Creating SPICE models

So I've been designing a Joerg-esque super cheap single diode sampler. With vaguely realistic parasitics, LTspice says it should give a sampling FWHM of 120 ps or a bit faster, all for well under a buck excluding the timing generation part, but including the pulse speedup part.

In the spirit of keeping things cheap and small, I'm using a single JFET switch and a hold capacitor to store the sampled voltage until the ADC gets round to coverting it.

Since it's all on a single 5-V supply, I need a small JFET with super low pinchoff voltage. The best one I can get for cheap seems to be the

2SK3796-3, which is about 9 cents in reels from distributors.

Unfortunately, there's no SPICE model that I can find.

So since I'm not going to buy PSPICE just for this one part, what else do you folks use for model creation?

Thanks

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs
Loading thread data ...

Are you overt while you are coverting ?>:-}

System modeling is easy, device models that are accurate require expensive parameter extraction test equipment.

If you can find a model with the same parasitics, but wrong threshold, you can just twiddle the threshold. ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
 Click to see the full signature
Reply to
Jim Thompson

Why not use a cmos opamp?

What are you using to make the sampling pulse? Transistor? SRD?

A dual-diode sampler isn't much more expensive. There are SRDs available for under a dollar, and dual diodes are cheap. One of the Tek plugins, the 1S2 I think, had a 2-diode sampler dumping into a big capacitor, which did the "smoothing" function and eliminated all sorts of concerns.

1-bit samplers are interesting too. Use a flipflop or a strobed comparator for the sampler. Everything is low impedance, and you don't need a narrow sampling impulse.
--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

Too slow. The front-end sample cap is only a few pF, so I have to do the sample-and-hold in a couple hundred ns at most. I have a discrete BJT amp for the first stage--its base current supplies most of the reverse bias for the sampling diode.

One of your LVDS receivers running nearly unloaded, followed by a practically infinitely fast transistor. Its voltage gain is what sharpens up the edge.

A buck is the budget for the whole thing. ;)

I'm differentiating the output of the fast transistor to make the sampling pulse.

So does anybody have a free/cheap way of making models from datasheet curves? PSPICE has a model editor. I can probably cruft together a test bench in LTspice, but I'd much rather learn from other people's mistakes. ;)

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

It's a hush-hush project. ;)

I've done that before, but it's kind of hit-or-miss. Ideally I'm looking for something like the PSPICE model editor without PSPICE attached.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

Shorted transmission lines work too.

I just breadboard stuff like this. It's faster, more fun, and you'll need to try it out in real life anyhow. Spice can run 12 orders of magnitude slower than real life.

--
John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  
 Click to see the full signature
Reply to
John Larkin

Am 26.10.2016 um 18:39 schrieb Phil Hobbs:

It used to be well known that if a BIP model featured a certain value for Rbb (50 Ohm or so), it was made by Pspice model generator. No matter how fat or tiny the transistor was.

Maybe they have improved in the last 20 years.

regards, Gerhard

Reply to
Gerhard Hoffmann

The PSpice Model Editor is junk. I've input data, generated a model, then tested the model and it doesn't return a match to the input data. ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
 Click to see the full signature
Reply to
Jim Thompson

Well, since you're PSpice's #1 fanboi, I guess that's pretty definitive.

Cheers

Phil Hobbs

Reply to
Phil Hobbs

As long as you don't have to stare at it, it seems like only 11 orders. ;)

The SPICEing was to see how good it ought to be. Once we get client approva l, the next stage is a board with a bunch of test structures and a best gue ss at the final configuration. We ought to zero in on the final final confi guration pretty easily if the test structures are well thought out.

Fun.

Cheers

Phil Hobbs

Reply to
Phil Hobbs

Yep. I never use it. ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
 Click to see the full signature
Reply to
Jim Thompson

That basic stuff, like seeing how much charge you'll get, closing the loop if any (is it a feedback sampler?) is good to Spice, but fast stuff is tough to sim accurately.

I actually possess a Lumitron single-diode sampler box, from roughly

1960. It uses an avalanche transistor and a 1N34 microwave diode, totally open-loop. It preceded the similar Tek type N sampler. Both were horrible.

formatting link

You should meet Mark Kahrs some day. He's the expert on the history of sampling.

formatting link

LeCroy did a very fast single-diode sampler not so long ago.

--
John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  
 Click to see the full signature
Reply to
John Larkin

I should try that sometime. In this case the signal level isn't that high--it's a special-purpose TDR.

The differentiator is actually a small cap running into a 50-ohm stub about 50 ps long, with a 20-ohm resistor at the far end. In spherical-cow spiceland that seems to give the best pulse shape, but we'll see. The circuit is simple enough to stuff several boards to hack up. I'm going to put a few dedicated ~25x capacitive-division things on it so that I can figure out what the actual waveform looks like, without screwing it up in the process. It's great having well behaved microwave devices available so cheap, such as the SKY65050 pHEMT.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

I'll try to find the time to try Intusoft's model maker.

--
 Thanks, 
    - Win
Reply to
Winfield Hill

*SRC=sm_NJFET;sm_NJFET;SpiceMod;Test;DUT *DWG CIRCUITS\SpiceMod\NJFET\NJFET.DWG
  • JFETs N;Depl;30.0V 5.00mA 200ohms .MODEL sm_NJFET NJF( VTO=-950m BETA=6.50m LAMBDA=1.00m RD=28.0
  • RS=25.2 IS=1.32f PB=1.00 FC=.5 CGS=2.00p CGD=3.65p )

I didn't try running the SpiceMod model.

--
 Thanks, 
    - Win
Reply to
Winfield Hill

Thanks, Win, I'll have a try using it tomorrow.

Cheers

Phil Hobbs

Reply to
Phil Hobbs

It had a parameter, maximum Id, which I didn't understand. I think I entered too high a value.

--
 Thanks, 
    - Win
Reply to
Winfield Hill

I'm enhancing it by about half a volt transiently, so it'll probably get there. ;)

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

I did a 4-layer PCB layout not long ago, with a bunch of test circuits for some new products. I threw in a TDR that uses a 1-bit sampler, which should get me well below 100 ps, maybe below 50. Unfortunately, I had MyroPCB make the board, and the pads keep falling off, so I haven't bothered to build it. When I get time, I might order another batch of boards from someone else and try it.

It wouldn't be as cheap as the thing you're doing. $20 or so with the timebase.

Do you have an SD14 probe sampler? 3 GHz and some absurdly low capacitance at the probe tips.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

I have two or three of them, as well as an SD-20 passthrough head. I also have a few P6249 FET probes for the TDS 694C, which are in the same league (120 ps) but allow continuous digitizing.

With only a ~4 pF sample cap, even half a picofarad is a lot. I'll put a pHEMT common-drain amp on the back of the test board, connected by the capacitance of a nearby via, and use the FET probes to check its response. (It's okay if the probe loads down the node as long as the pHEMT sees the same thing.) With a SKY65050, I don't even have to use bias resistors, so the amp's input capacitance can be below 1 pF, and its Zout can be 50 ohms. I did one like that for Samsung five years or so ago.

Got client approval today, tally-ho!

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

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.