FYI, a really interesting article related to this is "Secret world of oscilloscope probes" by Doug Ford. This was published by "Silicon Chip" in October 2009. One of the interesting features I was not aware of, is that the probe cables are lossy by design to suppress signal reflections. Another great video - thanks !
I think possibly you are mostly being bandwidth limited by the scope itself---1ns rise time is in general approximately 350MHz of bandwidth. Recently I looked at the output of an HP comb generator (step recovery diode with pulse output approximately 100-200ps) on my Lecroy 7200A with 500MHz 7242B plugin. I was only able to measure the rise time at about 900ps, and that was running directly into the scope with no probe. I had hoped to be able to measure it a bit more accurately but obviously my scope bandwidth was the limiting factor. So the solution is easy---go buy a Tek 11000 series sampling scope with a 20GHz plugin and redo the video...but don't forget to clear off 45 cubic feet of space on your bench first. :-D
Interesting. I have nothing certain to offer but a few things to consider. 1) The lead length of the 3904 is a bit long. 2) There exist avalanche specific transistors in SMD, I don't recall number. 3) When building the pulse circuit 50ohm output is not magic, try lower. 4) Build also a pulse circuit that connects directly to the scope thru a BNC w/o cable. cheers dan
If you are going to use or look at high speed, repetitive signals, you might try a sampling probe. It is a diode bridge made from 4 Schottky diodes. I used one such during the mid 90s and I could then use my 2.5 maxed out o-scope (like the ones that were once used to fix TVs with) to look at stuff right up to about 3 gig or so. The trigger that was used to open the bridge in the probe heat had 4 small coaxed cables running to it from the control box and used an avalanche set up like the one in the clip. The control box was all transistors, no ICs. ICs, I guess, might be pretty noisy at the higher frequencies.
That reminds me of a student project. For a pulse laser diode I had to build a square wave generator capable of 1A with a rise time of 1ns back in 1980. I used a 4 stage avalanche setup with open coax load line. Each stage produced a rise time of 250ps at 250mA triggered all at the same time to get 1A. The triggering also used an avalanche pulser. The open coax gives a nice square pulse compared to just using a cap
How bad were the harmonics of that generator? Did you get any letters from the FCC? It would be interesting to see the schematic of the generator, as I've struggled to make fast rise time clock generators with just transistors.
You're right up against the bandwidth limitation of the scope. Rise time is given by .35/freq, so for your 350 MHz scope, you can expect a rise time of 1 nanosecond. Add in the rise time of the probes to the rise time of the scope, and you are at the limit of your ability to measure. Time to spend more money!
Yup, I thought of exactly the same quick calculation that you used, and the 350 MHz bandwidth (3 dB down) scope is unfortunately not up to the task. This is where Jim Williams would just pull out the 1GHz bandwidth scope, attach a 500 MHz probe, show the improvement and ask if you have any questions. I think Jim always had the next bullet point already figured out and the appropriate test gear is just waiting in the wings incase needed.
One confounding factor could be that it appears to be a *very* narrow pulse, so the maximum observed amplitude is itself a function of the probe/scope’s bandwidth. The 10/90% points thus end up being relative to a level that’s in part determined by the probe’s bandwidth. I think this would tend to make slower probes seem faster than they actually are. I can’t properly think through the math in my head, but suspect you’d get significantly different results if you were looking at a step function vs an impulse. With a step, the ultimate voltage would always be the same regardless of bandwidth, which I think would tend to make the slower probes appear more proportionately slower than they seemed to be in the tests here. (Related to this, I wonder if the larger signal from the cheap Chinese probe that you remarked upon might simply have been a result of its apparently higher bandwidth, vs improper calibration: I’d be interested to hear how it’s calibration compared to the others when measuring continuous, low-frequency waveforms…) Thanks as always for a super-interesting video!
