Will Keysight upgrade the 2000, 3000T X-Series Oscilloscopes within a few months

[G0HZU]

Dunno if this is of any interest but I did a quick video showing the trace noise/wobble on my old scope when there is an intentionally generated 30dB S/N ratio on the test signal caused by wideband noise.

I scaled it down to 10MHz so to get a 30dB S/N ratio I needed -101dBc/Hz of noise in a 12.5MHz BW.
I produced this noise using a wideband noise source and a step attenuator, a 6dB splitter/combiner and a 12.5MHz LPF. The metal box with all the holes in it is a wideband high level noise source.

12.5MHz BW is 12.5e6 bigger than 1Hz. The marker shows -101dBm/Hz power in a 1Hz BW so the power in that 12.5MHz pedestal is 12.5e6 times bigger.
So the power in the 12.5MHz BW is 10*log(12.5e6) bigger or 71dB bigger.

-101 + 71 = -30dB so the S/N ratio is 30dB. I also checked the S/N ratio using a 20MHz Racal true rms meter and it was close to 30dB.

You can see the noise pedestal on the analyser and then the scope shows how much wobble you see with a 30dB S/N ratio.
I briefly increased the noise to a 20dB S/N ratio on the scope and that causes the brief burst of very bad wobble at 1:02.
Sorry about that, I turned the attenuator the wrong way briefly.

But you see the wobble diminish on the scope when I increase the attenuation to improve the S/N ratio to >= 40dB.
In the last few seconds of the video you can see me swapping back and forth between 40dB and 30dB S/N ratio and the wobble can be seen to almost vanish at the >= 40dB S/N ratio.

If I had done this at 100MHz with a 125MHz BW the wideband noise would need to be at -111dBc/Hz to give this amount of wobble.
If I had done this at 1000MHz with a 1250MHz BW the wideband noise would need to be at -121dBc/Hz.
If I had done this at 3000MHz with a 3750MHz BW the wideband noise would need to be at -126dBc/Hz.

A sig gen set to 3GHz is unlikely to be this noisy unless it was faulty but the scope system noise might be quite high so it might be possible that the cause of the wobble on the Tek scope may be partly due to wideband noise. But it's more likely to be something else IMO.

The video is just a quick and dirty demo so sorry it is so basic and brief...

https://www.youtube.com/watch?v=0vCgQtGJMno&feature=youtu.be

[G0HZU]

The noise source is band limited to 180MHz and I repeated the test with a 180MHz test signal with the 12.5MHz LPF removed from the system. It's as far as I can go with this noise source.

I therefore needed a pedestal of noise at -112.5dBc/Hz in a 180MHz BW to get a 30dB S/N ratio because 180MHz is 82.5dBHz.

When I set the 180MHz BW pedestal of noise down to this level I get the same amount of wobble on the scope on a 180MHz sinewave viewed on the scope. This is because there is the same amount of noise power as in the previous test.

i.e. -101dBm/Hz of noise band limited by a LPF to 12.5MHz BW is -30dBm or 1uW.

-112.5dBm/Hz of noise band limited by the noise source to 180MHz BW is also -30dBm or 1uW.

[G0HZU]

I also tested several of my GHz++ sig gens here for wideband noise floor when set to 2.4GHz and most were OK.
The worst offender was the HP tracking source designed for use with my HP8566B analyser. It gave a noise floor of about -127dBc/Hz right across a 3GHz BW. Not good at all.

The next poor performer was my HP/Agilent ESGD 4433 as this gave a noise floor of about -136dBc/Hz right out to 3GHz when set to 2.4GHz. Not good but the ESGD sig gens are economy sig gens and not really lab grade in terms of noise and spurious.

