EEVblog #610 - Why Digital Scopes Appear Noisy - Part 2

[G0HZU]

I had a play with a Tek MSO4104 at work today as well. I connected up the noise source via a high order 12MHz LPF. With the scope terminated in 50R and set to 100mV/div it showed 4 boxes peak to peak with the intensity set to fully opaque. i.e. the very skinny peaks of the noise covered about 4 boxes.

If I did the same test on the 465 with a reasonable brightness setting the (visible to the human eye) noise covered 3 boxes peak to peak at 100mV/div.

On a power meter the LPF filtered noise source measured -15.0dBm (50R) = 39.8mV rms.  On my old Racal true rms meter the noise measured just under 40mV rms.

The MSO4104 measured about 39mVrms which was a pretty good result.

My initial attempt to measure the noise figure of the MSO4104 came up with conflicting results depending on the bandwidth mode selected. The best result appeared to be in full bandwidth where it appeared to show a system noise figure of approx 28.6dB when set to 5mV/div. 

[rf-design]

I would like to see that noise measurement made on both an analog oscilloscope and several different modern DSOs which have a couple of ways to measure it directly without relying on the visual waveform rendition.  Two of my old DSOs should be able to do it but I lack a good test signal.

"which have a couple of ways to measure it directly"

So you place a cursor somewhere and get a rms number or in second popup window the distribution?

Give me an example.

Those two ways will work but would usually be applied over a window that could include the entire waveform record instead of multiple acquisition records with the measurement made at one point.

What would be interesting to see how they change depending on sample rate and record length since neither should except insofar as the total time acquired limits the low frequency noise which may or may not be significant depending on the source and noise bandwidth.  If you want to include noise at 0.1 Hz, then the acquisition has to be over at least 10 seconds.

My suspicion is that oscilloscopes that rely on post DSP processing to correct their frequency and phase response are going to have problems making this measurement accurately.  Anti-alias filtering is really going to screw it up.

My question is regarding if there is actual any scope which have one of our extensive discussed features. I know that these are possible but did not have the market overview and use only a limited number of scopes.

Regarding the rms noise or distribution at a specific point you could use only the triggered part of the sample points. Otherwise there is no overlay or rms statistic possible. If you want the hole sample you have to eleminate the trigger bling time. That is at the highest sample rate 90% and more. There are other concepts. For example software backprocessing with powerful machines which could nearly eleminate blind time by software triggering and subsample reconstruction and overlay. The record length then is the smallest interval between trigger conditions. Required bandwidth for 4 channels is then somewhere between a modern CPU and a GPU memory interface.

[robrenz]

.... This is not addressing the actual analog front end noise level of a top end analog scope vs a top end digital scope.  An old 7A22 differential amp has 16µV or 0.1 div whichever is greater displayed noise at 10µV/div and upper BW limit of 1MHz. Can you BW limit the Agilent to 1MHz and see what the noise floor is for comparison?

Answering my own question:
Infinivision 6000 X series  115µVrms (1mV/div) at 1GHz
Infiniium S-series  90µVrms (1mV/div) at 1GHz

Obviously better than the old school when you consider its at 1000X the bandwidth limit

[David Hess]

.... This is not addressing the actual analog front end noise level of a top end analog scope vs a top end digital scope.  An old 7A22 differential amp has 16µV or 0.1 div whichever is greater displayed noise at 10µV/div and upper BW limit of 1MHz. Can you BW limit the Agilent to 1MHz and see what the noise floor is for comparison?

Answering my own question:
Infinivision 6000 X series  115µVrms (1mV/div) at 1GHz
Infiniium S-series  90µVrms (1mV/div) at 1GHz

Obviously better than the old school when you consider its at 1000X the bandwidth limit

It does not scale that way.  The 7A22 suffers from low frequency noise (1/f) in a way that wideband noise in higher bandwidth instrument will conceal.  If you limited the high bandwidth instrument to the same low frequency range, it will not have proportionally lower noise.  To put it another way, limiting a high bandwidth oscilloscope to the bandwidth of a 7A22 will not yield the same or better performance without adopting the design of the 7A22.

[David Hess]

My question is regarding if there is actual any scope which have one of our extensive discussed features. I know that these are possible but did not have the market overview and use only a limited number of scopes.

I am not longer sure what you mean.

Most (or all now?) DSOs have an RMS measurement function which can be applied over the whole acquisition record or a selected part.  Some can calculate the standard deviation of the DC value over the record which should produce the same result.  Either should produce the same result as the analog oscilloscope tangential noise measurement and it should not very significantly with sample rate or record length except insofar as they affect the low frequency cutoff which will usually be minor.

