Yet another fast edge pulse generator

[Leo Bodnar]

Executive summary: (updated 1st September 2017)
40-50 picosecond risetime (and even faster falltime) generator with 1V peak to peak 10MHz squarewave output.  Has trigger output.
Powered and configured from USB.  Ships ready to use.
BNC jack for direct connection to your test equipment.
Designed and hand-built in the UK.
Available in stock for £49.99

Update January 2018:
2.92mm version (~30ps): http://www.leobodnar.com/shop/index.php?main_page=product_info&cPath=124&products_id=302
SMA version (~32ps): http://www.leobodnar.com/shop/index.php?main_page=product_info&cPath=124&products_id=303



--------------------------------------------------------------------------------------------------------------------------------------------
Here is a simple pulse generator I have made for testing medium BW scopes' risetimes.  This explains BNC connector on the PCB.
It produces flat top signal with clean and fast rise and fall edges.

I had a few different designs built, using CML and LVPECL drivers, fast comparators and laser diode drivers. Some are good, some bad,  some with 400mVpp, some 2Vpp, some well 50R terminated, some not.  Not a grave issue for launching pulse directly into oscilloscope 50R input but it might be a problem for some applications I have.  Still working on them - my target is ~40ps trise/tfall in the most compact volume.

With 4 or 5 Jim Williams avalanche generators I made over the years I never felt satisfied with the result.  Jim's pulser design is infinitely tweakable (which is a blessing and a curse) and needs some work before it becomes a risetime testing device from slew rate limit testing device. I wish Jim wrote an app note on the distinction between slew rate and rise time and their practical limits. A lot of people I have talked to don't understand the difference.   I had an opamp design guy being unable to explain why ideal RC network has characteristic risetime but no slew rate limit. Avalanche design does not have clean fall edge either. Anyway, I did not want another avalanche pulser.

This one is the simplest, based on 572 comparator from AD. 

It shows 60ps rise and fall time on CSA803A with SD-26 head.  Unlike most scopes, CSA803A cannot trigger from observed signal itself (only from external trigger input) so I had to use an ugly contraption of BNC-N-type-SMA with a T-piece to pick up part of the signal and feed it into the trigger. This does not help signal integrity but I'm reasonably happy with 60ps in such a setup.

I have attached a few oscilloscope screenshots I had around.
Tek CSA 803A with SD-26 (20GHz) 62ps risetime / 59ps falltime
Agilent MSO-X 3024A (200MHz) 1.6ns
LeCroy DDA-125 (1.5GHz) 295ps
LeCroy WavePro 7300A (3GHz) 143ps

All risetimes are for 10%-90% levels into internally terminated 50R scope inputs.

UPDATE =============== UPDATE =================== UPDATE

Gentlemen,

I have finished the configuration software and I call the project ready to ship. 
You can adjust output level from 50mV to 1.2V peak-to-peak and flip the trigger/output phase in case you are using sampling scope.
Settings are non-volatile, stored in the device and survive power off. 

There is a user calibration multiplier if you want to fudge voltage levels to read reasonably accurate value.
Output matches indicated value to about 5-10%.  But above 1000mVpp actual and indicated output start diverging due to output current source saturation.

Rise and fall times stay reasonably low throughout the whole output range (50-1200mVpp) but to keep specified 50ps edge time keep the output level below 700mVpp.

Order yours here: USB Pulser

Many thanks for your custom!
Leo

[fcb]

Very smart Leo.

I like the USB connector for the 5v (I built a USB powered JWG earlier this year), and also the tag connect for the PIC!

Any plans to sell these?

[Leo Bodnar]

I like the USB connector for the 5v (I built a USB powered JWG earlier this year), and also the tag connect for the PIC!
Any plans to sell these?

I'll be happy to build a few for sale if this is something useful to others.

