I am hoping someone with with high speed experience could point me in the r ight direction.
My issue is that I have an ~850pS pulse (somewhat random pulse output) tha t I need to lengthen to 25nS. I have built many oneshot/timer circuits over the years but never this fast. My research seems to indicate that most tim ers (available ICs) need a minimum trigger pulse of several to ~ hundred nS . Retrigger time doesn't seem to be an issue (yet).
Thanks in advance. Martin
Martin Gira Sr. Research Engineer Center for Visual Science (CVS) Center for Medical Technology & Innovation (CMTI) Neurobiology and Anatomy University of Rochester
That might be too fast for even the fastest one-shot, like an LVC1G123. But it would be easy to make a pulse stretcher. Assuming your source has reasonable drive strength, a small schottky diode in series, and an RC to ground would work.
You can use the Cin of the one-shot for the C. So, one diode and one resistor. Optimize the time constant per your expected trigger pattern.
I'm doing that in some GHz-range signal detectors now and it works fine with no discharge R at all. A little schottky like a BAT15 will have maybe 0.25 pF capacitance and 100nA or so reverse leakage.
Where is that fast pulse coming from? What's the logic swing?
--
John Larkin Highland Technology, Inc
Science teaches us to doubt.
Claude Bernard
I've built something similar here together with a colleague recently, using an MC100EP51 (a fast PECL FF). The input, from a fast photodiode straight into an ADCMP582, clocks H to the output, which then triggers the asynchronous reset some 10 ns later after a simple RC delay. This is perhaps a bit crude, but worked well enough for the application, as we didn't care about anything past the initial rising edge (provided the pulse just stayed high long enough for a later, slower trigger to catch it as well). Random jitter on the rising edge for the complete system was a few ps, almost certainly dominated by the LVTTL output buffer.
Best, David
On 01.09.20 4:45 pm, M Gira wrote: > Hi All, > > I am hoping someone with with high speed experience could point me in the right direction. > > My issue is that I have an ~850pS pulse (somewhat random pulse output) that I need to lengthen to 25nS. I have built many oneshot/timer circuits over the years but never this fast. My research seems to indicate that most timers (available ICs) need a minimum trigger pulse of several to ~ hundred nS. Retrigger time doesn't seem to be an issue (yet). > > Thanks in advance. > Martin > > > Martin Gira > Sr. Research Engineer > Center for Visual Science (CVS) > Center for Medical Technology & Innovation (CMTI) > Neurobiology and Anatomy > University of Rochester >
An ECL-output comparator is an open emitter, so add a cap and a weak pulldown to make a stretcher. Since the '582 is a dual, use the other half as a discriminator on the stretched pulse.
The 582 is expensive. One of the slower parts, like a 561, would work at these speeds.
--
John Larkin Highland Technology, Inc
Science teaches us to doubt.
Claude Bernard
Well, the source pulses in our case were ~5 ps, and only some 100 ps (at the trigger level) out of the photodiode. For a one-off lab gadget like this, it seemed less complex to first square it up as faithfully as possible while keeping the rising edge intact, and then manipulate the cleanly driven logic signals. Parts cost is of less relevance than the inelegance of my solution here. ;)
Using a dual like the ADCMP561 (the ADCMP582 is a single) for a stretcher is indeed a good suggestion, though. Speaking of which, what actually happens in practice if you drive fast comparators like those with a pulse shorter than the specified minimum pulse width? I've never tried?
two 100ep52 d-flops will capture this pulse with no problem. The first one captures the pulse and signals the second to start the delay. When the delay times out, it releases the first one so it is ready to find another pulse.
Most any flip-flop, CMOS or ECL, can be made to clear itself after it's clocked high. You can do that with a delay line from Q to Clear (PCB trace, or one R-L-C) but that has to clear out before the next pulse can be accepted. That can be worked around.
Just an RC from Q to Reset works with some logic families, but some don't like that, especially with long tau's.
On Wednesday, September 2, 2020 at 1:45:41 AM UTC+10, snipped-for-privacy@gmail.com wrot e:
right direction.
at I need to lengthen to 25nS. I have built many oneshot/timer circuits ove r the years but never this fast. My research seems to indicate that most ti mers (available ICs) need a minimum trigger pulse of several to ~ hundred n S. Retrigger time doesn't seem to be an issue (yet).