Regarding the rms noise or distribution at a specific point you could use only the triggered part of the sample points. Otherwise there is no overlay or rms statistic possible. If you want the hole sample you have to eleminate the trigger bling time. That is at the highest sample rate 90% and more. There are other concepts. For example software backprocessing with powerful machines which could nearly eleminate blind time by software triggering and subsample reconstruction and overlay. The record length then is the smallest interval between trigger conditions. Required bandwidth for 4 channels is then somewhere between a modern CPU and a GPU memory interface.

The noise can be measured with one acquisition record and the total time that the record encompass sets the low frequency cutoff so the dead time between acquisitions is irrelevant and triggering is not even necessary.  You cannot very well see 0.1 Hz noise unless the record lasts at least 10 seconds.  High frequency noise aliases to below the Nyquist frequency and gets included into the measurement which is why having a long record length to support a high sample rate is not necessary.

This is just one of those measurements that a DSO should make easy if it can be relied on compared to an analog oscilloscope.  The later can reliably make the measurement but it is time consuming and not a real time operation unless you just want to see relative levels.  At very low frequencies, I would not even want to try it on an analog oscilloscope although I have done it using a good digital voltmeter.

[rf-design]

I am not longer sure what you mean.

Most (or all now?) DSOs have an RMS measurement function which can be applied over the whole acquisition record or a selected part.  Some can calculate the standard deviation of the DC value over the record which should produce the same result.  Either should produce the same result as the analog oscilloscope tangential noise measurement and it should not very significantly with sample rate or record length except insofar as they affect the low frequency cutoff which will usually be minor.

Sorry David,

I did not make it clear enough. I mean the rms-noise at a specific time related to the trigger event not the rms over the hole trace lenght of a triggered waveform. Clearly everyone could do an average or rms over a trace lenght but we discuss in relation to this thread topic the vertical noise and noise distribution over a number of triggered waveforms. These are overlayed in some manner by DSO or combined by phospor decay time and our eyes in analog scopes.

BR
Reiner

[David Hess]

I am not longer sure what you mean.

Most (or all now?) DSOs have an RMS measurement function which can be applied over the whole acquisition record or a selected part.  Some can calculate the standard deviation of the DC value over the record which should produce the same result.  Either should produce the same result as the analog oscilloscope tangential noise measurement and it should not very significantly with sample rate or record length except insofar as they affect the low frequency cutoff which will usually be minor.

Sorry David,

I did not make it clear enough. I mean the rms-noise at a specific time related to the trigger event not the rms over the hole trace lenght of a triggered waveform.

I have done this by using either a gated measurement or delayed acquisition but that only makes the measurement on a single acquisition at a time instead of over multiple acquisitions.

DSOs that accumulate acquisition records into a DPO type display may be able to do this when combined with gated measurements.

DSOs that support histograms and statistics may be able to generate them over multiple acquisitions at one point or small area of the waveform but offhand I do not know of any that can do this.  I suspect that would be a pretty high end feature.

Clearly everyone could do an average or rms over a trace lenght but we discuss in relation to this thread topic the vertical noise and noise distribution over a number of triggered waveforms. These are overlayed in some manner by DSO or combined by phospor decay time and our eyes in analog scopes.

I do not think this has any meaning.  On an analog oscilloscope, multiple sweeps do not alter the measured value of noise because the sweeps are correlated in time.  If they did, then changing the sweep rate would change the measured value outside of the change in low frequency cutoff which is usually an insignificant effect. (*)  They make the trace "brighter" but do not change the distribution.

For a DSO it would depend on how the separate acquisitions are combined or measured in total.

(*) If I use a low-pass filter of say 10 Hz so I am only measuring low frequency noise and I change the sweep or acquisition length so a period of 10 seconds is acquired instead of 1 second, the noise definitely increases.  If I do the same thing without the high frequency cutoff filter, the change is not significant because the broadband noise swamps any extra contribution.  This is a common type of measurement for characterizing flicker noise which is difficult to filter out.  It is also inaccurate, difficult, or impractical to do on an analog oscilloscope.

[David Hess]

The rear vertical output does not necessarily match the bandwidth on Tektronix oscilloscopes (and it will be really slow if not terminated into 50 ohms) and it may only be roughly calibrated.  In addition, since the single ended rear vertical output signal is generated after the input signal is converted to a differential signal, it may suffer from excessive common mode noise.  My 7603 (just like in the video) had a problem with that until I made a design change to improve the common mode rejection of the conversion by more than 20 dB.

I like using the vertical output signal on my Tektronix 7000 mainframes so I can use my 7A22 and 7A13 differential vertical amplifiers with my DSOs.

