Electronics > Projects, Designs, and Technical Stuff
Usefulness of different TDR designs?
<< < (5/10) > >>
rhb:
I'm afraid I don't understand what you mean by  "sampling gate".  Would you mind providing a more complete description?  Both of my current units are covered with heat shrink tubing for mechanical and ESD protection.  IIRC it's a clock chip, MCU w/ USB interface and the laser driver plus the obligatory passives.

My understanding of the circuit is the 10 MHz clock is driving the the modulation input of the laser driver which dumps the output straight into the connector via a blocking capacitor. The Maxim part provides the slew rate for the sharp edge.  The MCU varies a digital pot to set the output voltage.  But that's only a rough guess.  I've never spent any time analyzing it.  It is so much faster than the 1.5 ns unit I built 30 years ago it's not even funny.

I've got a pair with 1 MHz clocks in place of the 10 MHz on order which makes them more useful for general TDR work.
David Hess:

--- Quote from: rhb on April 21, 2019, 08:54:00 pm ---I'm afraid I don't understand what you mean by  "sampling gate".  Would you mind providing a more complete description?
--- End quote ---

I mean the input side of the TDR which is typically a sampling bridge or other sampling structure to provide the highest possible bandwidth.  The clever part is that since sequential sampling can be used, there is no need to adjust sampling gain for random sampling which considerably simplifies things. (1)

Sacrificing amplitude for a faster edge does not increase performance unless the input bandwidth is sufficient to take advantage of it.  Of course an integrated laser driver or comparator with a CML output might still be preferable because of simplicity.  Just do not use it for the extra edge rate which will be irrelevant unless you can design and fabricate a sampler to take advantage of it.

(1) Offhand I cannot think of any advantage to making a random sampling TDR and there are several reasons not to.  A Tektronix 7T11 can be used as one and I have done it just to see how well it works but only for comparison purposes.
rhb:
OK, you're just describing my 11801.  BTW I just bought an SD-32 for $500 :-)  Speed limit?  What speed limit?

Interestingly, random sampling has a major advantage though I've not seen it exploited properly.  My LeCroy claims to have it, but I've not observed its implemented in a useful fashion.  However, I don't have a way to get machine readable data out of the DDA-125 yet.

Aliasing arises because the Fourier transform of a periodic spike series is a periodic spike series.  The Fourier transform of a random spike series is a single spike, so aliasing does not take place.  I first encountered the concept in a dissertation by one of Mauricio Sacchi's students at Alberta, but was completely unable to understand what was going on.  The focus of the work was regularization of irregularly acquired data via the minimum weighted norm.  It enjoyed a brief period of interest in the seismic processing community, but then faded away because of some mathematical issues which had made me drop work on an implementation.

However, David Donoho and Emmanuel Candes showed that random sampling was a major advantage.  Donoho developed the idea and coined the name "compressive sensing" but it was part of a staggering burst of work he did in conjunction with Candes.  There is a catch  though, sparse L1 pursuits, the term I use because the application of the mathematics is so much broader than just data acquisition, is computationally intensive.  When they first started doing it for MRIs at the Stanford pediatric hospital it took many hours to process the data. I know that significant speed ups were developed to replace the original linear programming solution, but I don't know what the state of the art is now.  I've not read any of the professional literature on the subject since 2016 which is almost an eternity in a field evolving as fast as sparse L1 pursuits.

Leo's pulser with a 1 MHz clock will do a very good job as a pulse source for TDR using a DSO.  Once I get mine and the resistive splitters I ordered from China I plan to do a long thread on using TDR to solve vector network analysis problems. the major problem with the typical DSO is the rectangular passband results in severe ringing after the step.  So you need about 3 m of cable for a delay line.  But once you can window off the reflection of interest, it's trivial to get magnitude and phase over the BW of the DSO with 100 dB dynamic range.  So adding VNA capability to a $400 DSO with an extra $100 in accessories should be a real boon to people interested in HF radio construction.