Yes. Rise time is generally defined for a long pulse input -- long enough for the response to settle to steady state. For the very narrow pulse input, as you point out, the steady state output condition may not be reached.
Read the app note. Start by driving the signal directly into the scope with 50ohm termination direct coupled and do the BW calculation. BTW, your dead bug construction is marginal for sub nanosecond rise time signals.
I recall using a simple pulse generator in a college Physics class. It consisted of a mercury wetted relay with a form C contact. A small, ceramic disk capacitor was connected between the common pole of the relay and ground. When the relay coil was de-energized, it charged the capacitor through a resistor from a DC power source. When the relay was energized, it discharged the capacitor directly into the load. This produced short pulses with a very fast rise time. Sorry, I don't have any actual rise time values to post here.
You did not mention that the measured risetime is the root of the sum of the 3 risetimes squared. These 3 risetimes are the risetimes of the generator, the probe and the scope. When the generator is fast enough (or assumed to be) and the risetime of the scope is given or measured via bandwidth the risetime of the probes can be measured. (It might not be a perfect measurement with the given set up.)
This is sorta true. The root sum of the squares only really works for combining independent systems. Systems that when you connect them together retain their properties. However, in this case the systems interact. What I mean by this is the probe affects the pulser, the scope affects the probe and vise versa, etc... This is especially true for modern scopes. The front ends on these are not Gaussian like the old analog scopes. They employ a lot of tricks to maximize bandwidth. The real math becomes quite insane and amounts to a set of simultaneous non-linear differential equations.
It would be interesting to see the effect of using the probes' ground leads - the additional series inductance could have quite an effect and would show the care needed when trying to measure a circuit so that you measure what you want to measure.
With a bit of cleaning up of your layout (VERY short leads and surface mount components) the risetime of that avalanche pulser is better than half a nanosecond. (limit of my 500mhz HP scope) Try directly connecting the circuit to a BNC that will plug onto the scope. Use the impedance switch on the scope to switch the input to 50 ohms instead of hi Z. This will give you the rise time of the scope. When using a slow scope like that it is difficult to accurately measure nanosecond pulses. That scope uses multiple iterations of the input signal to get the waveform. Any jitter in the signal (relaxation osc, oh my!) will skew the risetime edge measurement when at that short time scale ruining the accuracy. If you really want to measure that kind of speed, save up your money and get a good HP scope from ebay with 8ghz or higher sample rate.
*Summary* *Goal:* Measure the rise time of different oscilloscope probes. *Challenges:* * Requires a signal source with a rise time significantly faster than the probes being tested. * Scope bandwidth limits measurement accuracy (350 MHz scope = ~1 ns rise time). *Procedure:* 1. *(**0:00**) Initial attempt:* A 74AC-based pulser, achieving ~2.4 ns rise time, wasn't fast enough to differentiate the probes. 2. *(**1:58**) Improved pulse generator:* Built an avalanche transistor-based pulser aiming for sub-nanosecond rise times (design from Jim Williams' App Note 47). 3. *(**6:06**) Probe testing:* * Connected probes to the pulser using a low-inductance method. * *(**6:08**)* Measured the rise time of the Rigol PVP2350 (350 MHz) probe. * *(**6:52**)* Measured the rise time of an HP probe (bandwidth not mentioned). * *(7:49) Measured the rise time of a Tektronix P6106 (250 MHz) probe. * *(8:44) Measured the rise time of a Tektronix P2200 (200 MHz) probe. * *(11:06) Measured the rise time of a 100 MHz Tektronix probe. * *(12:17) Measured the rise time of a generic, unbranded "cheap Chinese" probe. * *(14:11) Measured the rise time of a vintage HP 275 MHz probe with a small form factor. *Results:* * The Rigol probes performed surprisingly well, comparable to higher-end probes in this specific test. * Measurement limitations due to scope bandwidth and pulse generator speed make it difficult to accurately differentiate the probes. *Future Plans:* * Obtaining a faster pulse generator (
Pretty long lead length there on that 2N3904. Scope rise time is probably limiting you also. Try direct 50 Ohm coax into scope terminated in 50 Ohms just to see what is what. Longer pulse duration also would help maybe? Just to make sure you get full swing before pulse cuts off.
