5 ms·

2 GHz Active Probe

98 points by hugolundin 2y ago

amelius 2y ago
Is there any open source DIY project to relatively cheaply measure the analog performance of USB 3.1 (10Gbit/s) signals?
- ComputerGuru 2y ago
  What does “analog performance” mean in this context? Are you talking about noise, reflections, etc or something else?
  - amelius 2y ago
    I mean anything that can give more information than just the digital bit errors. Like eye diagrams and such. Basically something that can rate the signal quality on some scale wider than "works? yes/no". Bit error rate is typically too low to give a useful indication.
    - hatsunearu 2y ago
      with the setup you're implying, the eye diagram is all you can get, which to be honest you can extract a lot of information about transmitter-side parameters like the jitter, but you can't evaluate receive performance at all. receive performance evaluation is done by injecting known amounts of jitter and seeing if the receiver can tolerate that jitter.
- dbuder 2y ago
  No, there was a DIY diode prob with a headline number high enough, but I think he said real world performance would be about 8ghz, that's only part of the problem anyway. I had a look, I can't find the forum post or the podcast he was on (amphour). Either rent an appropriate USB analyzer or scope. If you just want to verify the speed of your consumer equipment look for a software solution to measure throughput (there is significant protocol overhead). At the required speed you can't even use BNCs.
- rayray 2y ago
  Given the required front-end bandwidth, ADC bandwidth, and processing power, I don't think so. The closest thing you can do is perhaps get your hands on a FPGA board that has >10G serdes, and use the internal eye monitoring features to check the eye diagram, mask margin, etc.
- labcomputer 2y ago
  > Is there any open source DIY project to relatively cheaply measure the analog performance of USB 3.1 (10Gbit/s) signals? For this kind of thing (measuring eye diagrams and such, as you mentioned in another comment) you'd need a probe-oscilloscope system with (at minimum) 30 GHz at the probe tip. Realistically you'd want 50 GHz, and more is better. Recall that square waves are composed of odd harmonics of the fundamental (~10 GHz), and the more harmonics you capture the more fidelity you'll have to the original. I don't think there are any real-time scopes with this much bandwidth, so you will need a (equivalent-time) sampling oscilloscope. Keysight might make one, but it probably costs "call us" dollars. [Edit: Actually, it looks like the SD-32 sampling head for the Tektronix 11800 series of sampling scopes does claim a bandwidth of 50 GHz. That plugin does not have internal triggering though, so you would need to add a trigger recognizer head and arrange for an external trigger signal. Also, don't buy these used on eBay unless the sell offers returns because the sampling diodes in the frontend are very delicate.] One additional wrinkle is that super-speed USB uses a differential pair with a characteristic impedance of 50 Ohms (each side of the pair is 100 Ohms, and terminated with a 100 Ohm resistor). That means you'll need a differential probe (or two probes and an oscilloscope that can do math). I don't know of any DIY project with those ambitions, but the most likely route for something DIY is a passive Z0 probe: You want the DUT-probe interface to look like an impedance-matched power divider. That calls for a probe impedance of 10x (or even 100x) the transmission line, so 500 or 5000 Ohms. Anyway, I don't know of any DIY project that fits your needs, but the terms to search for include "Z0" and "passive".
  - amelius 2y ago
    I was thinking of skipping the probe entirely, and just having a USB receptacle on the board, then have a front-end IC that is hopefully available and that is fast enough to do the sampling. Along with some circuitry to make the other end talk USB3.1.
    - labcomputer 2y ago
      > [...] skipping the probe entirely, and just having a USB receptacle on the board [...] long with some circuitry to make the other end talk USB3.1. Practically, this probably ends up fairly similar to what I mentioned above, since "making the other end talk USB3" means receiving with the bits it sends and responding in a way it expects. But, yes, in-situ / solder-down probes can definitely simplify things because it gives you the flexibility to do things like remove the existing termination (assuming the termination isn't in the receiver IC itself) and use the 1x Z0 soldered-in probe as the terminator. I just probably wouldn't do this part: > then have a front-end IC that is hopefully available and that is fast enough to do the sampling This gets complicated fairly quickly. It sounds like you want both real-time sampling (so you can do something with the bits and "make the other end talk USB3") fairly high resolution (so can see