My other (newer model) ESGD 4433B was several dB better and the Marconi 2024 was much cleaner. The signal source from my E5071B VNA was very clean as well (in terms of broadband noise floor) when set to a cw signal at 2.4GHz. The old HP83752A 20GHz synthesised sweeper I have here was best of all. Very clean and I didn't bother to see how much cleaner. I suspect it is in the order of -150dBc/Hz. This was a nice surprise as I've never tested it like this before and I expected it to be quite noisy.

http://www.keysight.com/en/pd-1000001888%3Aepsg%3Apro-pn-83752A/synthesized-sweeper-001-20-ghz?cc=US&lc=eng

At work we made the mistake of buying some HP8648D sig gens about 12 or so years ago. These have a poorly designed levelling system which means they churn out a very high noise floor when set to levels in the -20 to -10dBm range across the HF/VHF range. Easily 20dB worse than a Marconi 2024. They were eventually banned from the labs because so many people were getting caught out by the high noise level when doing various receiver (and amplifier) tests with them.

[KE5FX]

Interesting stuff, all right.  It does make sense that a 30 dB SNR would cause display jitter on the order of a few percent, enough to be seen on a scope display.  I'd expect a trace jumping around on a scope display to look even worse, since it would be perceived as a "peak to peak" phenomenon.

There would be 95 dB/Hz of excess noise in a 3 GHz bandwidth, and since you've established we can 'see' a 30 dB SNR, a broadband signal source with -125 dBc/Hz or more would indeed be noisy enough to influence the TDS 694C's display.  At offsets beyond 1 MHz the 8672A I used should be in the -140 to -150 dBc/Hz range, so I still don't believe it contributes to the effect seen in the video.  The same is likely true for the aperture jitter of the ADCs in the scope themselves.  Their 10 GHz clock source, on the other hand, is an off-the-shelf ZComm VCO.  It might actually be one of the bigger contributors to the jittery display.

[nfmax]

Interesting. One thing struck me as slightly odd. As far as I can tell from the display (I'm not familiar with this scope model), the scope trigger level was set at 0V, and the delay set at 0s, so the trigger instant should be at the centre of the screen. Yet it was not, the +ve-going crossing was slightly early. Am I mistaken, or is there something wrong?

[nctnico]

This is normal because most oscilloscope triggers will have some hysteresis to avoid triggering on noise. At high frequencies close to the maximum bandwidth it often needs some fiddling with the trigger level to get a stable trigger because the sensitivity of the trigger will be significantly less.

[nfmax]

Agreed. But trigger hysteresis shouldn't affect the trigger threshold - i.e. if I set a threshold at 0.1 division, +ve going, then I would expect the trigger instant to be when the signal rises through the 0.1 division mark. If the trigger hysteresis was set to 0.3 division, I would expect the signal to have to rise to > +0.4 division and then fall to < -0.2 division before a new trigger could be recognised (at the 0.1 divison, +ve going point). Hysteresis shouldn't affect the trigger point.

[G0HZU]

I powered the HP54540C scope up again today and had a look through the settings and found it had a trigger offset of about 40mV programmed in (I think it was 43.75mV). Not sure how this happened but one of my beefs with the scope is that it is really easy to alter something by accident because it has multifunctions on the rotary controls and they are very sensitive. I'm guessing that I knocked the trigger offset away from zero somehow.

I tried doing a Recall CLR to default the scope and the offset now shows zero and the scope triggers in the centre. However, if I go down to 500ps/div the trace is still ever so slightly off to one side. Not sure if this an interpolation issue or if it would go away with a system calibration but I'm not too fussed really. It is only a trace width or so off centre at 500ps/div.

But yes, the scope display was a bit odd in terms of the trigger setting in the video on 50ns/div. You can see it is triggering at the programmed offset that was about 40mV. Sorry about that.

I really don't like scopes with shared controls and I try not to use this scope much. It was a freebie scope salvaged last year from the skip at work.

[Keysight DanielBogdanoff]

This is normal because most oscilloscope triggers will have some hysteresis to avoid triggering on noise.

Yes, you can often see this if you change the trigger mode to "rising OR falling edge." The intersection of the signals won't be right at the 0 time point because the scope can't vertically compensate for both a rising and a falling edge hysteresis offset.  In just rising or just falling edge mode, the scope will position the hysteresis so that the signal comes through the dead center of the screen.

[G0HZU]

At offsets beyond 1 MHz the 8672A I used should be in the -140 to -150 dBc/Hz range, so I still don't believe it contributes to the effect seen in the video.

