1. Home
  2. Electronics Design
  3. An Affordable 160 GHz Sampler

An Affordable 160 GHz Sampler

Conventional samplers for home brewers usually go to 1 GHz. The SD-32 sampler for the Tektronix 11801C mainframe goes to 50 GHz. The HP 110GHz oscilloscope costs around $1.3 Million USD, with a 10-bit resolution. Very impressive, and very expensive:

formatting link
I have invented a new sampling technology that promises 160 GHz bandwidth, yet is affordable to home experimenters. If you can afford an IPhone or IPad, you can afford this sampler.

This technology is not pie-in-the-sky. I made a basic 5 GHz version for the University of Ludwigshafen, Germany, and they were very pleased with the results. This was the first prototype, and I have made significant improvements since then.

The sampler offers substantial advantages over conventional samplers:

  1. conventional diode samplers can have significant loss using multiple stages to acquire the sample, perhaps 40 dB or more. This degrades the SNR, especially for low-level signals. The new sampler has no such loss and operates on the input signal directly, giving maximum SNR.

  1. conventional samplers produce samples that combine the actual signal plus noise. Averaging can improve the SNR up to a limit, where it takes too long to improve the SNR. The improvement is given by sqrt(N), where N is the number of samples. Improving the SNR means doubling the number of samples, and eventually the doubling takes too long to be practical. In addition, the signal can drift during long sampling times making the results useless.

The new sampler bypasses this limit by restricting the amount of change that can occur in each sample, so the conventional equations no longer apply. It can easily recover signals buried in 30 dB of noise, which is impossible with conventional samplers.

I have two questions for the SED group:

  1. where would a sampler with this bandwidth and performance be useful?

  1. where can I find signal sources at these frequencies to check the response?

Thanks

Reply to
Steve Wilson
Loading thread data ...

Impressive, if true. I'm afraid you'll have to be more convincing than that though.

Jeroen Belleman

Reply to
Jeroen Belleman

Sounds really impressive

That price/performance ratio seems like you could make millions just selling it

I guess that is why the details offered are limited

Reply to
Klaus Kragelund

The exact same posting (word for word) was made on the Time Nuts reflector today, only under the name "Mike Monett" snipped-for-privacy@teksavvy.com. So, who is "Steve Wilson"? Is either a true name?

It turns out that this first appeared circa 2003, on electronics newsgroups. Google on "Mike Monett ghz sampler" (without quotes) for the stories.

Such as:

formatting link

John Larkin was not impressed back then.

What has changed since then? It's been almost 20 years, and undersampling is now common, often being called "operating in the Nth Nyquist zone" (ie, above the Nyquist sampling limit) or the like.

Joe Gwinn

Reply to
Joe Gwinn

I did my EE senior project on the "binary sampler", which I called a slideback sampler. I used tunnel diodes. I did it all the night before the project was due. I won some awards and had to pesent it an an IEEE meeting full of old farts who couldn't understand it. My wife came along to hear me present, and we were privately pulled aside and told that women weren't allowed at IEEE meetings.

It was based on an idea in the GE Tunnel Diode Manual. I told Mike about that.

5 GHz bandwidth is fairly easy using a strobed comparator or a fast d-flop as the 1-bit sampler. We use a similar technique to make jitter measurements with a 30 fs noise floor.
Reply to
jlarkin

The same Mike Monett who used to endlessly witter on about colloidal silver and other lunatic-fringe theories? I hoped that we'd managed to chase him away..

Reply to
Clifford Heath

Mike Monett is the name on my patents. I do not use it online due to spam and risk of malware.

Excellent link. Thanks

Yes, I remember your slideback sampler. You didn't say it was from the GE Tunnel diode Manual. I had to find that out by myself.

When I suggested using a d-flop, your recommendation would have shorted the output of a mixer or emitter follower to ground through the 50 ohm input impedance. You also claimed the circuit would rail, even though it would be continuously sampling and following the input signal. You also said nothing about the SNR improvement over a conventional diode sampler, or the intrinsic noise rejection of the binary sampler.

These and other improvements are a far cry from the circuit we discussed back in the old days.

Here are the posts:

formatting link

Reply to
Steve Wilson

I will post complete details so anyone can make it.

What would you use it for?

Reply to
Steve Wilson

It uses conventional off-the-shelf components. If you give me your email address I can send you comparisons of waveforms with a Tektronix 1502 TDR and a photo of the top of the pcb.

If you are really interested, I can send you photos of the lab and the office of the graduate student I worked with in Ludwigshafen. He got his PHD as a result of this work.

Reply to
Steve Wilson

We were just talking, not writing a thesus. A simple "thank you for the idea" might be preferable to whining about all the things I never did for you.

Long ago National Semi did a paper about using a TTL dflop as a similar sampler. Can't find it online. I did a pretty fast sampler using a GigaComm flipflop, just for fun; it has all differential inputs and that's nice here.

I'll refund all your money.

The big bucks now are in telecom, which needs a realtime digitizing scope to nab constellations... over a megabuck worth in some cases. Samplers are not so popular these days.

Shock-line samplers peaked at about 100 GHz, LeCroy I think. There's apparently not much incentive left.

There are electro-optical samplers, based on femtosecond lasers, that do hundreds of GHz.

Here's the results from our jitter measuring gadget that uses a similar idea, walking the d and clock of a flop across one another:

formatting link
Standard dev is around 50 fs, so a 1-bit sampler that fast should be possible.

Reply to
jlarkin

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.