On my 465 the rear CH1 output port appears to have a pretty flat response out to about 100MHz Both the noise floor and the signal response looks good to 100MHz. It isn't calibrated but I can use the analyser to calibrate its 'gain'. I can also use the analyser to measure the S/N ratio and therefore predict the front end noise figure when set to 5mV/div and driven from a 50R source.

I am sorry that it took longer than I expected to respond.  I got distracted and then Firefox crashed losing my initial reply.

On other Tektronix oscilloscopes with vertical outputs that is not necessarily the case which is why I mentioned it.

Note: I'm terminating the output with the 50R input impedance of the spectrum analyser. Also, I found that selecting 20MHz BW on the front panel has no effect on the noise or signal response of the rear panel port because it is tapped off just before the part of the scope that limits the bandwidth to 20MHz when this feature is selected. So the bandwidth at the rear port can't be changed.

The bandwidth limiting for portable oscilloscopes usually occurs after the channel switch so it applies to all inputs and there are no issue with matching between channels.  I think this also helps with overload recovery.  Since the trigger pickoffs occur before the channel switch and the vertical output happens in parallel with the trigger pickoff, triggering and the vertical output are not affected by bandwidth limiting.

There are some weird exceptions though:

The 2230 DSO is designed like earlier portable oscilloscopes which as a side effect means that DSO mode and triggering is not affected by the bandwidth limit.  The 2232 which replaced it fixed this issue by moving the bandwidth limit to a spot earlier in the signal path and duplicating it for each channel.  As a side effect though, its bandwidth limit affects DSO mode *and* triggering.

The 7000 mainframes implement the bandwidth limit before the internal channel switch because of their modular nature; their vertical plug-ins do not have access to that point.  But since the signal and trigger paths separate in the plug-ins, some of them bandwidth limit the trigger and some do not.  Even worse, on at least one plug-in, it depends on if it is an early or late version.  The 7000 vertical output is picked off from the trigger signals after they are picked off from the vertical signals and selected which is nice since you can route any vertical signal to the vertical out with some limits whether it is displayed or not.

I don't know how much the rest of the scope path degrades the overall noise performance but the numbers suggest that the scope front end section has about a 20dB noise figure when set to 5mV/div. That seems pretty good to me.

I only mentioned it because it is something to watch for.  The differential signal is converted to single ended for the vertical output and may suffer from low common mode rejection.  My 7603 had a problem with this as the design only has 21 dB of common mode rejection leading to power supply ripple showing up on the vertical output at levels much higher than the noise level.  From the schematic, it looks like the 465 is a little better at 30 dB.

The consensus I have seen is that the vertical output is intended for things like frequency counters where any noise and exact calibration is unimportant.

I am including photos of the vertical output noise on my 7603 before and after I improved the design a little bit.  Note that the broadband noise level is not shown.

[robrenz]

I like using the vertical output signal on my Tektronix 7000 mainframes so I can use my 7A22 and 7A13 differential vertical amplifiers with my DSOs.

Great minds think alike   Here is a post of my 7603 with 7A22 feeding my Rigol. I didn't notice any great noise as seen in the 16µV PP noise sent to the Rigol

What is giving the cursor readouts on your 7603? I want one

[David Hess]

I like using the vertical output signal on my Tektronix 7000 mainframes so I can use my 7A22 and 7A13 differential vertical amplifiers with my DSOs.

Great minds think alike   Here is a post of my 7603 with 7A22 feeding my Rigol. I didn't notice any great noise as seen in the 16µV PP noise sent to the Rigol

I already had a 7603 mainframe and decided that a 7A13 and/or 7A22 was the most cost effective solution when combined with the vertical output signal to add high performance differential probing to my DSOs.  The 7A13 is just incredible to use.  The 4 slot 7000 mainframes that I acquired later do not have the same vertical output issue but they take up more space so I prefer to use the smaller 7603 which has a larger although not as sharp CRT anyway.

What is giving the cursor readouts on your 7603? I want one

The oscillograph is from my 2230 or 2232, I do not remember which, measuring the 7603 vertical output signal directly and showing power supply ripple from the +15 volt supply getting into the vertical output because of low common mode rejection of the differential to single ended conversion circuit.  I verified that other 76x3 oscilloscopes behave the same way and that it was not just mine.  Wideband noise is not present because it is was removed by low pass filtering and averaging.

Keep in mind though that the shown 6.2 millivolts of peak to peak power supply ripple amounts to about 1.24% of a CRT division on the 7603 which would correspond to 24.8 microvolts at the input when set to 2 millivolts/division.  Under normal conditions, it would be insignificant and that noise is only present at the vertical output.  The CRT would reject it since it accepts balanced drive.  The trigger might see it but again, it is insignificant and below the specified trigger threshold anyway.

That was mostly a test to verify where the problem was.  The circuit change is simple as shown but I may make a more comprehensive alteration in the future.