I'm also going to consider trying to beat Leo's rise time, but I'm not really likely to succeed.  I'm hoping to find a dead SD-24 cheap enough I can disassemble it and study the Tek pulse generator.
David Hess:

--- Quote from: rhb on April 22, 2019, 09:28:33 pm ---OK, you're just describing my 11801.  BTW I just bought an SD-32 for $500 :-)  Speed limit?  What speed limit?

Interestingly, random sampling has a major advantage though I've not seen it exploited properly.  My LeCroy claims to have it, but I've not observed its implemented in a useful fashion.  However, I don't have a way to get machine readable data out of the DDA-125 yet.
--- End quote ---

Sorry, I mean random sampling versus sequential sampling as it applies to a sampling oscilloscope.  Tektronix made two sampling oscilloscopes which supported random sampling.  The advantage is that no delay line or pretrigger is required.  Without a delay line, bandwidth of the sampler is maximized because the delay line needed to generate a pre-trigger is not needed.

To give a solid example, your SD-32 has a bandwidth of 50 GHz but if you must use the DL-11 delay line to generate a pre-trigger signal, then its bandwidth is limited to the DL-11's 5 GHz of bandwidth.  My older 7T11 with S-4 sampling head can operate in random sampling mode to make full use of the S-4s 14 GHz bandwidth without a delay line under conditions where no pre-trigger is available.

In a TDR, random sampling would allow the TDR to measure the reflected pulse immediately after or even before it is sent (1) but since a TDR can always generate its own pretrigger signal of whatever duration is required, random sampling is never needed.


--- Quote ---Aliasing arises because the Fourier transform of a periodic spike series is a periodic spike series.  The Fourier transform of a random spike series is a single spike, so aliasing does not take place.  I first encountered the concept in a dissertation by one of Mauricio Sacchi's students at Alberta, but was completely unable to understand what was going on.  The focus of the work was regularization of irregularly acquired data via the minimum weighted norm.  It enjoyed a brief period of interest in the seismic processing community, but then faded away because of some mathematical issues which had made me drop work on an implementation.

However, David Donoho and Emmanuel Candes showed that random sampling was a major advantage.  Donoho developed the idea and coined the name "compressive sensing" but it was part of a staggering burst of work he did in conjunction with Candes.  There is a catch  though, sparse L1 pursuits, the term I use because the application of the mathematics is so much broader than just data acquisition, is computationally intensive.  When they first started doing it for MRIs at the Stanford pediatric hospital it took many hours to process the data. I know that significant speed ups were developed to replace the original linear programming solution, but I don't know what the state of the art is now.  I've not read any of the professional literature on the subject since 2016 which is almost an eternity in a field evolving as fast as sparse L1 pursuits.
--- End quote ---

I remember when this technique was advertised in DSOs but I thought it was HP.  Either it did not remain in use for long or they stopped advertising it.


--- Quote ---But once you can window off the reflection of interest, it's trivial to get magnitude and phase over the BW of the DSO with 100 dB dynamic range.  So adding VNA capability to a $400 DSO with an extra $100 in accessories should be a real boon to people interested in HF radio construction.
--- End quote ---

I have been arguing for this capability in low end DSOs for years.  All it takes is adding differentiation and returning instead of discarding the FFT phase results although the proper averaging and noise marker functionality would make it even more useful.

I have been very tempted to get an old Tektronix TDS series DSO or even an 11k series because they are capable of this.

(1) A must when doing maintenance on a TARDIS.  This explains the rarity and high price of 7T11 and 7T11A plug-ins; those who still do TARDIS maintenance can pay unearthly (Gallifreyan) prices on the Terran market and can rebid on lost auctions if they choose.
rhb:
LeCroy offers a RIS function.  I cannot discern any utility to it, but in all fairness I've not spent a lot of time trying to devise a test case.  I've not played with a Tek sampler that had the feature.

Most certainly no one has implemented what Donoho and Candes described in a commercial product.  You cannot get a PhD by repeating someone else's work.  And failure is not an option.  If you fail you have to start over on a new project. At least that's the case at reputable schools.
Navigation
Message Index
Next page
Previous page
There was an error while thanking
Thanking...

Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod