Oscilloscope transient response: analog vs digital, fastrise, tunnel diode gen

[jonpaul]

Hello all,  during  scope calibrating and restoration, transient response is tested with tunnel diode adapter or a fast rise generator like TEK PG506. See for example the  TEK 2465B  threads.   I have tested TEK 7104/7A29,  2465B, 2467B and digital Yokogawa DL7440.  Interesting variations analog vs digital scopes and between different models, but even between CH 1 vs CH2.  Observing Preshoot, overshoot and abberations.

Anyone else interested in optimization of transient response, and types of pulse generators ?  Comments on analog and digital scopes?  your Feedback and images appreciated!

Cheers,

Jon Paul

[joeqsmith]

Hello all,  during  scope calibrating and restoration, transient response is tested with tunnel diode adapter or a fast rise generator like TEK PG506. See for example the  TEK 2465B  threads.   I have tested TEK 7104/7A29,  2465B, 2467B and digital Yokogawa DL7440.  Interesting variations analog vs digital scopes and between different models, but even between CH 1 vs CH2.  Observing Preshoot, overshoot and abberations.

Anyone else interested in optimization of transient response, and types of pulse generators ?  Comments on analog and digital scopes?  your Feedback and images appreciated!

Cheers,

Jon Paul

Hello Jon.  You could join the show us your square wave thread.  There you will find all sorts of combinations of test equipment looking at edge rates. 

https://www.eevblog.com/forum/testgear/show-us-your-square-wave/

[dzseki]

I have great interest in this topic.
Whenever I poke with high frequency linear circuits measuring rise and fall times is what I always measure -In abscence of a VNA

As for source I managed to get a hungarian made ECL pulse generator type EMG 12563, it ha a maximal output voltage of + or -2V (terminated on 50 Ohm), with +/-1V offset range, the maximum repetition range is 300MHz and the min. pulse width 4ns. The manual is somewhat vague on the rise time stating <1ns, but for 35USD I figured I couldn't be wrong with this.
At work the fastest oscilloscope we have is a LeCroy WavePro 725 (2.5GHz, 40Gs/s) with that the riste was measured to be 425ps 

On the other end I use a HP 1720A oscilloscope to catch the waveform, this scope is 275MHz bandwidth and therefore supposed to have 1.3ns rise time, although I often struggling to catch this because of the scope is rather old and sometimes actung up, resulting rise time mostly in the 1.5-2ns range. anyway here is a shot, probably the best I even seen from it.
That was made with the HP 1720A + 1120A 500MHz FET probe taping on the output of the above mentioned pulse generator (50Ohm terinated) the picture is 1ns/div

[jonpaul]

Hello:

RE TD pulsers:

TDs can be destroyed by static or 60 Hz stray (eg from soldering iron) or by excess Tjuct, as they are germanium.
I have soldered and curve traced with no problem on the TEK 10 mA parts.  A 20 - 100V peak pulse input needed for either avalanche transistor or TD. TD needs a current source,  a series R several K Ohm, as in TEK pulsers.

RE Leo Bodnar Pulser: must these be ordered from UK or is there a USA source?
http://www.leobodnar.com/shop/index.php?main_page=product_info&products_id=303

RE Aberrations:  response improves if you add a wideband resistive attenuator 6-20 dB, to reduce VSWR from the pulser to the scope input.
The CAL TEK 2465B,  transient procedure requires 2X, 2.5X and 5X attenuators.  I built 3 pcs each of  50 Ohm pi attenuators in Pomona box.
The attenuators reduce pulse aberrations to only 1-2% vs the direct connection pulser to scope.

I will soon post photos of my tests.

Ciao,


Jon

[jonpaul]

Hello again, fired up my old TEK 7104, 1 GHz, 7A29 VER and 7B15 HOR plugins, 4 shots 

L and R VER, direct from TU-5, and with 14 dB pi attenuator.

Interested in your comments and feedback.

Ciao,

Jon

[jonpaul]

Same setup TEK TU-5 driven from PG504, 20V pulse, into Yokogawa DL7440 500 MHz, using 14 dB pi attenuator

Jon

[dzseki]

The CAL TEK 2465B,  transient procedure requires 2X, 2.5X and 5X attenuators.  I built 3 pcs each of  50 Ohm pi attenuators in Pomona box.
The attenuators reduce pulse aberrations to only 1-2% vs the direct connection pulser to scope.

Interesting, my HP 1720A CAL procedure simply says when adjusting pulse shape, adjust the overshoot to the same level as it is specified on the PG's output.

[jonpaul]

Hello again

Any idea if the TD in 7B92A are fast enough for making a TD pulser to allign 7104/7A29, 2465B ?

Have 3 bad ones, one may be a parts unit.

