Measuring the WaveForms per sec for WaveJet without a trigger out, counter etc.

[jpb]

I hope that the following is of interest. Given that the LeCroy WaveJet is coming to the end of its commercial
life there might be some bargains around (I got my 334 cheap for a 350MHz four channel scope with 4 500MHz probes).

If nothing else it can make all you Agilent and Rigol owners feel smug at the superior performance of your scopes!

I've been wanting to measure the waveforms per second of my WaveJet since I got it but it lacks a trigger out, and I don't have a counter
nor do I have a signal generator (yet - I'm still trying to find what I want).

Then I realised that the history feature of the WaveJet would allow me to measure it using just my 10MHz OCXO with a double check from
my GPS module which produces a 100ms wide 1pps signal.

The history feature give a time stamp to 0.1secs and up to 1024 saved waveforms.

Measurement technique

I setup the WaveJet for the conditions that I'd use for looking at a signal for potential glitches -

Memory 500 points per screen (Giving a history of 1024 screens)
Peak detect (at slower time bases)
Lines off (except when looking at the history - NB lines aren't needed as
you are seeing 1024 screens at once so essentially 500,000 points on view.)

Note that this setup doesn't allow zooming and the sample rate is relatively slow at
longer time bases but with peak detect essentially the sample rate is 1GS/s. At the
faster time bases the sample rate is 2GS/s.

The scenario is if you see a glitch you look at it in the history and then set a
trigger to catch it in detail using more memory.

I did not set screen persistence as having 1024 screens on at once there is
persistence built in.

method 1

I set the 10MHz signal to channel 1 and the 1pps 100ms wide signal to channel 3.
Trigger was set to channel 1.
After pressing stop I went through the history, the level of channel 3 goes high
for 100ms and I could count the number of screens saved with the high signal giving
me a direct measurement of wave forms in 1/10 sec.

method 2

Again trigger of the 10MHz signal and the note the time stamps of the saved wave forms.
Look for the wave forms on either side of where the 0.1 second time stamp changes and
then count (by looking at their numbers) the number of wave forms a whole number of 1/10ths
of a second apart.

I used method 2 for two channels to get a comparison with method 1 and as can be seen from
the table there is good agreement.

I then repeated using just one channel which gives the fastest wave form rate and matches
the conditions in Agilents app note.

Results

The table gives the results. The first two columns are from Agilent's app note. The first is
for the Agilent 2000X series, the second is Agilents results for the WaveJet. The columns
for my results are labelled pps for method 1 and ts (time stamp) for method 2.

Note how pessimistic Agilent's results for the WaveJet are. Particularly striking are those
for the longer time bases where Agilent set the WaveJet up for the maximum memory, e.g.
the difference at 20us/div Agilent have the value of just 35 whilst I measured 2,550!
Also note at faster time bases Agilent's WaveJet figures top out at 1000 whilst I measured up to
4,700 - given that the WaveJet doesn't have a trigger out and has a history memory of around 1000
I suspect Agilent just used this.

The nominal maximum for the WaveJet is 3,600 so I found I got slightly better results than this.

One odd anomaly is those at 5ms/div where I measure only 8.5 instead of it being nearer 20. I
remeasured this a number of times and always got the same results (which equal the nominally
slower 10ms/div results). I don't know why this is the case.

Conclusions
Though the WaveJet does not have a fast update rate compared to the Agilent 2000X series, the latest
Rigol (and certainly not the Agilent 3000X series!) it is not nearly as bad as Agilent
paints it in their app note.

In fact at slower timebases it is faster than Agilent!

	Agilent	Agilent	2 channel	2 channel	1 channel
Timebase	2000X	WJ	(pps)	(ts)	(ts)	Max Possible
1ns/div	NA	NA	2350	2377	4,700	100,000,000
2ns/div	54000	1000	2360	2370	4,650	50,000,000
5ns/div	54000	1000	2320	2340	4,660	20,000,000
10ns/div	54000	1000	2200	2210	4,380	10,000,000
20ns/div	54000	1000	2000	2020	3,990	5,000,000
50ns/div	54000	1000	2080	2090	4,070	2,000,000
100ns/div	52000	1000	1870	1910	3,825	1,000,000
200ns/div	49000	1000	1810	1830	3,610	500,000
500ns/div	43000	1000	1530	1550	3,095	200,000
1us/div	35000	625	1450	1450	3,065	100,000
2us/div	26000	300	1580	1610	3,355	50,000
5us/div	18000	150	1640	1670	3,185	20,000
10us/div	9000	70	1580	1600	2,955	10,000
20us/div	4500	35	1470	1480	2,550	5,000
50us/div	1800	35	1190	1200	1,803	2,000
100us/div	900	35	900	910	942	1,000
200us/div	460	35	470	471	466	500
500us/div	170	25	190	190	191	200
1ms/div	60	20	95.6	96.1	95.7	100
2ms/div	43	15	48	48	48	50
5ms/div	18	10	8.5	8.5	8.5	20
10ms/div	9	7	-	8.5	-	10
20ms/div	5	4	-	4.8	-	5
50ms/div	2	2	-	1.9	-	2
100ms/div	1	1	-	0.96	-	1

PostScript added 23 May 2013

Did a few more experiments turning various features on and off. Found that turning lines off during capture makes a big difference at 500ns/div and very little difference at 1ns/div (which is not surprising as it only plots a few points).