Ghiggino, K.P., Phillips, D., and Sloman, A.W. "Nanosecond pulse stretcher" ,Journal of Physics E: Scientific Instruments, 12, 686-687 (1979).
We didn't stretch the pulse to 25nsec - we only needed to get it long enoug h to trigger some ECL logic.
As is mentioned in the paper, the two transistors configured as an emitter- coupled monostable would probably have done a better job.
You probably can't now buy the broad-band (5GHz) PNP transistor we used but equivalent NPN transistors are widely available.
The BFR92 NPN transistor should work fine.
formatting link
Element 14 has 23,000 in stock , and they aren't expensive. It's surface m ount part, but through the board mounting doesn't work well at this kind of speed.
You need to put a surface mount "base-stopping resistor" - something like 2
7R worked for me - right up against the base pin to prevent fast parasitic oscillations. There are better (but trickier) ways of doing this, but a bas e-stopper won't slow the circuit enough to stop it doing your job.
E-mail me with more detail, and I can probably whip up something in LTSpice that you could play with in simulation. I've posted at least one .asc here recently that included an emitter-coupled monostable and the Spice model f or a BFR92 (but I was exploiting the low collector capacitance of the BFR92 rather than its speed).
The difference between a fast op amp and a fast comparator is that the gain of a fast op amp drops as a linear function of frequency, and the gain of a fast comparator drops as the square or cube of frequency (two stage of g ain, or three).
The small low frequency content in the pulse hitting the comparator does ge t amplified - so you might see a ripple on the output, but that's it.
Broad band transistors don't have a lot of gain, but what they have rolls d own to 5GHz or higher.
formatting link
The BFU768F offers 110GHz for less than a dollar. Stopping it from self-osc illation might be tricky.
On a sunny day (Tue, 1 Sep 2020 08:45:33 -0700 (PDT)) it happened M Gira wrote in :
Apart from the solutions like diode + RC time and D-flipflops there is a simple circuit I used many times:
formatting link
C1 R1 is a differentiator, C2 R2 sets the output pulse width, you need to slice that somewhere. Of course the Ccb and Cce of the transistor counts too. Using 10 GHz transistors should work (have not tried that high).
If you bias the base just below Vbe it can work as pulse detector too. May save components.
R2 will be low for short output pulses, could even be 50 Ohms perhaps. Takes 5 minutes to test.
Yeah, they've been glacial about that stuff. They don't even add their existing models to the library, e.g. ADA4899.
The selection dialogues for the built-in library are one of LTspice's worst features. You have to run down the whole list using the arrow keys to find such illuminating descriptions as
"Voltage comparator" or "uPower dual comparator"
Some of the newer parts actually have a good enough description that you can tell whether it's worth looking at the datasheet, e.g. LT6703-2 "uPower, Low Voltage, SOT-23, Dual Comparator with 400mV Reference"
Sure would be nice if there were a detailed list view available.
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510
http://electrooptical.net
http://hobbs-eo.com
I'm not sure that I would trust the models when used outside the datasheet operating conditions. I'm sure they have internal transistor level ADICE models that would accurately predict behaviour in these conditions, but then IIRC they tend to give some idealised model to customers instead of the accurate transistor level model. I can't understand the motivation for this. The idea that this would stop a competitor from reverse-engineering it is silly - 1. any serious competitor would easily be able to open one up and extract the schematic and 2. they probably already bought the competitor anyway.
I've seen some bizarre behavior. Sometimes an internal current source makes an opamp into a perpetual atomic battery. One opamp and one capacitor becomes a kilovolt (or teravolt) power supply.
--
John Larkin Highland Technology, Inc
Science teaches us to doubt.
Claude Bernard
hat I need to lengthen to 25nS. I have built many oneshot/timer circuits ov er the years but never this fast. My research seems to indicate that most t imers (available ICs) need a minimum trigger pulse of several to ~ hundred nS. Retrigger time doesn't seem to be an issue (yet).
Is the pulse at the volt level or so? For 0.8 ns to 25 ns I think you might be able to get by with a fast comparator (from LT/ AD, If not at the volt level.) And then a Schmitt trigger inverter with a diode and RC on the input. something like the pulse stretchers here.
formatting link
.html But much smaller C's! (maybe the input C of the inverter is enough as JL suggested.)
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.