Leo

[tggzzz]

An alternative approach is:

• one "section" consisting of a 74lvc1g14 + 143ohm series resistor to give an output of 0->1.5V to 0->5V depending on Vcc, with an output impedance of ~150ohms
• three of those sections in parallel, to give an output impedance of ~50ohms
• driven by whatever signal is available, including a simple RC relaxation oscillator at a suitable frequency

That certainly gives a clean signal with <1ns transition time into 50ohms. I can't directly measure any faster with my 350MHz scopes. However, indirect frequency-domain measurements indicate a transition times of ~650ps, and a quick test by someone else (i.e. not verified by me) indicated ~300ps rise/fall times.

[smoothVTer]

Very useful.  I am also interested in acquiring one of these if you'd be willing to sell one.  What would you charge for a completed assembly?

[tggzzz]

Very useful.  I am also interested in acquiring one of these if you'd be willing to sell one.  What would you charge for a completed assembly?

I'm not really interested in starting a cottage industry, thanks.

I can't guarantee those results since I can't measure them properly; I only have a Tek 485. So the price would be £1000, to enable me to buy a scope that could measure the output properly

They aren't that difficult to make on a double-sided pcb. The major issue is to get good power supply decoupling, which implies broad and short tracks to the Vcc and ground pins.

If you are prepared to do a bit of experimentation, then it might even be possible to make a tolerable one on a breakout board. I'd start by trying a single 74lvc1g* and a 43ohm series resistor, with an 0603 decoupling capacitor mounted diagonally on top of the 74lcv1g*; I used wirewrap wire to connect the capacitor to the pins. Edge launch SMA connectors would be the simplest connector option.

[H.O]

I'll be happy to build a few for sale if this is something useful to others.

Leo

I'd gladly purchase one if you decide to build a batch.

[WastelandTek]

That looks like just the ticket, especially as it seems Vincent is too busy at Tesla to make his Jim Williams implementation available.

What is the plateau duration on this?  I understand that is a problem with the Williams pulsers.

[Leo Bodnar]

Thank you for the comments, guys.  I will make a batch and let you know - I have some PCBs and components left from prototyping.

This is essentially 50% duty ratio rectangular pulse so plateau duration depends on the frequency.  This particular design uses internal PIC oscillator which is slaved to the USB clock to keep it simple and compact.  Output frequency is 12MHz. This works fine for ordinary scopes that trigger off the same signal front they are capturing which is almost all the scopes out there apart from few sampling ones like CSA 803 / Tek 1180x / 8300. In that respect it's no different in operation than vanilla JW pulser.

Leo

edit: s/square/rectangular/

[rx8pilot]

Love it.

I would love to have one if it somehow becomes available. If you prefer not to make a (small) business out of it - I bet a few forum members would be willing to do the assembly and sell them as a courtesy service. I could run a few panels through my pick and place line and have 100 in a few hours.......not to make any money but because these things are interesting.

[fcb]

Thank you for the comments, guys.  I will make a batch and let you know - I have some PCBs and components left from prototyping.

This is essentially 50% duty ratio rectangular pulse so plateau duration depends on the frequency.  This particular design uses internal PIC oscillator which is slaved to the USB clock to keep it simple and compact.  Output frequency is 12MHz. This works fine for ordinary scopes that trigger off the same signal front they are capturing which is almost all the scopes out there apart from few sampling ones like CSA 803 / Tek 1180x / 8300. In that respect it's no different in operation than vanilla JW pulser.

Leo

edit: s/square/rectangular/

My JWG just used USB for power - be good if it didn't need a 'real' USB port to work, so it could run from everything including USB power-banks.

[Leo Bodnar]

My JWG just used USB for power - be good if it didn't need a 'real' USB port to work, so it could run from everything including USB power-banks.