something in the eye diagrams). That calls for a digitizer with like >=5 bits and a sample rate of >100 Gsps... Which basically doesn't exist, AFAIK. Oscilloscopes which sample that fast are interleaving multiple digitizers, and this isn't really something you can DIY. However, if you just want eye diagrams, you can use ET sample techniques (running at a pedestrian ~10-100 ksps), together with something to synchronize the sample to the USB clock, to fill in the entire horizontal width of the eye diagram. There are several catches: First is that it takes a very large number of triggering events to fill in the diagram. Second, the 100 ksps sampler needs a very short capture window (on the order of picoseconds). Third is that you "miss" most bits in the data stream, so you don't capture any real USB data while making the eye diagram. Fortunately, you can buy a used sampling oscilloscope which does this for you. It captures the samples, relates them to the USB clock, and places them in the correct place on the x-axis. You just need to construct probes to bring the signal into the scope. The 11800 I mentioned is one such example, but there are others, too.
- KK7NIL 2y ago
  Not that I'm aware of. You could generate an eye diagram with an old sampling scope or BERT for probably less than $2k, but those likely wouldn't have the proper clock recovery, which means your horizontal would be sort of pointless. You'd have to make a separate clock recovery module to sit between your USB 3.1 DUT and the instrument. The other option would be to try and make a real-time ADC with enough BW (at Tek we demoed USB 3.1 debugging with just 10 GHz on an MSO6B, even though compliance testing requires 12 or 15 iirc) and memory depth to do the clock recovery in DSP. This would be a very significant challenge, but might be possible, depending what ADC's are on the market now a days.
  - CamperBob2 2y ago
    The other option would be to try and make a real-time ADC with enough BW (at Tek we demoed USB 3.1 debugging with just 10 GHz on an MSO6B, even though compliance testing requires 12 or 15 iirc) and memory depth to do the clock recovery in DSP. This would be a very significant challenge, but might be possible, depending what ADC's are on the market now a days. It strikes me that if you ran it long enough, there is probably no reason why a sampling scope couldn't do the same thing. It would need even more DSP eggheadery than the realtime scope, of course.
    - KK7NIL 2y ago
      No because modern high speed serial uses clock recovery for the timing (implemented as a first or second order PLL in the receiver), so you need to implement this in your scope too. In a real-time scope this can be done in software but in a sampling scope it has to be done in hardware and then fed into the trigger. If you just trigger off the edges in the signal all you'll be measuring is the scope jitter.
      - CamperBob2 2y ago
        If all you want is an eye diagram, you may not necessarily care about clock recovery, though. You'd just need a high degree of control over your own sampling clock (i.e., better than whatever you're trying to measure.)
- hatsunearu 2y ago
  if you want to do this, you generally want a breakout board that breaks out the high speed differential signals to 50 ohm RF connectors and hook that up to a 50 ohm terminated real time scope, and not use a probe. also for 10gbps per lane, you need at minimum 5ghz bandwidth.
cushychicken 2y ago
In case folks don’t know about the “El Cheapo Special” - you can make very respectable 2GHz passive scope probes with 150ps rise times out of RG-174 coax cable and 1k resistors. Clip the cable in half, solder the core to the 1k resistor, and the other end of the 1k resistor to the signal you want to probe. Solder the shield braid to board GND. 50ohm terminate your scope, set probe amplification to 20x, and voila! Excellent tip recommended both by Paul Horowitz (Art of Electronics) and Howard Johnson (High Speed Digital Design). Works like a charm!
- 0l 2y ago
  The catch however is that this can cause (relatively) high resistive loading, so isn't really suitable for all applications.
  - jjoonathan 2y ago
    It really is suitable for most transmission line applications, though, because transmission line impedance is a fixed target. Top-shelf 20-30GHz probes use this exact technique as their "front end" followed by a 50/100 ohm amplifier to lock in the noise figure: https://www.tek.com/en/datasheet/trimode%28tm%29-probe-family-0 https://www.tek.com/en/datasheet/trimode%28tm%29-probe-famil...
  - cushychicken 2y ago
    True - but those are the exceptions, not the rule. Most modern systems aren’t discrete transistor amplifiers where an arbitrary 1k to gnd will massively affect bias point or circuit operation. For digital interfaces, or power supplies below ~24V, it’s a great high bandwidth probe option. Much cheaper than active probes too.