[G0HZU]

Can I ask a general question about how you are all taking images of the CRT screens? The camera used by Dave in the video blog is very impressive because it gives very lifelike images. The still images taken by Dave Hess also look very natural as well.

Do you do anything special with the lighting or camera settings before taking these images? Both of my cameras struggle to take good images of CRT displays. One camera is very old but the other is a fairly new and decent mid range camcorder (Panasonic) that normally has very good image quality.

I did a few comparisons between my TDS2012 and the 465 in terms of noise performance and it would be nice if I could post up better quality images of the 465 CRT. So do you guys use any clever lighting etc?

[David Hess]

Can I ask a general question about how you are all taking images of the CRT screens? The camera used by Dave in the video blog is very impressive because it gives very lifelike images. The still images taken by Dave Hess also look very natural as well.

Do you do anything special with the lighting or camera settings before taking these images? Both of my cameras struggle to take good images of CRT displays. One camera is very old but the other is a fairly new and decent mid range camcorder (Panasonic) that normally has very good image quality.

I had the same problem using an old Sony DSC-S70 and solved it by buying a better camera in the form of a Canon SX150.

I did a few comparisons between my TDS2012 and the 465 in terms of noise performance and it would be nice if I could post up better quality images of the 465 CRT. So do you guys use any clever lighting etc?

I have just been taking the shots by hand in automatic mode under random conditions and they turn out better than the older Sony camera which I was using under controlled conditions.  At some point I would like to build a camera hood for my Tektronix oscilloscopes which are designed to accept one:

http://www.amplifier.cd/Technische_Berichte/Photographie_Oszilloskop/Foto_Halterung.html
http://www.cembreros.jazztel.es/camerascope/camscope_en.htm
https://www.eevblog.com/forum/projects/quick-and-simple-oscilloscope-camera/
http://paulorenato.com/joomla/index.php?option=com_content&view=article&id=92

[cbattles]

Sorry to resurrect an old thread, but I have a question.  Forgive me if this is the wrong place, but I'm new to the forum and am still working my way up from episode #1 (and this is as far as I've gotten :p)

With the advent of "digital phosphor" and the higher BW, sampling depth, update speed, etc., of the current lower end digital scopes, is there still a need (or a desire) for an analog scope for general electronics work?  Especially for the hobbyist?  Basically should I spend $700-$900 on a nice used Tek 2455/65 or go with something like the Siglent 2000 series or the Rigol 2000 series, which I can get new for a similar price, at least in a 2ch version.

I'm not an electronics guy (I actually have a degree in Mechanical Engineering) and I'm not a kid anymore (pushing 44) but I'm becoming interested in electronics as a hobby.  I currently have a freebie tek TDS210 (yes it is 'slow as a wet week') and am looking to augment/upgrade.  I have a couple of projects planned, including a simple tube guitar amp and a TTL single tube nixie clock.  I'm having fun breadbording and playing around for the past year (keeps me out of the bars ), but I would like to have some better equipment.  Call it a mid life crisis if you want   I actually recently purchased a nice BK 393 DMM to replace my freebie Omega scope, recently realized I now need an LCR meter (salvaging parts from an old power supply for my junk bin and have no way to measure the inductors) and I think it may spiral a bit from here....  But I digress.

Having watched 600+ episodes so far in the past few months, spanning 5 years of real time, It appears to me that the original argument of the true benefit of the "analog scope for beginners" was 2-fold:  1.  Cheap.  Yes you can find a nice $50 analog scope on ebay anytime you look.  And 2:  The visual persistence of a CRT simply blew away the digital scopes before the advent of "digital phosphor".  But when you get into the $800-$900 range it seems that the playing field may have levelled and I would like the opinions of the real experts.

Thanks for your time,
Chris

[EEVblog]

With the advent of "digital phosphor" and the higher BW, sampling depth, update speed, etc., of the current lower end digital scopes, is there still a need (or a desire) for an analog scope for general electronics work?

No, essentially none.

  Especially for the hobbyist?

You could argue that an analog scope provides an easier insight into how an oscilloscope works, but it's a moot point.

  Basically should I spend $700-$900 on a nice used Tek 2455/65 or go with something like the Siglent 2000 series or the Rigol 2000 series, which I can get new for a similar price, at least in a 2ch version.

Without question the digital scope.

It appears to me that the original argument of the true benefit of the "analog scope for beginners" was 2-fold:  1.  Cheap.  Yes you can find a nice $50 analog scope on ebay anytime you look.  And 2:  The visual persistence of a CRT simply blew away the digital scopes before the advent of "digital phosphor".  But when you get into the $800-$900 range it seems that the playing field may have levelled and I would like the opinions of the real experts.

Correct!