Simple Technique to measure Waveform Update Rates: DSOs w/either Edge Triggering

[ivan747]

Actually it made me think an alternatives using double pulse train to do the same thing. Say, the 1st pulse has a width of 10ns and the 2nd, 20ns. The freq of the double pulse train needs to be well below the expected update rate. Now, one can adjust the time interval btw the two pulses from small to large. The moment the 20ns pulse starts to show up at the trigger point, it defines the update rate.

Yes, this is a known (and published) technique    I was trying to come up with something easier.

Where can I find information about this technique?  I've been messing with a microcontroller to produce a pulse train described above, but the counter on the scope (rigol) would count both pulses even if it doesn't flicker back and forth between the the two different sized pulses.

My interpretation of what he meant is on page 5, you can take a look at that. Maybe it's not the correct interpretation but it works for our purposes.

@Harvs: I will confirm if my DS1102E shows the same update rate with the method I described earlier. I'm not opening my oscilloscope. Too expensive for me to void its warranty. By the way, what firmware are you using? And please try Marmad's method to see if it works, because on my FW 00.04.00 it doesn't work, so I made a work-around. It's described in page 5 of this thread.

And wouldn't an isolated trigger output be a sweet mod to try? 

[marmad]

Where can I find information about this technique?  I've been messing with a microcontroller to produce a pulse train described above, but the counter on the scope (rigol) would count both pulses even if it doesn't flicker back and forth between the the two different sized pulses.

I described it in-depth before in another thread here - and I think you can also find similar techniques written about elsewhere online - although I don't have any links offhand.

BTW, the method referred to in the link describes using a dual-output function generator - but that is simply to make it easy to adjust the distance between the pulses. If you are generating the pulses with software (e.g. a microcontroller), a single output is fine.

[marmad]

I had a frequency counter on the external trigger out while triggering on a 27MHz waveform, is that an OK way to measure? I'm still pretty new to scopes, is this update rate pretty good?

Sure, that's the easiest (and best) way to measure. And yes, those are definitely good update rates - perhaps the Tek TDS-700 series were the top of the wfrm/s  heap before the Agilent 3000/4000 X-Series came along?

[Harvs]

Thanks Marmad.  I gave it a go, but the results weren't particularly conclusive.  I'm probably doing something wrong, I assumed that by feeding in a 40MHz siginal to trigger it, then I would be able to trigger it every time.  But I get different minimums depending on whether it's 40MHz or 35MHz.

Anyway, I think for the DS1000 series I've probably put about as much effort into this one as I'm going to.  I've got some reasonable results, I know that above about 40 wfms/s it'll start dropping the odd trigger (which is what I saw in the original method.)

The trigger out on the other hand could be worth putting some time into.  I don't think isolation is needed, as for anything I can think of it'll be triggering another mains earth reference device.  But a high speed 50ohm driver would be a definite requirement.  But yeah, that would need to go to the end of the list of other projects I've got on the go at the moment.

[KedasProbe]

On the DSO with 'alternate', the waveforms per second becomes a third of the maximum rate - missing or ignoring 2 triggers between each acquisition.
On the DSO with 'either', the waveforms per second becomes half of the maximum rate (as shown in the video)  - missing or ignoring 1 trigger between each acquisition.

If you would have to choose between 2 then alternate is better, you do sometimes (not always) miss a 'trigger' but it guarantees than you will always see both edges on your screen for all frequencies. But obviously your method needs a missing edge on the display for some frequencies.

[darrylp]

Yep, it just goes to show how much faster the DS2000 is.  It may well have varied by 1 or 2ms, but the overall time is so long it's hard to notice properly.

But I am curious to know what the DS1000E series wfrm/s rate is in the 50ns - 500ns range (where most DSOs hit their peak speeds); I've heard many figures bandied about.   

Results I got, DS1052E hacked to 1102,

500ns/div 36 or 37 wfms/s
50ns/div 39 wfms/s

@Harv, are you sure you didnt miss the flickering at approx half these values ?   are these number from the on screen counter ?  which is twice the input freq. ?

my posted 20Hz input was showing 40Hz on the DSO counter.

thus, thinking your 39wfms is ( close to my 20 ... if you were using the scopes numbers that is ;-) )

oh dear, now being awake... I realise my posts last evening here in the UK,  it was early morning, and being tired you shouldnt be posting tech info !

20Hz input double ( edge triggered ) and detected by the scope was like 40Hz on its counter, and that indeed is what the scope saw, and updated on. so it was updating at 40wfms, not my half posted values of the input. so i was seeing ~40wfms

sorry for my confusion !

Darryl

[marmad]

oh dear, now being awake... I realise my posts last evening here in the UK,  it was early morning, and being tired you shouldnt be posting tech info !

Ah..yes... I have suffered the same fate too many times to count.  No worries - that's what Modify and Remove are for! 