98 points by hugolundin 2y ago

Is there any open source DIY project to relatively cheaply measure the analog performance of USB 3.1 (10Gbit/s) signals?

ComputerGuru 2y ago
What does “analog performance” mean in this context? Are you talking about noise, reflections, etc or something else?
- amelius 2y ago
  I mean anything that can give more information than just the digital bit errors. Like eye diagrams and such. Basically something that can rate the signal quality on some scale wider than "works? yes/no". Bit error rate is typically too low to give a useful indication.
  - hatsunearu 2y ago
    with the setup you're implying, the eye diagram is all you can get, which to be honest you can extract a lot of information about transmitter-side parameters like the jitter, but you can't evaluate receive performance at all. receive performance evaluation is done by injecting known amounts of jitter and seeing if the receiver can tolerate that jitter.
dbuder 2y ago
No, there was a DIY diode prob with a headline number high enough, but I think he said real world performance would be about 8ghz, that's only part of the problem anyway. I had a look, I can't find the forum post or the podcast he was on (amphour). Either rent an appropriate USB analyzer or scope. If you just want to verify the speed of your consumer equipment look for a software solution to measure throughput (there is significant protocol overhead). At the required speed you can't even use BNCs.
rayray 2y ago
Given the required front-end bandwidth, ADC bandwidth, and processing power, I don't think so. The closest thing you can do is perhaps get your hands on a FPGA board that has >10G serdes, and use the internal eye monitoring features to check the eye diagram, mask margin, etc.
labcomputer 2y ago
> Is there any open source DIY project to relatively cheaply measure the analog performance of USB 3.1 (10Gbit/s) signals? For this kind of thing (measuring eye diagrams and such, as you mentioned in another comment) you'd need a probe-oscilloscope system with (at minimum) 30 GHz at the probe tip. Realistically you'd want 50 GHz, and more is better. Recall that square waves are composed of odd harmonics of the fundamental (~10 GHz), and the more harmonics you capture the more fidelity you'll have to the original. I don't think there are any real-time scopes with this much bandwidth, so you will need a (equivalent-time) sampling oscilloscope. Keysight might make one, but it probably costs "call us" dollars. [Edit: Actually, it looks like the SD-32 sampling head for the Tektronix 11800 series of sampling scopes does claim a bandwidth of 50 GHz. That plugin does not have internal triggering though, so you would need to add a trigger recognizer head and arrange for an external trigger signal. Also, don't buy these used on eBay unless the sell offers returns because the sampling diodes in the frontend are very delicate.] One additional wrinkle is that super-speed USB uses a differential pair with a characteristic impedance of 50 Ohms (each side of the pair is 100 Ohms, and terminated with a 100 Ohm resistor). That means you'll need a differential probe (or two probes and an oscilloscope that can do math). I don't know of any DIY project with those ambitions, but the most likely route for something DIY is a passive Z0 probe: You want the DUT-probe interface to look like an impedance-matched power divider. That calls for a probe impedance of 10x (or even 100x) the transmission line, so 500 or 5000 Ohms. Anyway, I don't know of any DIY project that fits your needs, but the terms to search for include "Z0" and "passive".
- amelius 2y ago
  I was thinking of skipping the probe entirely, and just having a USB receptacle on the board, then have a front-end IC that is hopefully available and that is fast enough to do the sampling. Along with some circuitry to make the other end talk USB3.1.
  - labcomputer 2y ago
    > [...] skipping the probe entirely, and just having a USB receptacle on the board [...] long with some circuitry to make the other end talk USB3.1. Practically, this probably ends up fairly similar to what I mentioned above, since "making the other end talk USB3" means receiving with the bits it sends and responding in a way it expects. But, yes, in-situ / solder-down probes can definitely simplify things because it gives you the flexibility to do things like remove the existing termination (assuming the termination isn't in the receiver IC itself) and use the 1x Z0 soldered-in probe as the terminator. I just probably wouldn't do this part: > then have a front-end IC that is hopefully available and that is fast enough to do the sampling This gets complicated fairly quickly. It sounds like you want both real-time sampling (so you can do something with the bits and "make the other end talk USB3") fairly high resolution (so can see something in the eye diagrams). That calls for a digitizer with like >=5 bits and a sample rate of >100 Gsps... Which basically doesn't exist, AFAIK. Oscilloscopes which sample that fast are interleaving multiple digitizers, and this isn't really something you can DIY. However, if you just want eye diagrams, you can use ET sample techniques (running at a pedestrian ~10-100 ksps), together with something to synchronize the sample to the USB clock, to fill in the entire horizontal width of the eye diagram. There are several catches: First is that it takes a very large number of triggering events to fill in the diagram. Second, the 100 ksps sampler needs a very short capture window (on the order of picoseconds). Third is that you "miss" most bits in the data stream, so you don't capture any real USB data while making the eye diagram. Fortunately, you can buy a used sampling oscilloscope which does this for you. It captures the samples, relates them to the USB clock, and places them in the correct place on the x-axis. You just need to construct probes to bring the signal into the scope. The 11800 I mentioned is one such example, but there are others, too.
KK7NIL 2y ago
Not that I'm aware of. You could generate an eye diagram with an old sampling scope or BERT for probably less than $2k, but those likely wouldn't have the proper clock recovery, which means your horizontal would be sort of pointless. You'd have to make a separate clock recovery module to sit between your USB 3.1 DUT and the instrument. The other option would be to try and make a real-time ADC with enough BW (at Tek we demoed USB 3.1 debugging with just 10 GHz on an MSO6B, even though compliance testing requires 12 or 15 iirc) and memory depth to do the clock recovery in DSP. This would be a very significant challenge, but might be possible, depending what ADC's are on the market now a days.
- CamperBob2 2y ago
  The other option would be to try and make a real-time ADC with enough BW (at Tek we demoed USB 3.1 debugging with just 10 GHz on an MSO6B, even though compliance testing requires 12 or 15 iirc) and memory depth to do the clock recovery in DSP. This would be a very significant challenge, but might be possible, depending what ADC's are on the market now a days. It strikes me that if you ran it long enough, there is probably no reason why a sampling scope couldn't do the same thing. It would need even more DSP eggheadery than the realtime scope, of course.
  - KK7NIL 2y ago
    No because modern high speed serial uses clock recovery for the timing (implemented as a first or second order PLL in the receiver), so you need to implement this in your scope too. In a real-time scope this can be done in software but in a sampling scope it has to be done in hardware and then fed into the trigger. If you just trigger off the edges in the signal all you'll be measuring is the scope jitter.
    - CamperBob2 2y ago
      If all you want is an eye diagram, you may not necessarily care about clock recovery, though. You'd just need a high degree of control over your own sampling clock (i.e., better than whatever you're trying to measure.)
hatsunearu 2y ago
if you want to do this, you generally want a breakout board that breaks out the high speed differential signals to 50 ohm RF connectors and hook that up to a 50 ohm terminated real time scope, and not use a probe. also for 10gbps per lane, you need at minimum 5ghz bandwidth.