Yes, the HP8672A looks like a decent generator. I've not used one but the specs look good and it would have to be faulty to generate that much broadband noise. When I tested my HP54540C up at 600MHz with several sig gens here I took the precaution of using a very narrow 600MHz BPF inline just to be safe but it made no difference with/without the filter as the scope only has 500MHz BW and my regular sig gens are all clean enough for this test.
When tested at 600MHz this scope is beyond the 500MHz -3dB BW stamped on the front panel but it is still just inside -3dB at 600MHz when tested. It shows a fair bit of wobble on the sine wave when tested up here though. Very similar to your Tek scope display.

[siggi]

Dunno if this is of any interest but I did a quick video showing the trace noise/wobble on my old scope when there is an intentionally generated 30dB S/N ratio on the test signal caused by wideband noise.

On a tangent, it should be possible to get an idea of the noise characteristics of the front end of the TDS scopes by looking at the signal feed-through. My TDS784D has channel 3 fed through to the back. Looking at the spec, though, man: "11mV/division +-20% into a 50Ohm load.", that's pretty loosy-goosy.

It also looks like the 694C doesn't have that feed-through.

[G0HZU]

I took a quick video of my HP54540C (500MHz BW and 2GS/s) with it trying to measure the 500Hz squarewave test waveform from the scope. This waveform can be used to set up the probes on the scope etc.

It has a very fast edge on the waveform and you can see in the video that 2GS/s isn't a high enough sample rate because you can see overshoot/wobble when sampling at 2GS/s. It gets even worse at 1GS/s.

However, the waveform cleans up nicely when the Repetitive mode is selected. The waveform loses the undershoot on the start of the leading edge and just looks better.

So in this case, having a scope branded as having 500MHz BW and sampling at 4 times the BW isn't enough to prevent waveform issues. There aren't enough samples to prevent corruption from aliasing.  This waveform is a 500Hz squarewave but it has frequency content that goes well above 1GHz. The Repetitive mode has a much higher effective sample rate and so this gives a much nicer result.

It's taken with a mobile phone so it is a bit flakey but I think it is clear enough from this demo that there are times when a 4:1 sample rate to BW ratio isn't enough when you feed a real signal to a real scope

Note that there is a trigger offset on this video again. I used default and autoscale to set up the scope and set CH1 to 50R input. But it has added an offset again so it doesn't trigger in the centre.

https://www.youtube.com/watch?v=8yz662DeElA&feature=youtu.be

I also looked at the test signal on a spectrum analyser and it shows frequency content that extends well above 1GHz. This should be no surprise as the risetime of the waveform is probably less than 500ps. Too fast for this scope to measure accurately. But the lowpass filtering in the analogue front end of the scope isn't steep enough to prevent the >1GHz frequencies from reaching the sampler stage in the scope.

[nctnico]

If this scope has equivalent time sampling then it doesn't surprise me the analog front end filter doesn't roll off very steep (probably a Gaussian response).

[G0HZU]

I think some of the wobble effect is from the interpolation when in real time mode. I'm guessing that the fast risetime means that the interpolation can't cope very well with the limited qty of samples. So you get a wobbly effect and the introduction of the negative bump at the start of the leading edge?

[nctnico]

When the frequency contents get close to fs/2 (say >0.42 fs) then the interpolation could get upset.

[G0HZU]

It wobbles on a clean sinewave at 600MHz with 2GS/s as well and this is well below Fs/2. It basically looks the same as the Tek scope. I've seen this on other scopes at work too.

A real scope like this one will have noise, (and maybe alias terms on complex signals) and sampling/timing jitter to cope with. It will have real world filters in it and this will all mean that the interpolation result will reflect all these factors. So you can't expect it to draw an accurate sinewave once you get this high in frequency.

It's going to affect the amplitude and also it will introduce distortion into the overall shape of the sinewave.

[HackedFridgeMagnet]

With the release of the new R&S RTB2000 series hopefully Keysight will bring out something soon.

I'm wanting to buy very soon, is it worth waiting Keysight?