical specifications table doesn't match what's in the bias resistor section? The specifications table (for the 66+) says "Device operating voltage 3.0-3.5-4.0 V" (min, typ, max). But the bias section seems to require at least 7V. What's going on with this? Can the device be run off 3.3V, or does it really need 7 to operate?
You loose the rest of the voltage across the bias resistors :-)
Why did you use 1ohm and not 0ohm resistors in place of the ferrite beads? Also, send me the checklist!
I have a very intelligent and advanced answer for that......... I ran out of 0 Ohm resistors :-D. Link is in the description. BR, Hans
great job can you pls make a video about how to measure and design wifi and ble pcb antennas with the nanovna or librevna and how to make sure that my antenna will work properly like i expecting
That is on my list but may take a while before I get to it...
@@HansRosenberg74 thanks 😍
I hate MELF parts! The little buggers roll around!
yes, I noticed, this was my first experience with MELF........ I'm on the fence now what is best. The 0805 components have the drawback that they're upside down half of the time (also annoying) but they do have the great advantage you can stack them when experimenting, which is tricky with a MELF........ so not sure what to do with my next power amplifier yet.....
Why you speak fast. Slow down, so that your audience could understand what you are teaching.
Hi, thanks for the feedback. You are literally the first to mention this! This 'high information density' is a very conscious decision. I've had many comments of viewers who really appreciate that. According to research, talking fast is an advantage on youtube, Apparently the majority of viewers do not want to waste any time getting their information. What you could do is play the video at half of trhee-quarter speed. I had one viewer that was soo happy with no BS high speed approach, he said that he normally played clips at double speed but he had to slow mine down to normal speed :-D Best regards, Hans
Personally I would say better fast than slow. You can repeat as needed.
This video is full of design flaws. 42 years experience sitting at the bench?
Thanks for the feedback. Can you please point out my mistakes?
Please elaborate on this.