wideband FM demodulator

That is how it's done in FM ultrawide bandwidth radio (UWB). Page 26 here:

If the signal is always strong and of more or less constant amplitude a simple phase discriminator in an FPGA plus an external delay line might suffice:

High precision will be a challenge because most delay lines are temperature sensitive. Then there either has to be an occasional "pilot tone" for auto-calibration or a 2nd thermally coupled delay line and phase detector with a reference signal going into it. Plus a heater underneath (meander trace, resistor, FET, or other).

Wide bandwidth SAW filters might also be useful and save real estate.

There's probably a patent somewhere.

But scopes don't need a lot of linearity or dynamic range. Most are barely 8 bit machines. They can, probably do, put a switchable attenuator ahead of the FM isolation link.

The dominant distortion of a delay discriminator running at constant amplitude is cubic. That can be corrected using a second discriminator with three times the delay, and adding something like 1/3! of its output to the first one. (The trigonometric series isn't the same as the Maclaurin series.) That would leave the dominant distortion as fifth-order, which ought to be small enough to ignore.

Cheers

Phil Hobbs

There's a faction here that just wants to digitize the signal at 250 Ms/s and ship it as 8B10B serial data. Makes some sense. The ADC will cost $60 or so, but it's a lot easier to think about than FM. I took one course in communications theory and it was hard.

Some fun dithering things could be done to keep me amused from an otherwise boring digital design.

But it's too pedestrian. There has to be some sportsmanship in engineering.

The ADC will

Ouch. Why that much? Lots of bits?

FM with a delay discriminator should be easy but wouldn't just PWM-ing it over be enough?

thanks, I didn't note that he had such a wide band-width until after I posted, it's not like this is a 5Mhz tv signal

onsdag den 31. oktober 2018 kl. 21.04.57 UTC+1 skrev John Larkin:

how many bits?

I requested an RF chapter in AOE3 but...oh well.

If you understand patent claims, you can easily find a way around them. There are probably many patents for transferring information across a barrier. Even a slight difference constitutes a different invention. This is how pharmaceutical companies get new patents on virtually identical products. Your application is different, therefore most existing patents won't apply.

True, but that has nothing to do with your application.

Often that's just synchronous AM. Which may also be doable in John's case but hard to say without knowing dynamic range specs and stuff.

Most patents these days are broad grab attempts along the lines of "Let's throw all these spaghetti at the wall and see which ones stick". It is assumed that litigation will weed out bad claims or maybe whole patents.

I may be wrong but I doubt that they have an FM link.

Careful with that advice. Patents are legal documents more than technical ones. What constitutes a sufficient difference to establish noninfringement depends on a lot more than the claims. Google "doctrine of equivalents" for instance.

I understood that to mean that a solution good enough for 7-8 bits wouldn't necessarily be good enough for John's.

I'll have to do the math for the trig series and try out a linearized delay discriminator. I've been doing way too many BOMs and ECOs and stuff lately.

Cheers

Phil Hobbs

A target might be 50 to 100 MHz bandwidth and 75 dB dynamic range. The comm link will be fiber.

If we go digital, we could digitize to 14 bits at around 300 MHz. With some carefully planned framing overhead, thet would fit into a 5 Gbit/sec fiber link.

My Tek DPO2024 does, I think, use a roughly 3 GHz FM link for the channel and trigger isolation. Tek does have some products that use optical isolation.

If you go as far as fifth-order, you can get the nonlinearity below 0.1% of reading in a fractional bandwidth of 65%.

See

Cheers

Phil Hobbs

We'd need 14 to meet one customer's dynamic range spec. We have a nice

I could just buy a 310 MHz 14-bit ADC and pass along the cost.

Sampling at, say 300M, we could use 8B10B serial transmission at 6 Gbits/sec.

Dang, all this digital stuff makes design boring.

Am 01.11.18 um 18:15 schrieb John Larkin:

AD9694. 4 ADCs 14 Bits/500MSPS each, JESD204B serializeres included. Up to 15 GBit/s per lane if wanted.

That's where digital turns analog again.

regards, Gerhard

torsdag den 1. november 2018 kl. 18.15.39 UTC+1 skrev John Larkin:

HMCAD1520, ADS4129, ADS41B49 seems to be in the ballpark substantially cheaper

With that 75dB spec I'd use FM and maybe a delay line detector.

You could do that but it'll be much more hardware effort than FM.

They have nice iso scope and I could really use one right now. However, for just one job I can't justify spending the equivalent of a small car.

How would I use that?