Super fast sample & hold under $10?

Am 05.02.2014 22:01, schrieb Joerg:

You are not the only one, I was looking for sth. like that also. You can get them if you buy a fast ADC on the same chip.

There is a fast T&H from Inphi, but not for $10. I have found nothing else. :-(

regards, Gerhard

I did go through those but, as you hinted, anything in a reasonable price range only has a T&H that goes to 250MHz. One went to around

So, looks like I have to roll my own. Again. ...

I think I'll hop on my street bike now and pedal away the frustration. Always seems to help and sometimes I get some fresh ideas.

Well, if you don't mind biasing it yourself, there's the HFA3102.

Cheers

Phil Hobbs

Thanks. But that's just a transistor array, like using the old diode quad. I guess I'll have to roll my own again.

This would be the real deal:

However ... 300 bucks :-(

For my masters project I could still obtain pouches of four matched RF diodes each. That is becoming increasingly difficult and nowadays, while you can get low capacitance bridge quads from Skyworks and others, they are typically non-stock.

What always boggles my mind is this: The cost difference between making your own beer versus buying a 12-pack of the good stuff is maybe 1:3. Makes sense. For very fast S&H or T&H it is easily 1:50. Makes no sense.

That's a nice chip. I can see a push pull with emitter limiter transistor application there.

Or maybe a gilbert cell, any one up for a mixer ? :)

Jamie

The capacitances are nice and low, very few hundred femtofarads. Unfortunately it's a SO-14 package which introduces lots of inductance. On fast switches that can really get you.

Some ADCs have screaming fast multi-GHz front-end s/h speeds, intended for sub-Nyquist sampling of RF stuff. 1 ns isn't especially fast in that world.

For $10 ? Put me down for one.

Tom

AD9204-20 has 700 MHz s/h bw, maybe 500 ps, $5.

I'll put you down for two.

I have never done fast S/H, so I'll excuse my suggestion forehand:

Wrap the sample and hold around a very fast comparator, like the:

You probably need an open drain output type, let a cap ride on the output. Charge the cap, Then enable the comparator which in turn via a resistor discharges the cap and the comparator shuts of its own discharge. Start AD conversion.....

Cheers

Klaus

Of course, the precision is perhaps not up to your requirements, since it has 150ps prop delay...

What is the requirement for accuracy?

Cheers

Klaus

But when looking at the fine print in here ...

... it says quote "The user can sample any fs/2 frequency segment from dc to 200 MHz". Something doesn't seem to compute with the data on page

Do you know any cheap one that's faster? I sure wish that instead of all the no-connect pins they'd pipe out the analog output of the S&H so people could use it sans the ADC section.

The problem is the switch, making the pulse is not a problem. The comparator doesn't have any switch but you must disconnect the cap from the signal source. Even Skyworks diodes are still 0.5pF (and unobtanium ...) which presents a ton of leakage at a GHz. Making a T or double-T switch out of something like that gets really big and unwieldy.

The prop delay does not matter. I need 100psec or so accuracy in pulse position but the delay is not a concern.

The data sheet is a tad ambiguous about what that 700 MHz thing means.

The s/h may not actually exist as such inside.

Do you need a s/h alone, or could you use an adc with a fast s/h inside? There are probably more ads with fast front ends... I just nabbed that one as an example.

I don't quite understand the application.

Could you use a diode mixer and a narrow pulse generator? That could be cheap.

Sounds like a job for a couple of pHEMTs in cascode. The SKY65050 is a very nice device if its high 1/f corner (~50 MHz) doesn't worry you too much.

Cheers

Phil Hobbs

We use SMS7621 diodes, around 0.25 pF. I can send you some.

A PHEMT might make an interesting series switch. They behave like jfets on steroids. An NE8509 has about 0.35 pF drain capacitance and is about 6 ohms Rds at zero gate bias. The capacitance could be "neutralized" with a balun transformer inverter and another cap, or some dual-phemt balanced thingie.

Small-signal? All the real data stops at 200 MHz.

"Analog input bandwidth is the analog input frequency at which the spectral power of the fundamental frequency (as determined by the FFT analysis) is reduced by 3 dB."

with the beer, you'll keep on drinking...