Low noise, high bias voltage on picoAmp TIA's input, howto?

The voltage output of a transformer is proportional to the d/dt of its primary current. Constant voltage output results from a triangle-wave input current, except at turnovers. So, the output-side rectified output only has a blip to cover with its filter capacitor, rather than the second harmonic of the drive frequency

Bill Sloman

I put the zip onto the little site (last position):

Thanks. I made an OCRed version using

(I already have a copy of the paper book, but it ain't searchable.)

Cheers

Phil Hobbs

It wants a password, and my security software was wailing about it.

Joe

Am 21.05.21 um 14:08 schrieb snipped-for-privacy@ibtk.de:

Hi, Timo! since you are interested in measuring small currents, you might like this one:

They even T-compensated the FET bias current.

I'm currently out of sync with the wall clock but happy since I repaired and calibrated my no longer booting HP 54750A sampling scope mainframe today. In the end, it was an empty Lithium battery on the CPU board. It was COMPLETELY empty, not a mV! With the new battery, it complained about missing cal-factors. I wonder how that thing worked the last years at all. Wow! 68020 with floating point coprocessor!

Note to selfes: Don't try to remove the box/handles/bottom feet etc. Impossible or futile. Spot-welded, unlike the usual HP/Agilent boxes. Just remove the backplate.

Cheers, Gerhard

Except that the blip is the current moving from one rectifier to another. The two alternative paths both have inductance, and the rectifiers have been known to act as step-recovery diodes, and that can generate some very high rates of change of current, and corresponding large voltage spikes.

The zip file isn't easy to open as text, and LTSpice 4 threatens to dump most of the content. E-mail of an un-zipped file would be easier. I can read German - if not all that well - but my wife could help with any difficult bits.

Bill Sloman, Sydney

LTSpice 17 did better, but the circuit is horribly messed up with resistors not hooked up and in the wrong place. I've made a start at making the schematic look more like something that would work, but it isn't going quickly, and the saturatable transformer model doesn't seem to have made it.

Using the John Chan model for saturation is nice, but does it seem to make the schematic very messy.

Using a bipolar transistor to drive the transformer may not help the efficiency - they do need base drive (though the FTZ694B is a high gain part) and they don't pull the collector as low as a good MOSFET pulls it drain, particularly when you have Schottky diodes in series with the emitters ....

Bill Sloman schrieb am Freitag, 28. Mai 2021 um 07:24:52 UTC+2:

Good morning Bill,

excuse me please, I did not tell about usage. You should expand the zip with another program than LTSpice and put all files in _one_ (working) directory. You may prefer a new one but not necessarily. LTSpice grabs all needed symbols from there, also the modified res.asy. This is explained here:

The chan model or the schematics become much more clear by using the Symbols: Trafo_1pri_1sek.asy or Trafo_3pri_2sek.asy which has to carry the identical name as ist's nets: trafo_1pri_1sek.asc or Trafo_3pri_2sek.asc

The model description is put into the *.asc files, the upper-/lowercase does not seem to be a problem.

The original resistors use a bad placed origin, it should be in the middle of the symbol. Additionaly I like to know, which is pin 1 in the schematic (e.g. to determine the direction of the current flow).

One may prove the power wasted by e.g. the bipolars using [Alt]+[left mouse button] on it. It is low in comparison to the whole efficiency loss. If I prefer FETs, I have to watch Vgs or complicate the gate-controls due to the 24V power which translates into (Pi)*24V=75V at the drains.

I am still not sure which version would be the more suitable for that application. At least I saw the bipolar version working. ;) The Schottkys are a try only together with the BE-diodes. Delete them...

Cheers, Timo Kirschke

Hi Gerhard,

Gerhard Hoffmann

You provide very interesting ideas every time! I remember your amplifier attempts. Our "customers" like to see 0,5pA as resolution, not accuracy, so I'm in a relaxed situation with the ADA4350-1. Much simpler than the swiss approach. Interesting, that they seem to use ordinary FR4 instead of ceramic based layers. Congratulations, that you managed the scope fail! In earlier times we left the disco at that time, not the lab...

Thank you very much! Cheers, Timo

Thanks. I had to put in three ferrite beads - I picked the 1uH Wurth parts from the LTSpice list - and add a Spice model for the FTZ694B - but after that the circuit did simulate and produced sensible waveforms. I had to stretch the simulation periods out to 10msec - there was something odd happening around 2.2msec.

I agree that MOSFETs are bit trickier to drive than bipolar transistors - I tend to use a one turn base drive coil between the bases of a PNP long-tail pair, and switch the tail current between two gate to source resistors. but you typically have a 20V rating between gate and source, which allows plenty of head-room.

Bipolar transistors can give you squegging - which LT Spice doesn't simulate because it seems to depend on something odd that inverted bipolar transistors that the Gummell-Poon model doesn't capture.

No password required when I use it.

Cheers

Phil Hobbs

I just checked again, got the djvu password request.

You may have your credentials cached, so it doesn't keep asking you.

If you come in using a private browser window, what happens?

Joe Gwinn

In fact, that seems to have been what was gong on from 2.0 msec to about 2.7msec.

The current through L7 started going negative - to about -5mA, and there were a negative current spikes through Q1 and Q2.

The problem is that this goes away in Spice simulations, but have been known to persist in real life. In real life you'd probably reduce the inductance of L7 until it went away.

Thank you for sharing your valuable experiences!

Weird. If I just type djvu.org into the address bar, it comes up in http:// and works normally. The https:// thing works as you say.

Use the http:// one. ;0

Cheers

Phil Hobbs

Using plain http did get rid of the password dialog, and did get me to the any2djvu page, which is not an OCRed copy of anything.

Joe Gwinn

Try it out on an actual PDF. There are several options for getting embedded text, two of which involve OCR. I've done hundreds of documents that way, from one-pagers up to full-length books. It's Good Medicine.

Cheers

Phil Hobbs

Ahh. I thought that an already OCRed pdf was there. I don't have the original pdf yet. Where do I find that?

Thanks,

Joe Gwinn

Timo's link above.

Cheers

Phil Hobbs