[HN Gopher] 2 GHz Active Probe
       ___________________________________________________________________
        
       2 GHz Active Probe
        
       Author : hugolundin
       Score  : 73 points
       Date   : 2024-06-16 07:36 UTC (15 hours ago)
        
 (HTM) web link (jmw.name)
 (TXT) w3m dump (jmw.name)
        
       | amelius wrote:
       | Is there any open source DIY project to relatively cheaply
       | measure the analog performance of USB 3.1 (10Gbit/s) signals?
        
         | ComputerGuru wrote:
         | What does "analog performance" mean in this context? Are you
         | talking about noise, reflections, etc or something else?
        
           | amelius wrote:
           | 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 wrote:
             | 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 wrote:
         | 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 wrote:
         | 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 wrote:
         | > 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 wrote:
           | 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.
        
           | KK7NIL wrote:
           | The old rule of thumb that you want the bandwidth to be 3 to
           | 5x your baud rate is no longer true. It has been found that
           | BW as low as 0.7x has very little effect in TDECQ and so the
           | newest optical ethernet standards have moved to that for
           | standards testing and the electrical standards will likely
           | soon follow. This BW is also applied as a 4th order Bessel-
           | Thompson instead of an oldschool 1st order gaussian response.
           | 
           | Real-time scopes go up to 110 GHz now! Sampling scopes are
           | unfortunately dying out now a days.
        
             | CamperBob2 wrote:
             | There is an interesting new 6 GHz sampling scope that could
             | potentially be useful for some of this work:
             | https://www.eevblog.com/forum/testgear/pocket-
             | sized-6-ghz-1-...
             | 
             | Haven't used it myself, but they seem to have a really
             | solid customer support ethic, given the way the thread
             | developed.
        
               | wrs wrote:
               | And re the original topic, that thread has a reference to
               | a 2.7GHz active probe for $226!
               | 
               | https://www.lasmux.com/product/single-ended-active-
               | probes/
        
             | nyanpasu64 wrote:
             | Is the analog Nyquist frequency of the signal 0.5x the
             | digital baud (effectively sampling) rate?
        
               | bobmcnamara wrote:
               | Not exactly. All other parameters the same, the digital
               | baud rate plays into how much bandwidth the signal uses,
               | but not directly where it is in the spectrum.
               | 
               | Ex:9600 baud FSK signal centered at 440MHz still needs
               | something capable of capturing around 440MHz, but this
               | has nothing to do with baud rate. A 500MHz scope would do
               | well for acquiring the signal.
               | 
               | Practically radios often divide the problem into two
               | parts, first use a mixer to shift the signal down and
               | filter out the parts you don't want, then decode the FSK
               | which remains at a much lower frequency.
               | 
               | This is the governing limit: https://en.m.wikipedia.org/w
               | iki/Shannon%E2%80%93Hartley_theo... and it considers
               | things like SNR as well.
        
           | photon_rancher wrote:
           | Realtime scopes exist at this bandwidth but are crazy
           | expensive, and then they have frequency-stitching artifacts
           | all over the fourier plane. (Because they use mixing
           | techniques)
           | 
           | Sampling systems are about $20k - 50k without probes in this
           | bandwidth.
           | 
           | Keysight DCA-M and picoscope have the best public prices, and
           | exfo has an eye analyzer in that price range. There's also
           | DCA-X but nobody wants to pay 10k extra for a touchscreen.
           | 
           | Unfortunately tek doesn't make electrical sampling systems
           | anymore, just optical.
        
         | KK7NIL wrote:
         | 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 wrote:
           | _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 wrote:
             | 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 wrote:
               | 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.)
        
               | KK7NIL wrote:
               | For amplitude (the vertical part of the eye), yes. For
               | jitter (the horizontal part), which is usually the most
               | critical, especially for NRZ protocols like USB 3.1, no,
               | you need clock recovery.
        
         | hatsunearu wrote:
         | 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 wrote:
       | 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 wrote:
         | The catch however is that this can cause (relatively) high
         | resistive loading, so isn't really suitable for all
         | applications.
        
           | jjoonathan wrote:
           | 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-
           | famil...
        
           | cushychicken wrote:
           | 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.
        
       ___________________________________________________________________
       (page generated 2024-06-16 23:01 UTC)