cushychicken 2y ago

In case folks don’t know about the “El Cheapo Special” - you can make very respectable 2GHz passive scope probes with 150ps rise times out of RG-174 coax cable and 1k resistors. Clip the cable in half, solder the core to the 1k resistor, and the other end of the 1k resistor to the signal you want to probe. Solder the shield braid to board GND. 50ohm terminate your scope, set probe amplification to 20x, and voila! Excellent tip recommended both by Paul Horowitz (Art of Electronics) and Howard Johnson (High Speed Digital Design). Works like a charm!

0l 2y ago
The catch however is that this can cause (relatively) high resistive loading, so isn't really suitable for all applications.
- jjoonathan 2y ago
  It really is suitable for most transmission line applications, though, because transmission line impedance is a fixed target. Top-shelf 20-30GHz probes use this exact technique as their "front end" followed by a 50/100 ohm amplifier to lock in the noise figure: https://www.tek.com/en/datasheet/trimode%28tm%29-probe-family-0 https://www.tek.com/en/datasheet/trimode%28tm%29-probe-famil...
- cushychicken 2y ago
  True - but those are the exceptions, not the rule. Most modern systems aren’t discrete transistor amplifiers where an arbitrary 1k to gnd will massively affect bias point or circuit operation. For digital interfaces, or power supplies below ~24V, it’s a great high bandwidth probe option. Much cheaper than active probes too.