[tinhead]

Where can I find information about this technique?  I've been messing with a microcontroller to produce a pulse train described above, but the counter on the scope (rigol) would count both pulses even if it doesn't flicker back and forth between the the two different sized pulses.

I described it in-depth before in another thread here - and I think you can also find similar techniques written about elsewhere online - although I don't have any links offhand.

BTW, the method referred to in the link describes using a dual-output function generator - but that is simply to make it easy to adjust the distance between the pulses. If you are generating the pulses with software (e.g. a microcontroller), a single output is fine.

here

http://www.storeinfinity.com/blog/Waveformrefreshrate

[marmad]

On the DSO with 'alternate', the waveforms per second becomes a third of the maximum rate - missing or ignoring 2 triggers between each acquisition.
On the DSO with 'either', the waveforms per second becomes half of the maximum rate (as shown in the video)  - missing or ignoring 1 trigger between each acquisition.

If you would have to choose between 2 then alternate is better, you do sometimes (not always) miss a 'trigger' but it guarantees than you will always see both edges on your screen for all frequencies.

Spoken like someone who's DSO does 'alternate'.    Seriously though, this isn't a contest - and both methods have their advantages. The 'either' method guarantees a higher wfrm/s rate: for example, at 5us/div on my DSO - given a 5kHz waveform input frequency - my wfrm/s rate is 5000 - but it would only be 3750 using the 'alternate' method.

Ideally you would be able to chose the method you prefer, but only a few DSOs offer both options (Agilent X-Series, LeCroy WaveSurfer, some R&S DSOs, etc). My personal feeling is that if I don't have a choice, I would prefer that the DSO doesn't perform 'logic' on my triggers - so I would favor the faster wfrm/s. If I just happen to be observing an exact waveform frequency that lands in the blind time (like the above example: causing me to lose an edge - but maintain a higher wfrm/s rate), a simple menu button press allows me to alter which Edge I'm looking at. Also, the DS2000 has 16 different trigger types, so if I need to see both edges at the same time I have other options. 

[KedasProbe]

I would be a nice feature if the scopes just display the time between the last sample and the next first sample (can't be a hard thing to do in the firmware). Because that's what you want to know, how much time (or %)  am I missing on the screen.
Different settings are changing this value so I would almost think it's a must have when you measure something, so why aren't they adding it.

[marmad]

I would be a nice feature if the scopes just display the time between the last sample and the next first sample (can't be a hard thing to do in the firmware). Because that's what you want to know, how much time (or %)  am I missing on the screen. Different settings are changing this value so I would almost think it's a must have when you measure something,

Absolutely! Very easy to do in firmware, I would think - although sometimes it's bouncing around a lot - so with some kind of averaging would probably be best.

so why aren't they adding it.

Why do you think? Did you notice the posted speeds of the DS1052E? They don't really want comparisons.

[alank2]

If you're not sure if your DSO supports every Edge, you can test very quickly by setting it to 10ms/div, then sending it a square wave of 1, 2, 3, 4, then 5Hz. If one of those frequencies does NOT produce a stable, non-moving waveform, then your DSO is skipping potential triggers - and this technique won't work.

I tried this last night with older firmware (DS1102E) and I couldn't get a stable waveform.  I updated to newer firmware and the same result.

10ms timebase, edge trigger, slope has both up and down arrows, sweep normal, sensitivity 0.1div, holdoff 500ns

1 Hz to 5 Hz square wave all produce a back and forth waveform where it captures the rise on one and then the fall on the next, like a rail road crossing alert...

Thanks,

Alan

[marmad]

I tried this last night with older firmware (DS1102E) and I couldn't get a stable waveform.  I updated to newer firmware and the same result.

And yet some here have posted that it works on (at least certain versions) of the DS1052E FW. Strange.

[Harvs]

If you're not sure if your DSO supports every Edge, you can test very quickly by setting it to 10ms/div, then sending it a square wave of 1, 2, 3, 4, then 5Hz. If one of those frequencies does NOT produce a stable, non-moving waveform, then your DSO is skipping potential triggers - and this technique won't work.

I tried this last night with older firmware (DS1102E) and I couldn't get a stable waveform.  I updated to newer firmware and the same result.

10ms timebase, edge trigger, slope has both up and down arrows, sweep normal, sensitivity 0.1div, holdoff 500ns

1 Hz to 5 Hz square wave all produce a back and forth waveform where it captures the rise on one and then the fall on the next, like a rail road crossing alert...

Thanks,

Alan

10ms/div is a bloody long time.  There's no point measuring at this level as it'll be mostly all acquisition time.

If you have a look at the tables I've produced earlier, use these as a starting point (i.e. known value) and you should be able to wind it back from there and find it works fine.  The real problems are encountered when you are triggering off integer multiples and don't know it.

EDIT: Sorry I just read the post you were replying to.  That came across the wrong way.   The method works fine though with the DS1000 series (within reason.)

