No Script, No Fear, All Opinion
RSS icon Home icon
  • EEVblog #591 – Agilent 54622D Retro Mixed Signal Osciloscope Review & Teardown

    Posted on March 16th, 2014 EEVblog 8 comments


    Dave checks out one of his favorite “old” oscilloscopes, the 2000 vintage Agilent 54622D mixed signal oscilloscope with the original megazoom ASIC technology and high resolution CRT display.
    Megazoom technology presentation.
    Datasheet
    Forum HERE

    Hi-res teardown photos:

    Be Sociable, Share!
    • Joker

      Just as a side note, me and my coworkers jokingly call my boss a “mixed signal” engineer. Pun intended.

    • http://www.cadsoftusa.com Jorge Garcia

      It may sound silly, but I’m not sure how this scope can be specified as having a 100MHz bandwidth with only 200MSa/s. Technically I know it meets Nyquist but usually you want at least 5 times the sample rate for a given bandwith in order to capture the waveform accurately.

      I know that Dave is aware of this recommendation as well. So I honestly want to know if there’s maybe a document or something that explains how this scope can be specd to 100Mhz with only 200MSa/s.

      Is it the sinc x/x interpolation that allows this to be the case?

      Honest question, sorry if it in any way sounds pretentious.

      Best Regards,
      Jorge Garcia

      • numpty

        Simple:
        It has 100MHz analog bandwidth (i.e. the point at which the amplitude response of the vertical amplifiers is 0,707x that of the low frequency.

        It has a 200Msps ADC architecture.

        Those are actually independent specifications.
        Nyquist just tells you the sampling rate you need to capture an input signal of a given bandwidth.

        You can also get (older) scopes with 20Msps sampling and 500MHz bandwidth – i.e. in order to get a meaningful 500MHz (or even 100MHz) signal, the scope has to capture multiple cycles of the waveform.

        • http://www.cadsoftusa.com Jorge Garcia

          Seeing them as separate specs sounds reasonable. However it seems overkill and unnecessary to give a scope a 500Mhz front end when it can only accurately display in the best theoritical case a 10Mhz signal.

          The times 5 rule I mentioned earlier comes from the fact that from an aribitrary waveform if you truncate it’s fourier series to the 5th Harmonic you’ll have a 97%(If memory serves I read this from a tek app note) accurate waveform.

          Thanks for the response Numpty, I know this is a piece of the puzzle but something still seems to be missing.

          Best Regards,
          Jorge Garcia

          • Wim Libaers

            Nothing is missing. The trick is, if you have a repeating signal (same waveform over and over again), you can slightly shift the sampling point each time.

            For example, the scope gets a trigger signal, and samples on the trigger point, 100 nanoseconds after the trigger point, 200 nanoseconds after the trigger point…

            The next trigger event, it samples 1 ns, 101 ns, 202 ns after the trigger point.

            After 100 triggers, you have your waveform sampled at 1 ns intervals, even though your sampling hardware can only sample once every 100 ns. But of course, it’s essential that you have a nice repeating signal. It doesn’t work for single shot triggers (which is why scopes using this method will sometimes give both repeating signal bandwidth, and single shot bandwidth, with the second one being limited by the sampling hardware).

            • http://www.cadsoftusa.com Jorge Garcia

              Hi Wim,

              I think that was the part I was missing. Nice repeating waveforms is pretty strong requirement though, many of the waveforms one might be interested in vieweing are not repetitive.

              Thanks for the clarification to both of you.

              Best Regards,
              Jorge Garcia

              • David Hess

                The 54622D specifications do not highlight it but its equivalent time sampling rate is 40 GSamples/second.

                Before real time sampling became inexpensive and thereby ubiquitous, most DSOs supported equivalant time sampling to some performance level allowing them to use their full input bandwidth. The requirement for a repeating waveform is only an issue in a minority of applications. This is especially the case when the sample rate is twice the bandwidth. The 100 MHz bandwidth implies a rise time of 3.5 nanoseconds but the sample rate allows capturing glitches down to 5 nanoseconds. Think time domain instead of frequency domain. Finding a single shot signal which could not be reliably captured would be exceedingly difficult.

                My oldest DSO when configured for its maximum bandwidth of 14 GHz (yes, GHz) has a sample rate of about 50 kSamples/second but an equivalent time sampling rate of 100 GSamples/second. Obviously it is completely unusable for single shot events (unless it is being used to simultaneously demodulate) but when you need that kind of capability, single shot events are in the minority of applications anyway.

    • Mark Allen

      I found one of these exact scopes for $650 USD. I was thinking it was a bit much but after watching your teardown I am rethinking my position. Does anyone have an opinion?