@@argcargv You're right, I had that backwards. In fact I have a 100:1 250MHz probe that I was playing with, and it was getting slightly better results than a 10:1 500MHz probe (on a 1GHz scope). And FWIW, I experimented again with other settings that I mentioned and really didn't see much difference... so you can ignore my whole ignorant post :)
I would have liked to see you use a short piece of coax, fold a little shield back as your ground and expose the center conductor all to fit in your fixture, I know the probe tip ficture as a Curly Q. Use a BNC on the other end. Then terminate the coax at the scope with 50 ohms. Yes, it is low impedance, but I'd like to see the waveform.
@@IMSAIGuy So, you got 1ns rise time using (What kind?) coax and how long, terminated at the scope with 50ohms? Have you figured out what the rise time of your scope is?
I believe you can get the frequency response using a spectrum analyser with tracking generator. Personally, I would try to fi nd the -3dB frequency by using a sine wave and changing its frequency. of course, the probe bw should be lower than that of the oscilloscope.... Nice vídeo ;)
Bandwidth x risetime = 0.45 in modern scopes -> 350MHz bandwidth equals roughly 1.28ns rise time. That is what you are measuring. The probes have barely an influence I guess
Your measuring is limited buy your scope. What your rise time with you pulse generator connected directly to your scope? Before you tested did you adjust the probes with your scope for a nice square wave?
Hi for traveling a very fast rise time pulse signal from one board to another have i use coax??? If I use normal two wires can it decrease the rise time of pulse???
Looking up the specs of the scope you are using I see that it has an analog bandwidth of only 70 MHz so these measurements are severely tainted by using the inadequate scope for your probe comparisons.
I think C1(22pF) is too large. this causing pulse height too high, looks like about 5V. I have avalanche pulse generator too. it is without C1, parasitic capacitance only. in #838, you measure 3db point is more than 500MHz. 1.2ns risetime is not your scope limit I guess (but for 5V pulse... maybe).
@@matteo234321 I tried it. it is certainly interesting. I'll do a video on it in the future. Don't think it tells the whole story but is another datapoint
FYI, a really interesting article related to this is "Secret world of oscilloscope probes" by Doug Ford.
This was published by "Silicon Chip" in October 2009.
One of the interesting features I was not aware of, is that the probe cables are lossy by design to suppress signal reflections.
Another great video - thanks !
Thanks! I just downloaded it.
Cool article
Yup, the center conductor is very fine with significant resistance
I think possibly you are mostly being bandwidth limited by the scope itself---1ns rise time is in general approximately 350MHz of bandwidth. Recently I looked at the output of an HP comb generator (step recovery diode with pulse output approximately 100-200ps) on my Lecroy 7200A with 500MHz 7242B plugin. I was only able to measure the rise time at about 900ps, and that was running directly into the scope with no probe. I had hoped to be able to measure it a bit more accurately but obviously my scope bandwidth was the limiting factor. So the solution is easy---go buy a Tek 11000 series sampling scope with a 20GHz plugin and redo the video...but don't forget to clear off 45 cubic feet of space on your bench first. :-D
Interesting. I have nothing certain to offer but a few things to consider.
1) The lead length of the 3904 is a bit long.
2) There exist avalanche specific transistors in SMD, I don't recall number.
3) When building the pulse circuit 50ohm output is not magic, try lower.
4) Build also a pulse circuit that connects directly to the scope thru a BNC w/o cable.
cheers dan
This!
@crash burn no, it is driving 10meg. Think.
