A low noise byproduct

[syn08]

A byproduct of the HPS 3.0 project.

2 sq. inch of PCB with 64 unsorted BF861B in parallel (no source degeneration) cascoded with a 2SC5171 followed by a AD797BRZ for a total gain of 1000, SMD tantalum oxide thin film resistors, 1uH gate inductance for each 4 JFETs. Total drain current is about 750mA. Power supply is +/-18V from a low noise HP6627A plus a local parallel (bipolar, MOSFET was way to noisy) regulator with a OPA211 opamp, decoupled locally by Nichicon CG series, 13mOhm ESR, 1000uF electrolytics. Everything is SMD except the 2SC5171. JFET Vds is 2.7V to keep the dissipation low. A polypropilene cap between the JFET stage and the opamp.

0.12nV/rtHz

Temperature in my basement: +19 centigrades. A small fan is blowing over the PCB, however bringing my finger at 1cm over the JFET area raises the noise to 0.18nV/rtHz in about 10-15 seconds. Don't ask me to explain the spectra, it seems to have 1/f, some GR noise and god knows what else (at LF).

If somebody wants to build a phono amp based on this, let me know. I don't plan to 😉

Edit: don't know what KE5FX plotter emulator is doing (obviously something is wrong, missing symbols, etc...), but this is the result of 1200 averages over an entire night.

 

[jackinnj]

What does the input impedance look like?

In lower America I am running the basement at 22 C.

 

[infinia]

What does the input impedance look like?

In lower America.

Where would South America be then? lower lower and more lower America😉

 

[scott wurcer]

So the L's worked. If the scale really reads 8dB/div looks like 10dB/decade below 100Hz or so, almost exactly 1/f. A 50-60Hz corner is typical of what I find with these, pretty good for RF FET's. They probably can't afford not to have a clean fab these days. This would make a nice transformerless ribbon mic input but the gain might be too high.

 

[syn08]

What does the input impedance look like?

In lower America I am running the basement at 22 C.

1. I don't know.

2. This is Canada, mind you. I recall the good ol' times down in Galvestone running the AC in the New Year's eve.

 

[scott wurcer]

The FET board would probably still work at 77K.🙂

 

[syn08]

So the L's worked. If the scale really reads 8dB/div looks like 10dB/decade below 100Hz or so, almost exactly 1/f. A 50-60Hz corner is typical of what I find with these, pretty good for RF FET's. They probably can't afford not to have a clean fab these days. This would make a nice transformerless ribbon mic input but the gain might be too high.

I can adjust the gain, 1000 is for measurement purposes only. Can't go lower, required RBW would be to small.

Yes, L's work fine. BTW, the equivalent drain resistor is about 15 ohm (16 x 249 ohm in parallel).

Edit: yep, its 8dB/div and hence 1/f but only down to 2-3Hz. Below that there's some GR noise, probably slow surface states exchanging charge.