Yep, this is how it works.  USB connectivity is there "for free" and not of much importance or necessary for operation.

rx8pilot, thank you for your offer, I have few dozen PCBs remaining and it's not a big deal to assemble them manually.  I'd be amazed if the world market is that huge.  But stranger things happened

Leo

[rx8pilot]

I'd be amazed if the world market is that huge.  But stranger things happened

Leo

Me too - I would guess a low number which is probably why these things are not commercially made even by niche manufacturers. It is a fun and educational tool to learn about though. I have been diving into high-speed digital signal integrity over the past year and it is like drinking from a fire hose. This little circuit could be fun for TDR experiments, PCB structures, and other mysterious challenges related to high-speed design. Pretty much anything even close to those rise times is huge and expensive.

[MK]

do you have a circuit diagram perhaps?

[edavid]

rx8pilot, thank you for your offer, I have few dozen PCBs remaining and it's not a big deal to assemble them manually.

Did you hand solder the ADCMP572?

Also, have you thought about adding a pretrigger output?

[Leo Bodnar]

MK, formalising the schematic is not a priority at the moment.

Did you hand solder the ADCMP572?
Also, have you thought about adding a pretrigger output?

No, this has been reflowed.
Pre-trigger would need another connector.  I'd like to keep this simple for now until it's working well.
Leo

[lukier]

I'd be amazed if the world market is that huge.  But stranger things happened

I would like to buy one unit (if possible). I was thinking about this pricey comparator before, but here you have well evaluated design, proven by equipment I don't have (sampling or > 1GHz oscilloscope) - that's a big plus.

[edavid]

Pre-trigger would need another connector.

It could just be a couple of header pins.

[Neomys Sapiens]

U would like to register too for such a device. Albeit I concur with edavid that a pretrigger output would come handy.

[richnormand]

This looks great to check my S-6 sampling head on my trusted Tek scope.
Had to change a switching amp in it.
It does seem to function fine now but this would be the acid test as the risetime is supposed to be 30ps.
In theory the switching amp should not have affected the sampling diodes but who knows.
I would be interested in getting one too. External trig output is also a neat idea.
Cheers and thanks.

[Leo Bodnar]

I already have some PCBs here so I'd like to use them but I'll see what I can do with pre-trigger output for the next batch.

What time delay (if any) between the pulse and pre-trigger are you thinking of?

Leo

[edavid]

I already have some PCBs here so I'd like to use them but I'll see what I can do with pre-trigger output for the next batch.

What time delay (if any) between the pulse and pre-trigger are you thinking of?

The Tek 7T11 requires 75ns, so you could just make it one clock period.

I don't know if there are other samplers that require a greater delay.

[Leo Bodnar]

OK, for the next spin I am going to remove the MCU, use 10MHz TCXO as a frequency reference and break out LVCMOS buffered pre-trigger.

With 10MHz stable repetitive signal sampling scopes can trigger on the nearest edge which is 50ns away or next one at 100ns.

Buffered TCXO cycle-to-cycle jitter is probably going to be in the order of of few ps but this is not going to be a deal breaker - hopefully.

In a pinch this can also be used as a 10MHz reference oscillator or source of harmonics-rich signal.

Leo

[richnormand]

As an example my S-53 trigger recognizer can take a trigger input up to 100 MHz so your present rep rate of 12 MHz should not be an issue at all.

Since this is a sampling scope the time delay is not that important (the main plugin is used as a TDR after all, so I can add a delay at will) but trig jitters are.

Often I have to amplitude divide the signal and go through an amp and divide by 100 Hittitte chip to generate the trigger. Not ideal as this can corrupt the signal.

In it's present form just a 12 MHz pulse in sync with either the rise or fall of the waveform would be OK since for a sampling scope the sync jitter will be the real issue.

It looks like I could use your present design directly. But just a test point on the board with a trigger output ( buffered from the test signal) would be very nice indeed.

Have fun!

[Leo Bodnar]

Update: I have added an SMA with trigger output and re-spun the boards.
I should have them tuned and tested by early next week.
Thanks
Leo