If you are going to use or look at high speed, repetitive signals, you might try a sampling probe. It is a diode bridge made from 4 Schottky diodes. I used one such during the mid 90s and I could then use my 2.5 maxed out o-scope (like the ones that were once used to fix TVs with) to look at stuff right up to about 3 gig or so. The trigger that was used to open the bridge in the probe heat had 4 small coaxed cables running to it from the control box and used an avalanche set up like the one in the clip. The control box was all transistors, no ICs. ICs, I guess, might be pretty noisy at the higher frequencies.
That reminds me of a student project. For a pulse laser diode I had to build a square wave generator capable of 1A with a rise time of 1ns back in 1980. I used a 4 stage avalanche setup with open coax load line. Each stage produced a rise time of 250ps at 250mA triggered all at the same time to get 1A. The triggering also used an avalanche pulser. The open coax gives a nice square pulse compared to just using a cap
How bad were the harmonics of that generator? Did you get any letters from the FCC? It would be interesting to see the schematic of the generator, as I've struggled to make fast rise time clock generators with just transistors.
You're right up against the bandwidth limitation of the scope. Rise time is given by .35/freq, so for your 350 MHz scope, you can expect a rise time of 1 nanosecond. Add in the rise time of the probes to the rise time of the scope, and you are at the limit of your ability to measure. Time to spend more money!
Yup, I thought of exactly the same quick calculation that you used, and the 350 MHz bandwidth (3 dB down) scope is unfortunately not up to the task. This is where Jim Williams would just pull out the 1GHz bandwidth scope, attach a 500 MHz probe, show the improvement and ask if you have any questions. I think Jim always had the next bullet point already figured out and the appropriate test gear is just waiting in the wings incase needed.
One confounding factor could be that it appears to be a *very* narrow pulse, so the maximum observed amplitude is itself a function of the probe/scope’s bandwidth. The 10/90% points thus end up being relative to a level that’s in part determined by the probe’s bandwidth. I think this would tend to make slower probes seem faster than they actually are. I can’t properly think through the math in my head, but suspect you’d get significantly different results if you were looking at a step function vs an impulse. With a step, the ultimate voltage would always be the same regardless of bandwidth, which I think would tend to make the slower probes appear more proportionately slower than they seemed to be in the tests here. (Related to this, I wonder if the larger signal from the cheap Chinese probe that you remarked upon might simply have been a result of its apparently higher bandwidth, vs improper calibration: I’d be interested to hear how it’s calibration compared to the others when measuring continuous, low-frequency waveforms…) Thanks as always for a super-interesting video!
Yes. Rise time is generally defined for a long pulse input -- long enough for the response to settle to steady state. For the very narrow pulse input, as you point out, the steady state output condition may not be reached.
Read the app note. Start by driving the signal directly into the scope with 50ohm termination direct coupled and do the BW calculation. BTW, your dead bug construction is marginal for sub nanosecond rise time signals.
I recall using a simple pulse generator in a college Physics class. It consisted of a mercury wetted relay with a form C contact. A small, ceramic disk capacitor was connected between the common pole of the relay and ground. When the relay coil was de-energized, it charged the capacitor through a resistor from a DC power source. When the relay was energized, it discharged the capacitor directly into the load. This produced short pulses with a very fast rise time. Sorry, I don't have any actual rise time values to post here.
Nice Vid
I have a Leo Bodnar pulse gen , BNC version
32ps rise , 30ps fall
Measured with CSA803 w/SD-30 head
12:19 jab at the "Cheap Chinese probes" which actually kicked ass with one of the fastest rise times.
These $10 probes show heavy ringing. Should the test pulse be a square wave, instead of a spike, the difference would be obvious.
You did not mention that the measured risetime is the root of the sum of the 3 risetimes squared. These 3 risetimes are the risetimes of the generator, the probe and the scope. When the generator is fast enough (or assumed to be) and the risetime of the scope is given or measured via bandwidth the risetime of the probes can be measured.
(It might not be a perfect measurement with the given set up.)