Thanks

Jon

[rhb]

Twenty years ago I tried to build a fast rise time square waver generator using a 10 MHz clock and a sub nS MECL buffer.  I failed miserably.  But it was a wonderful education.  Circuit layout is absolutely critical.

I have Leo's  BNC square wave pulser and have a 100 pS pulse version in transit.  I'm planning to do a bunch of scope AFE testing with the 100 pS pulse version to post to the T&M section.

The only "improvement" on Leo's unit is a big piece of heatshrink to provide mechanical and ESD protection.

Once you have one it will be your first input to a new scope every time.

Edit:  The spike generator just arrived.  < 50 pS rise time, 100 pS duration on Leo's CSA803.  I got it so I could easily evaluate the BW of an AFE on a DSO using the FFT function. With an Owon XDS2102A sampling a 1 GSa/S in 8 bit mode, the FFT clearly shows that it will alias at 500 MSa/S in 12 bit mode.  It's less than -15dB  between 250 & 300 MHz and only -10dB at a Nyquist of 250 MHz.

[jonpaul]

What was cost of the Leo pulser with shipping to USA? How long to arrive?

Definitely a good design.

Has anyone tried to salvage the TDs in old/junk 7B plugins like 7B92A? Seems they must have fast TDs.

Jon

[rhb]

It took 9 days to central Arkansas  and cost $76.78.

I've attached the plots that came with my spike unit.  The square wave unit plot is MIA at the moment.

The asymmetry is due to the AFE of the sampling head being minimum phase.  The ripple is probably due to the BNC adaptor used to connect to the sampling head.

They are also available with microwave connectors, but at higher cost because of the connector costs.
The repetition rate is 10 MHz, so an FFT shows the BW as a set of spikes at 10 MHz spacing.

[jonpaul]

RHB, many thanks but this short pulse is NOT what a scope transient CAL needs!

It should be a wider pulse with a flat top to see the aberrations in square wave response, like the old TEK pulsers produced.

Can the Leo pulser do that? Does it have both short pulse and longer output?

Jon

[rhb]

Leo's standard pulser produces a <40 pS rise time 10 MHz square wave and has a trigger out for use with sampling scopes.  The 100 pS spike version is a modification available by request.  I have both versions.

Leo supplies a printout from a CSA803 showing the measured rise time and performance for each unit.

 A lot of the people on the TekScopes list have them. 

There's a long thread with lots of examples from many scopes.

https://www.eevblog.com/forum/projects/yet-another-fast-edge-pulse-generator/

[macboy]

Jonpaul, you are quite right that an impulse/spike is not suitable for calibrating scope front end response, nor for rise time measurements. However, the standard sqaure wave pulser from Leo is actually perfect for transient response testing of most scopes a few GHz and below. Beyond that you start to need to consider the rise time and shape of this generator's pulse when interpreting the result. The small-ish 1 Vpp (terminated) voltage might be a limitation. This can be reduced with 50 Ohm attenuators, but then the response of the attenuator needs to be considered.

In the thread referenced by rhb I demonstrated a way to directly determine frequency response of a scope using Leo's pulser. This requires that the scope has math-on-math capability. You display a single rising edge, then average many waveforms (remove noise, increase resolution), take the derivative d/dt (to get an impulse from the step), then take the FFT of that impulse. The displayed FFT is the actual frequency response of the system (scope+generator). Since the generator has quite a well behaved step and is much faster than most scopes, the response can be interpreted as that of only the scope in most cases.

Leo's spike/impulse generator version should be directly usable for this (just take the FFT of a single impulse or average of many) without all the extra math. I'd imagine the spectra of the impulse version to be slightly less flat than that of the step version (when used as above), but I don't know if that has been tested.

With either method (step generator with d/dt, or impulse generator), you must limit the FFT to a single impulse rather than a train of them, or the FFT will be 'spiky' instead of smooth and continuous.

Aaron

[rhb]

I just found two  pulsers I built 25 years ago.  Clock oscillator and an MC74F00.  One has a 74LS74 to divide the clock and make it symmetric.  They are actually much better than I thought despite being badly laid out on protoyping PCBs.  Rise times (2.86 & 3.35 nS) are in line with the data sheet for the MC74F00.  The 74LS74 needs a decoupling cap on Vcc, but other than that I did as well as could be expected using what I had available.  That's a pleasant surprise.  I was repairing a Tek 465 and Dumont 1060 at the time and needed a fast step.  Shortly after I picked up a couple of Tek 106s which are much more appropriate for the task.

Apparently the EE at work who helped me test them and I didn't set the 485 coupling to 50 ohm DC so we thought they were ringing really badly.

I'd forgotten how much stuff I built in '85-'93.  I have all these mystery boards I have no idea what they do or even how to connect them.  Fortunately, none is complex.  Plus lots of things in unlabeled boxes.