PSRR of frontend buffer amplifier

[gf]

I'm trying to get a clue what PSRR one could expect from the "FET IMPUT AMP" subcircuit in this schematic:
http://www.eevblog.com/files/Rigol-DS1054Z-Schematic-FrontEnd.pdf
(particularly for the negative V- supply rail)

The buffer amp circuit in the frontend of my Hantek 2D72 looks almost identical (as far as I briefly investigated the PCB, there seem to be only minor differences and slightly different component values) and I'm facing significant noise on the two scope channels which originates in the output amplifier of the built-in AWG (when it drives a 50 Ohm load with a 500 Hz square wave @ 2V amplitude). This noise primarily is a (distorted) 500 square wave too, with some additional higher-frequency components, and it is obviously injected into the frontends' FET buffer amplifier via the power supply rails - I think in particular via the two current sources. It is not present at the input (gate of the FET), but appears first at the source of the FET. And it is present on the power supply rails, too.

EDIT:

A more detailed inspection reveals the my scope's FET buffer stage and its component values are closer to this one
https://www.mikrocontroller.net/attachment/175587/Hantek_Tekway_Voltcraft_DSO_hw1007.pdf, page 2.
(Still the basic circuit with source follower + emitter follower + two current sources is pretty similar)

Initial thoughts regarding the circuit:

The current sources are biased with a resistive divider R01_23 and R01_24 from GND to the (the noisy) V-, therefore I'd expect their "constant" current to carry the noise as well. The bypass capaciter C01_11 from the divider's tap to V- seems too small to help at 500 Hz. I'm wondering in which amount the noisy "constant" current affects the buffer's output voltage then - I guess this transimpedance dominated by the output impedance of the source follower, is this correct?

Besides the fast path via the FET, there is still the the low-frequency/DC path via the opamp, with feedback from the output, but this path seems to join-in only at even lower frequencies (50 Hz or even lower), so its feedback can't compensate the injected noise either at 500 Hz.

[Kleinstein]

The output resistance of the source follower with a current sink at the source side is essentially the inverse of the  trans-impedance of the JFET. So with a 10 mS JFET like J309 at some 1-5 mA this would be on the order of 100 Ohms.

The negative side PSRR could be one path, but there are other paths too, especially a direct ground loop.

[David Hess]

I agree with Kleinstein.  Further the transconductance of the emitter follower is higher even at the same current so it has an even better PSRR.

The designs looks essentially identical to me.

Tektronix did it a different way with the low frequency correction amplifier controlling the JFET's current sink so power supply noise was suppressed.  I suspect the advantage in these examples is that no separate compensation adjustment is required.

[gf]

I agree with Kleinstein.  Further the transconductance of the emitter follower is higher even at the same current so it has an even better PSRR.

The latter does unfortunately not help if already the 1st stage suffers. My guess was too, that the source follower's output impedance is not that low (making also the 2nd emitter follower stage necessary). Thanks for the confirmation, Kleinstein. Considering only the 500 Hz square wave component of the ripple/noise, I measure ~7mV pp on V- and 2mV pp at the FET's source, so given the resistor values in the circuit, a transconductance of about 12-13mS (70-80 Ohms) would indeed fit with the observed picture and lead to the observed PSRR. There is no ripple at the input (gate) yet, on the signal path it firstly appears at the source of the FET. I also can't see it at various ground points.

But the low-frequency V- PSRR is only ~10dB then . IMO not to much for a frontend which is supposed to deal with millivolts...

The designs looks essentially identical to me.

I first saw this particular dual-path design in the two linked schematics, later I found it also in the frontend of my Hantek 6xx4, and now also in the Hantek 2000. Seems to be much favored, for different models, and even by different manufacturers. Even with identical, or very similar component values. Just wondering who derived from whom...

[David Hess]

The designs looks essentially identical to me.

I first saw this particular dual-path design in the two linked schematics, later I found it also in the frontend of my Hantek 6xx4, and now also in the Hantek 2000. Seems to be much favored, for different models, and even by different manufacturers. Even with identical, or very similar component values. Just wondering who derived from whom...

This design goes back a long ways.  Kenwood used it in their analog oscilloscopes.

I have never done a detailed performance analysis comparing the two different designs but I suspect the major difference is that this one allows the AC coupling to be implemented with a low voltage capacitor and switch.