[alank2]

So I switched to a sine wave, and here it is stable:

1Hz   stable
2Hz   not stable
3Hz   stable
4Hz   not stable
5Hz   stable

Triangle works the same as sine.  Back to square wave and all are not stable.

[Harvs]

If you're not sure if your DSO supports every Edge, you can test very quickly by setting it to 10ms/div, then sending it a square wave of 1, 2, 3, 4, then 5Hz. If one of those frequencies does NOT produce a stable, non-moving waveform, then your DSO is skipping potential triggers - and this technique won't work.

I haven't properly read this before, and I'll caveat what I'm about to say by the fact it's getting late here.

But this seems to assume that your scope is going to capture an even multiple of 10ms divisions right?  Or am I off here?

Unlike on the DS2000, the DS1000 hardly ever displays all of the data captured.  In fact some time bases are effectively just zooms of others (i.e. they have the same capture length at the same sample rate, and just reduce the amount that it shows on screen.)  So the timebase and screen divisions is not an effective measure of the acquisition time.  As I've said earlier in the thread, if you need to know the acquisition time on a DS1000, just hit the menu button to get the sample rate, and divide sample memory depth by it.

[ivan747]

So I switched to a sine wave, and here it is stable:

1Hz   stable
2Hz   not stable
3Hz   stable
4Hz   not stable
5Hz   stable

Triangle works the same as sine.  Back to square wave and all are not stable.

Seems like I am the only one that cannot get this to work on my DS1102E. I still think it's the firmware.

Can anyone confirm Marmad's method doesn't work on a DS1000 series scope with a 00.04.00 firmware?

Thanks

[marmad]

10ms/div is a bloody long time.  There's no point measuring at this level as it'll be mostly all acquisition time.

The 10ms/div setting is just a simple test point to see if your DSO does 'either' or 'alternating' Edge triggering. You could test at any time base setting - but the slower ones only have a limited range of frequencies before one of the edges HAS to hit the blind time and then vanish ('either') or stay visible ('alternating').

The method works fine though with the DS1000 series (within reason.)

But the point is, he has one of the DS1000 series - and he's saying it doesn't work with his model/firmware combination.

[marmad]

But this seems to assume that your scope is going to capture an even multiple of 10ms divisions right?  Or am I off here?

No, this has nothing to do with divisions per se - it's only related to the acquisition cycle time - which, as I mentioned in the video, is always the longer of either (time base * divs) OR (sample length / sample rate).

Unlike on the DS2000, the DS1000 hardly ever displays all of the data captured.  In fact some time bases are effectively just zooms of others (i.e. they have the same capture length at the same sample rate, and just reduce the amount that it shows on screen.)  So the timebase and screen divisions is not an effective measure of the acquisition time.

I'm pretty sure most DSOs work in (more or less) the same way. The sample time is controlled by the sample rate; when you set a sample length (e.g. 1kPts), the DSO sets the sample rate to (at a minimum) fill the display time. So if you're at 1ms/div (x 10divs = 10ms display time) each sample has to be 10us - so the effective sample rate is set to 100MSa/s. Some DSOs actually change the frequency of the ADC clock - some just keep the same ADC clock rate and decimate unused samples. But at smaller sample lengths (< ~100kPts) and larger time base settings (>= ~1us) the samples often just fill the screen with no overflow.

[alank2]

My first report was with long memory.  Normal memory depth on sinewave:

1 Hz  stable
2 Hz  not stable
3 Hz  not stable
4 Hz  stable
5 Hz  not stable

[marmad]

My first report was with long memory.  Normal memory depth on sinewave:

1 Hz  stable
2 Hz  not stable
3 Hz  not stable
4 Hz  stable
5 Hz  not stable

Alan - I'm not sure if your DSO is correctly doing 'either' Edge triggering - but it might be that I've miscalculated just how slow the DS1000 series is.

Why don't you try 2ms/div? This is more reliable because we have data from this setting. According to Harvs chart below, the wfrm/s rate at that setting is 5.62 - which means the acquisition cycle time is ~178ms - with a blind time of ~154ms. That means if you give the DSO a sine wave of 4Hz (250ms period = 125ms edge to edge), it shouldn't trigger on the opposite edge (it will definitely fall in the blind time) - so if you see both edges, it's doing 'alternating' Edge triggering - if it remains locked on the rising (or falling) Edge, it's doing 'either'.

[alank2]

2ms timebase, normal mem depth, edge, slope both up and down, normal sweep

4 Hz sinewave - rising only (stable)
4 Hz triangle - rising only (stable)
4 Hz square - alternating rising and falling (unstable)

[marmad]

2ms timebase, normal mem depth, edge, slope both up and down, normal sweep

4 Hz sinewave - rising only (stable)
4 Hz triangle - rising only (stable)
4 Hz square - alternating rising and falling (unstable)

And is your trigger level at 0V?

[alank2]

I mean the trigger sweep mode=normal (not auto or single).

4 Hz is stable.

[alank2]

Yes, Trigger level at 0V