This is sorta true. The root sum of the squares only really works for combining independent systems. Systems that when you connect them together retain their properties. However, in this case the systems interact. What I mean by this is the probe affects the pulser, the scope affects the probe and vise versa, etc... This is especially true for modern scopes. The front ends on these are not Gaussian like the old analog scopes. They employ a lot of tricks to maximize bandwidth. The real math becomes quite insane and amounts to a set of simultaneous non-linear differential equations.
ECL logic provides good rise time. I used it for CERN project while back.
It would be interesting to see the effect of using the probes' ground leads - the additional series inductance could have quite an effect and would show the care needed when trying to measure a circuit so that you measure what you want to measure.
With a bit of cleaning up of your layout (VERY short leads and surface mount components) the risetime of that avalanche pulser is better than half a nanosecond. (limit of my 500mhz HP scope)
Try directly connecting the circuit to a BNC that will plug onto the scope. Use the impedance switch on the scope to switch the input to 50 ohms instead of hi Z. This will give you the rise time of the scope. When using a slow scope like that it is difficult to accurately measure nanosecond pulses. That scope uses multiple iterations of the input signal to get the waveform. Any jitter in the signal (relaxation osc, oh my!) will skew the risetime edge measurement when at that short time scale ruining the accuracy. If you really want to measure that kind of speed, save up your money and get a good HP scope from ebay with 8ghz or higher sample rate.
Thanks a ton!
*Summary*
*Goal:* Measure the rise time of different oscilloscope probes.
*Challenges:*
* Requires a signal source with a rise time significantly faster than the probes being tested.
* Scope bandwidth limits measurement accuracy (350 MHz scope = ~1 ns rise time).
*Procedure:*
1. *(**0:00**) Initial attempt:* A 74AC-based pulser, achieving ~2.4 ns rise time, wasn't fast enough to differentiate the probes.
2. *(**1:58**) Improved pulse generator:* Built an avalanche transistor-based pulser aiming for sub-nanosecond rise times (design from Jim Williams' App Note 47).
3. *(**6:06**) Probe testing:*
* Connected probes to the pulser using a low-inductance method.
* *(**6:08**)* Measured the rise time of the Rigol PVP2350 (350 MHz) probe.
* *(**6:52**)* Measured the rise time of an HP probe (bandwidth not mentioned).
* *(7:49) Measured the rise time of a Tektronix P6106 (250 MHz) probe.
* *(8:44) Measured the rise time of a Tektronix P2200 (200 MHz) probe.
* *(11:06) Measured the rise time of a 100 MHz Tektronix probe.
* *(12:17) Measured the rise time of a generic, unbranded "cheap Chinese" probe.
* *(14:11) Measured the rise time of a vintage HP 275 MHz probe with a small form factor.
*Results:*
* The Rigol probes performed surprisingly well, comparable to higher-end probes in this specific test.
* Measurement limitations due to scope bandwidth and pulse generator speed make it difficult to accurately differentiate the probes.
*Future Plans:*
* Obtaining a faster pulse generator (
As I mostly work on audio frequenccies I have nothing to say, but, interesting results. )
Pretty long lead length there on that 2N3904. Scope rise time is probably limiting you also. Try direct 50 Ohm coax into scope terminated in 50 Ohms just to see what is what. Longer pulse duration also would help maybe? Just to make sure you get full swing before pulse cuts off.
I recently did. limited to 1ns. working on a new idea
Before testing scope probe connect with shortest 50R coax direct to scope get reference waveform. 0.22/350MHz should be around 650pSec rise time.
or 0.35/350MHz ?
You may be running into the rise time limits of the scope and the acquisition setup. Specs for 350MHz I believe say "
Also make sure you're using the probes in 1x mode, not 10x?
@@sorcererstan no, to get the highest bandwidth you need to use the 10x option.
