Nanoamp calibrator again

"Couldn't find symbol npnsub"

Fixed--see the web page again.

Thanks

Phil Hobbs

Phil Hobbs a écrit :

I did thought of that then, but for some reason the post didn't get through.

Didn't find obvious solutions. Anyway some 'random' thoughts:

- lower the cap as much as possible to get a better L/C ratio

- use several smaller Cs in //

- lowering the cap means you need lower non linear parasitics. One way is to use a resistor as a current source and a low frequency parallel path to linearize the exponential ramp.

- esl is a pb so 'eliminate' it, that is induce a compensating emf (have

GND ---||---sig2 sig ---||---GND GND ---||---sig2

With the right caps ratio there might be a solution :

---+------+------+ | | | [R1] [R2] [R1] | | | sig2+ | + sig2 | | | | | | 0.05p | sig+--------------||-->

| | | C1 | |C2 |C1 --- M --- M --- --- --- --- | | | | | | GND GND GND

M is the mutual between the C1s and C2

Another try might be to precompensate for the L-C high frequency response. Tried it: not so easy :-(

Also, for the resistor current source, the parasitic cap might as well be a problem...

Finally, maybe you can have 2 coupling pads, one with fast, low amplitude pulse (OK saw tooth) and one with a slower, higher amplitude one for different frequency band of interest. This might ease the L pb somewhat.

I just did but the sym is still missing. Have you built it and tried meandering back and other tricks to neutralize some of the inductance? Can't do that on the simulator though, at least not on LTSpice.

orand see.

Only the brute force approach. (Is L4 the 'offending' inductance and C10 the gimmick?)

Less C, more R and more V.

Could you make the C a transmission line type thing?

George H.

Weird. All of those are defined right there in the .asc, which runs fine here.

I took some suggestions and added a few more tweaks, including a critically damped RLC filter at the output, and it looks okay now with 1 nH in series with each of 10 parallelled 39 pF capacitors.

The transient gate current of pHEMTs can be pretty significant when you're only running them at 70 uA I_D. Also the beta linearity and Early voltage of PNP RF transistors is really, really horrible.

With reasonably realistic strays put in, it looks like the transients are all over in a nanosecond, and they're not a lot larger than the DC, so it should reproduce the 3.5 ns rise time of the preamp pretty well.

Cheers

Phil Hobbs

OOPS! tried the LTC program and got the following errors: Questionable use of curly braces in "symattr spiceline rser=0.005 lser={lcap}" Error: undefined symbol in: "[lcap]" Questionable use of curly braces in "symattr value {lbead}" Error: undefined symbol in: "[lbead]" Questionable use of curly braces in "symattr spiceline rpar={rpbead}" Error: undefined symbol in: "[rpbead]" Questionable use of curly braces in "symattr value {rbi}" Error: undefined symbol in: "[rbi]" Questionable use of curly braces in "symattr value {lbead}" Error: undefined symbol in: "[lbead]" Questionable use of curly braces in "symattr spiceline rpar={rpbead}" Error: undefined symbol in: "[rpbead]" Questionable use of curly braces in "symattr value {lbead}" Error: undefined symbol in: "[lbead]" Questionable use of curly braces in "symattr spiceline rpar={rpbead}" Error: undefined symbol in: "[rpbead]" Questionable use of curly braces in "symattr value {lbead}" Error: undefined symbol in: "[lbead]" Questionable use of curly braces in "symattr spiceline rpar={rpbead}" Error: undefined symbol in: "[rpbead]" Questionable use of curly braces in "symattr value {lbead}" Error: undefined symbol in: "[lbead]" Questionable use of curly braces in "symattr spiceline rpar={rpbead}" Error: undefined symbol in: "[rpbead]" Circuit: Version 4

Fatal Error: Multiple instances of "Flag" Maybe one of the problems was that all of the files were in a folder on another drive.

Look harder..

• Cute!

• Or maybe some feed forward or some positive feedback?

Check; it disliked the ASC - a fair bunch of errors.

ip

the

, of

r 1

more

ic,

lder

Excellent! (that's semi-brute force)

George H.

t -

Ran first time for me.

lder

Lucky you. I just got a long string of error messages, suggesting that whatever end-of-line symbol Phil uses in his longer Spice command strings isn't recognised as such by my version of LTSpice.

-- Bill Sloman, Nijmegen

I posted a rebiased version that runs off +5V and gets the edge artifacts down to something reasonable. It uses single-ended drive for the pHEMT differential pair Z2/Z3, in order to prevent capacitive gate current in Z3 from corrupting the output. Using ten 39 pF caps in parallel reduces the equivalent series inductance, and adding an RLC filter to the output makes the transition a lot cleaner.

The output as before is I(C5). The simulation shows all sorts of ripples at up to 12 GHz elsewhere in the circuit, which I really don't think are real. (The pHEMTs are that fast at 20 mA of drain current, but surely not at 70 uA.)

Anyway, I think this can be made to work, assuming that the layout person knows something about RF.

Cheers

Phil Hobbs

I took the .asc file and copied it directly to the web host using sftp. I suspect that your client mungs the line endings somehow.

Cheers

Phil Hobbs

I still get "Couldn't find symbol npnsub".

It uses single-ended drive for

And think about tricks like meander-back to reduce inductance.

The .asy file is there on the webpage--the LTspice NPN symbol doesn't work with a subcircuit, so I had to make another one.

Care to expound?

Cheers

Phil Hobbs

It's usually best to include all non-standard models right inside the SPICE file and then declare a regular parts a subcircuit.

The BFP640FESD model is sort of large and complicated, and I don't think you can put a symbol in a .asc file--can you?

Cheers

Phil Hobbs