Diode sampler (again)

Maybe you could start here:

But in essence, you just build it. I designed my first really fast one at around age 25. Never looked at publications because I found this didn't make much sense. I had to make do with what I could get, what I could buy. Fast diodes where I could obtain quads without breaking the bank, toroids that could be had without too much of a wait, and so on. Then ... click ... turned on the Weller.

Here's a dual-diode sampler I built a while back.

ftp://jjlarkin.lmi.net/Sampler1.JPG

It had about a 70 ps rise time, 5 GHz bandwidth. I'm working just now with a couple of EE students, on their class project, to do a more modern, uP/USB version, TDR maybe, deconvolution math and such. I expect fun.

John

Tektronix once published a series of books on 'scope design. IIRC one of them described some of the basic equations associated with diode samplers.

Looks simple. I like that. No equations though :-)

I'm thinking along the same line.

It seems the book "Sampling Oscilloscope Circuits" is one I need to look for...

Just very basic stuff. Some transmission line calcs, simple charge calculations, opamps and such. 4-function-calculator math, Appcad for the trace widths.

I included a blowby compensation circuit, but it didn't seem to need it. I'm still not sure what all that was about.

Our plan is to build a fairly ugly TDR system (which is what limited experience/time/resources will allow) and beautify it with a deconvolution-trained FIR filter, in the PC application.

John

Thats just experience talking :-) Anyway, I got something working in the simulator after I figured out current needs to flow one way through the sampler diodes. Feed-thru is still horrible.

The circuit on the left I presume.

I can't decide yet whether I just want to try and learn something interesting or turn this project into a scope.

What diodes are you assuming? You need tiny schottkies, 0.2 pF maybe. SMS7621 types. I have some beam lead samples from Aeroflex that run around 0.08 pF, but you'd have to be a real pain lover to mess with them. Somebody makes a 0.1 pf class schottky in a managable surface-mount package, but I can't remember who.

The diodes would be back-biased by ballpark 2 volts, and get whacked by, say, 4 volt bipolar sampling impulses. Allowable input signal swing will be small, +-1 volt or less. Sampling efficiency (voltage glitch coming out of the sampling caps, compared to actual signal swing) is typically just a few per cent in fast samplers, so you need a lot of downstream signal processing, opamps and gates and integrators and such, for it all to make sense.

Yes. I'll have to think through the blowby thing again, some day soon.

Do both! This picosecond stuff is really cool.

John

Good question. I got some diode parameters from a research paper. I was under the impression that pin diodes where the way to go but I can imagine the much higher forward voltage (compared to schottkey) may be a killer. I'll substitute some models from real world diodes later on.

I wasn't planning on going that far. 2GHz bandwidth is OK for me which I think should be feasible with 'standard' components. I'm planning on using a 100kS/s multiplexing ADC so it will be slow already.

If time permits :-)

PINs are p-n junctions, so are too slow (forward and reverse recovery, both) to use as samplers. The very first semiconductor samplers used silicon point-contact diodes, which have schottky junctions.

You could use regular schottkies, 1N5711s maybe, at a GHz or so.

John

I simulated with BAS70 jelly-bean schottkies (

What are you using to generate the sampling pulses?

John

come

range,

I'm thinking about a pulse transformer and some fast logic. First I want to simulate that as well. It depends on what cheap pulse transformers are available. Maybe a common LAN transformer will do.

My circuit looks much like the circuit in this paper:

come

can't

range,

bandwidth.

one

the

on.

Wind your own. Seriously. First time I learned that lesson was on my masters project when I needed fast samplers for the CCD camera I was developing. I searched for at least a couple of days, then cussed a bit (I try not to do that anymore ...) and wound my own within a couple of hours or so. Could have saved myself those two days.

The trick is a good core and strict adherence to multifilar winding techniques. Courtesy of the ARRL Handbook in my case. There are also nice coax transformers you could make.

t) to come

t I can't

age range,

s is how

andwidth.

some

fast one

d this

t, what I

aking the

nd so on.

ge

d for

in

be.

can

be

on.

y,

I'm with Joerg here. The last time i seriously needed a fast transformer - twenty years ago - I carefully wound one with minature coax, and it worked beautifully.

Just to anticipate the obvious question, I then wound a transformer with a twisted pair of enamelled transformer wire (which was a lot cheaper and we already had two different colours of enamelled wire in the company stock). It worked better than the minature coax - mainly because I had space for more turns of transmission line, which did mean that the high frequncy limit (set by the length of the transmission line) was lower, which didn't matter in my application.

At the frequency we were working with at that time, I could get away with winding the twisted pair onto a standard RM-core coil form and soldering the ends of the wires onto the pins of the form - much easier, cheaper and more compact than minature coax connectors on flying leads.

Does anybody do a surface mount coil form these days?

-- Bill Sloman, Nijmegen

to come

can't

range,

how

bandwidth.

fast one

this

what I

breaking the

so on.

Lodestone Pacific.

John

come

can't

range,

how

bandwidth.

one

what I

the

on.

Shazam! We finally do agree on something :-)

A lot in transformer design boils down to the art of winding them. I found it tough to explain and convince even seasoned engineers of coaxial transformer techniques. Often when I wind one myself, put it in their circuit and some hardcore problem vanishes they still think this is all black magic, voodoo, shamanism, coincidence or luck. Especially since I normally use their cores.

One engineer insisted that my results were pure coincidence. So I wound several more which all worked. "How many more should I make?" ... "Ahm, nah, that's ok. But please write a very, very detailed instruction on how to manufacture those. With lots of drawings in there. Now can you also make this table move without touching it?".

Yup. Besides John's recommendation there are also places in Europe which should be easier for you to order from. One example is on page 83:

Can't help you with Australian sources though :-)

to come

can't

range,

how

bandwidth.

fast one

this

what I

breaking the

so on.

I think I put this project on ice for now. I want something simple which can be build from cheap components. I need to let the whole thing sink in first.

:-) LOL Knowledge is (magical) power.

Samplers don't need transformers.

Quitter!

John