@@argcargv You're right, I had that backwards. In fact I have a 100:1 250MHz probe that I was playing with, and it was getting slightly better results than a 10:1 500MHz probe (on a 1GHz scope). And FWIW, I experimented again with other settings that I mentioned and really didn't see much difference... so you can ignore my whole ignorant post :)
@@sorcererstan The 100:1 have a lower capacitive loading on the source.
I would have liked to see you use a short piece of coax, fold a little shield back as your ground and expose the center conductor all to fit in your fixture, I know the probe tip ficture as a Curly Q. Use a BNC on the other end. Then terminate the coax at the scope with 50 ohms. Yes, it is low impedance, but I'd like to see the waveform.
1ns. working on something elese
@@IMSAIGuy So, you got 1ns rise time using (What kind?) coax and how long, terminated at the scope with 50ohms? Have you figured out what the rise time of your scope is?
Should have gone back to the first probe to see if the weird fall off was caused by heat/degradation of the transistor.
Were they all compensated? What effect would that have on this test
I believe you can get the frequency response using a spectrum analyser with tracking generator. Personally, I would try to fi nd the -3dB frequency by using a sine wave and changing its frequency. of course, the probe bw should be lower than that of the oscilloscope.... Nice vídeo ;)
several people have suggested that, the data is very strange also. I will do a video in the future
Bandwidth x risetime = 0.45 in modern scopes -> 350MHz bandwidth equals roughly 1.28ns rise time. That is what you are measuring. The probes have barely an influence I guess
Your measuring is limited buy your scope. What your rise time with you pulse generator connected directly to your scope? Before you tested did you adjust the probes with your scope for a nice square wave?
12:54 measured rise time is lower but... fall time has more divs.
Leo Bodnar is amused.
Do you have access to any tunnel diodes ? You could build a simple TD pulser and generate very fast rise times.
no, I looked for some but didn't want to have to buy new ones
With 70MHZ bandwidth you are getting the right result. You need a faster scope to see a sub-ns risetime
the scope is 350MHz tested to 500MHz
it has the software unlocked
Hi for traveling a very fast rise time pulse signal from one board to another have i use coax???
If I use normal two wires can it decrease the rise time of pulse???
most modern designs use differential pairs.
@@IMSAIGuy in my case my signal is a high power and single ended so I can't use differential pair.
ok, I would recommend a transmission line then (coax). make sure the input, output, and coax, all have the same impedance.
@@IMSAIGuy OK thanks
Inverse Fourier transform of frequency response == impulse response
I didn't see you do probe compensation?
doesn't mean I didn't do it
Use an HDMI source for fast edges.
WHY?
Looking up the specs of the scope you are using I see that it has an analog bandwidth of only 70 MHz so these measurements are severely tainted by using the inadequate scope for your probe comparisons.
my scope runs to 500MHz, upgraded: ruclips.net/video/eaoHYWYLRV0/видео.htmlsi=eCDisqU3-fSAvPbG
Isn't the fact that you have an impedance mismatch a problem? Your generator is 50 ohm but a 10 x probe is nowhere near 50 ohms.
don't think so. It is like using a scope probe on a dummy load to look at a transmitter. or a multimeter checking the voltage on a load.
I think C1(22pF) is too large. this causing pulse height too high, looks like about 5V.
I have avalanche pulse generator too. it is without C1, parasitic capacitance only.
in #838, you measure 3db point is more than 500MHz. 1.2ns risetime is not your scope limit I guess (but for 5V pulse... maybe).
Why not use a network analyzer?
Hmm, have to think about that
I was thinking about using a spectrum analyzer with tracking gen. Just need an adapter to convert type N to BNC and then just probe the gen output.
@@matteo234321 I tried it. it is certainly interesting. I'll do a video on it in the future. Don't think it tells the whole story but is another datapoint
oh I know what's wrong .
says right on the label:
MADE IN CHINA
😁
Better get it from MADE IN MARS