At 500nS div with lines on the rate dropped to only 150 wf/s whilst turning lines off and using points it went back to 2,820 wf/s (a little less than the previous result of 3,095 but I may have had a few other things different).

In most circumstances the lines add little when the memory is set to 500 as the screen is showing 1024 lots of 500 points each slightly offset because the scope and the waveform won't be exactly in sync so the net effect is a very complete curve. The only thing is with glitches, they are slightly less visible if they are very sharp and only contain a few points.

[marmad]

Funny!    I just posted a video of a much simpler technique I came up with for quite accurately measuring waveform update rates on any DSO with or without Trigger Out: you just need a simple variable sine or square wave - a PC sound card will do.

[jpb]

Yes, I was just responding to your post while you were responding to this!

[Hydrawerk]

Given that the LeCroy WaveJet is coming to the end of its commercial
life (...)

Really? The obsolete Tektronix TDS3000 from 1998 is still being produced... Ant it has 10points memory, LOL.
http://www.tek.com/oscilloscope/tds3000
On the other hand, the 5GS/s per channel is still very good and even better than WaveJet with 1GS/s per channel.
EDIT: And this scope is quite expensive!

[jpb]

Given that the LeCroy WaveJet is coming to the end of its commercial
life (...)

Really? The obsolete Tektronix TDS3000 from 1998 is still being produced... Ant it has 10points memory, LOL.
http://www.tek.com/oscilloscope/tds3000
On the other hand, the 5GS/s per channel is still very good and even better than WaveJet with 1GS/s per channel.

The WaveJet is 2GS/s for 2 out of 4 channels but only 1GS/s for all 4 channels.

I don't know that the WaveJet is at the end of its life but a lot of places seem to no longer be selling it. Also when I inquired about the LAN interface for it the LeCroy salesman said that they were offering up to 55% discounts (though this may be for education). So this seemed like it was being superseded perhaps.

On the other hand Iwatsu are still selling a scope which is only a small evolutionary improvement on the WaveJet so perhaps it will keep going for a while. I just can't see them being able to sell it at the present price levels given its age and spec, but on the other hand it is not a "cheap" scope so they really need to upgrade it.

[Hydrawerk]

Well, the DS5500 is the only digital scope sold by Iwatsu. http://www.iti.iwatsu.co.jp/en/pdf/ds-5500_e.pdf

[jpb]

Did an additional small experiment and found turning lines on during capture makes a major difference - see ps on my original post.

[KedasProbe]

Depending on all kind of settings on/off it will be different, statistics should also drop it (CPU intensive)
On the Hameg scope there is also a setting where you can choose the priority 'sample rate' or priority 'waveform rate' (there is also an 'automatic' setting that puts the speed in between if possible)
if you choose waveform rate the sample rate will drop but the waveform rate will go up.

I made a feature request to Hameg to display the time between triggers.
They already show a timer on the screen but only if the time between trigger events is bigger than 1 second.

[jpb]

One drawback of doing these experiments, I've just discovered, is that I've got RSI in my right wrist from twiddling the history knob!
(Hopefully it is just a minor pain and will go away in a day or two.)

My day job is as a Computer Scientist so I'm very aware of RSI as a potential problem (Carpal tunnel syndrome) and have generally managed to avoid it by being lazy and typing as little as possible - so beware knob twiddling on oscilloscopes!

[marmad]

Did an additional small experiment and found turning lines on during capture makes a major difference - see ps on my original post.

Note how pessimistic Agilent's results for the WaveJet are

Perhaps this explains the smaller numbers from Agilent? The X-Series always has interpolation on - so they would certainly have that feature turned on when making comparison measurements with other DSOs.

[marmad]

My day job is as a Computer Scientist so I'm very aware of RSI as a potential problem (Carpal tunnel syndrome) and have generally managed to avoid it by being lazy and typing as little as possible - so beware knob twiddling on oscilloscopes!

Unfortunately you can get it from just mousing/trackballing too much as well. 

[jpb]

Perhaps this explains the smaller numbers from Agilent? The X-Series always has interpolation on - so they would certainly have that feature turned on when making comparison measurements with other DSOs.

It is probably a contributing factor but Agilent's figure was much higher than my lines on figure. This was at 500ns. Agilent's figure (for the WaveJet) was 1,000. My figure with lines off was around 3,000 but with lines on was right down to 150. But Agilent set things up to give maximum sample rate so they probably set the memory to 5000ns*1GS/s or 5k but the nearest memory point is 10k which ironically means the history reduces to 50 from a 1000 so the lines on/off may make less difference.

But Agilents figures are very odd anyway as down at 2ns/div they should have set the memory to 500 and there would have been only a few points so the lines on/off makes little difference yet they were still on a fixed 1000 whilst I measured nearer 4600 for the same setup. I wonder if they did what I did, looked at the history, but perhaps looked only at whole seconds (perhaps they had the real time clock set rather than trigger marker times) and so the maximum rate they could get was the maximum history of 1000 in one second.

Another factor is Agilent were measuring the 100MHz model WaveJet and mine is 350MHz so perhaps there is a difference in the hardware (there may be a dividing line between the 100MHz model and the 200/350/500MHz models - the latter have 2GS/s while the former only go up to 1GS/s) .

[Hydrawerk]

Agilent guys use Default setup on all the measured scopes...  They apparently don't use the best waveform update rate setting.