The bugs aren't awful. It does some cute things, like protocol analysis and NRZ eye diagrams without a trigger.
The bugs aren't awful. It does some cute things, like protocol analysis and NRZ eye diagrams without a trigger.
-- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
It wouldn't take a lot of equalization to bang flat the step response. It wouldn't need enough gain anywhere to increase the noise much.
I did a software adaptive equalizer, in case I ever decide to sell a TDR system.
Given an actual waveform and a perfect target waveform, it evolves a FIR filter to beautify things. If you try to increase the gross bandwidth too much (like, say, 2:1) you do get a noise explosion. But fixing little ringies and reflections and such works fine. So I could make a fast but ugly TDR and fix it in software.
This is "the deconvolution problem", namely finding filter F such that
S ** F = T
where ** is convolution, S is some signal, and T is the desired target response. This is sometimes called an "ill-posed problem."
-- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
Deconvolution isn't il-posed unless the SNR goes towards zero, e.g. trying to deconvolve out a true Gaussian, or (my favourite example) using a tungsten filament at different temperatures and deconvolving out the reflection spectrum of some object. Now _that's_ ill_posed. But for other things, e.g. my NA6 microscope, deconvolution works great, and the filter is easy to calculate.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 hobbs at electrooptical dot net http://electrooptical.net
The conventional approach - the one that seems to create the most academic papers - is to FFT the signal S and FFT the target T and do a point-by-point complex divide to get the Fourier transform of F, then reverse FFT to make the FIR filter. That tends to create divides by very small numbers, or even divides by zero, so has noise problems or worse. So folks try to mess with the denominator spectral points to avoid the singularities but not trash the waveform too much. At that point, it's ill-posed.
The thing I did is all in the time domain and is really simple, and it doesn't do any divides, so it behaves pretty well.
-- John Larkin Highland Technology, Inc lunatic fringe electronics
Divides don't introduce anything much by way of roundoff--it's subtractions of nearly-equal quantities that do that. Apart from roundoff, time- and fr equency-domain deconvolutions are identical, so your scheme that worked in the time domain would also have worked in the frequency domain, provided it was really the same.
I agree that sometimes people get into trouble with deconvolution, usually by waving FFTs around like Gomez Adams's witch doctor--ignoring the noise g ain, not bothering with windowing, that sort of stuff. But that's mostly c arelessness and a bit of stupidity rather than any fault of the basic idea.
Unless they're attempting recreational impossibilities such as removing a b road, smooth instrument function such as the Planck curve, it works pretty well. You do have to keep an eye on the noise gain, for sure. In the cases where I use it, it's no more ill-posed than an audio tone control.
Cheers
Phil Hobbs
Divides don't introduce anything much by way of roundoff--it's subtractions of nearly-equal quantities that do that. Apart from roundoff, time- and fr equency-domain deconvolutions are identical, so your scheme that worked in the time domain would also have worked in the frequency domain, provided it was really the same.
I agree that sometimes people get into trouble with deconvolution, usually by waving FFTs around like Gomez Adams's witch doctor--ignoring the noise g ain, not bothering with windowing, that sort of stuff. But that's mostly c arelessness and a bit of stupidity rather than any fault of the basic idea.
Unless they're attempting recreational impossibilities such as removing a b road, smooth instrument function such as the Planck curve, it works pretty well. You do have to keep an eye on the noise gain, for sure. In the cases where I use it, it's no more ill-posed than an audio tone control.
Cheers
Phil Hobbs
Well, a graphic equalizer where the sliders go to +400 dB.
-- John Larkin Highland Technology, Inc lunatic fringe electronics
Am 19.07.2016 um 03:57 schrieb John Larkin:
..
..
The HP 54750 scope has convolution in its TDR software to improve on the source rise time. There is an interesting app note from Pico Second Pule Labs, Boulder that compares the available scopes and that discusses this.
Still looking for a pulse generator that is worthy of my dual
50 GHz / 9ps rise plug in.regards, Gerhard
My volume control only goes up to 11. ;)
Cheers
Phil Hobbs
A couple of people sell NLTL (shock-line) edge enhancer gadgets. I have a sample around here somewhere. They are often sold as frequency-domain comb generators, but make really fast steps or impulses.
The really fast sampling heads use a shock line to make the sampling gate.
-- John Larkin Highland Technology, Inc lunatic fringe electronics
I've been wanting an excuse to build one of those, but none has turned up so far. Such a nice antidote to the usual tendency of transmission lines to round off sharp edges.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 hobbs at electrooptical dot net http://electrooptical.net
Ferrite beads and type 2 capacitors? :)
Tim
-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website: http://seventransistorlabs.com
Something like that. 6 ps edges don't happen by accident either.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 hobbs at electrooptical dot net http://electrooptical.net
The really fast ones are ICs, with numbers like 100 or so varactor diodes along a line. I have seen shock lines built with discrete inductors and diodes; McEwan, the impulse radar guy, did some. My friend Mark Kahrs has done some work on discrete varactor NLTLs, if anybody is interested.
High voltage NLTLs have been done using the nonlinearity of ceramic caps.
I have fiddled a bit with varactor NLTLs in simulation. What I learned is that this is NOT a problem that is well suited to fiddled simulation.
There is also a NLTL based on successive saturating inductors, and it can make monstrous pretty-fast pulses. Saturating magnetics is an obscure and interesting niche. We recently met the guy who designed the megawatt laser pulsers used in the Cymer DUV lithography lasers, the ones that we do the timing for. He said that the only known failure mode was the cooling oil boiling off.
-- John Larkin Highland Technology, Inc lunatic fringe electronics
Am 19.07.2016 um 11:53 schrieb JW:
Buying does not count. :-)
30 ps rise time is still somewhat easy. CML logic and 10 GB laser drivers can do that. We did it in our XFP FO modules on FR4, with the TOSA / ROSA attached with flex Kapton, after 3D EM simulation nightmares. <Maybe one could use a CML driver, clip its output with beam lead schottkies and be happy with 100 mV swing. That has a taste of fraud, but should be good enough for TDR.
regards, Gerhard
All the Picosecond Pulse Labs app notes are off their website. They were bought by Tektronix. All their cool stuff has been replaced by fluff app notes from Tektronix Marketing.
Would you care to share the file comparing scopes?
Thanks,
Steve
Nevermind, I found a copy on the web.
Steve
JFTR Jim Andrews has them on his website
Lots of good stuff there.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 hobbs at electrooptical dot net http://electrooptical.net
They also sound totally cool. "Magnetic compressor". "Electromagnetic shock line". Star Trek wishes they had need of these components.
Physicists get to play with all the cool stuff.
Tim
-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website: http://seventransistorlabs.com
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.