Discrete AD797 up and running

[nelsonvandal]

Not really an AD797 since it's a JFET input amp. Seems stable, sounds good. Sounds close to the original AD797. Maybe a little cleaner than my discrete AD829 clone, but I can't be sure. I'm going to give it some more time.

I use Sanyo SA1016 and SC2362, not the Toshibas in the schematic. There are no good models for them. The Sanyos sound more neutral and just as detailed. The perfect small signal transistor.

It's battery powered, 8 x AAA, and pocket sized.

I think it's the floating current mirror that makes it sound different from "AD829", more so than the "magic" transistor or the decompansating capacitor.

[aparatusonitus]

Originally posted by nelsonvandal
I think it's the floating current mirror that makes it sound different from "AD829", more so than the "magic" transistor or the decompansating capacitor.

Hi Nelson,

If you are referring "magic transistor" to T1, floating current mirror couldn't exist without one. Have you consider cascading J1/2 with something like 2SK246 alla Borbely? Do you have the same DC operating conditions at input and output of current mirror, could you post few turns of transient analyses from input/output from CM and opamp output?

[jcarr]

Nelson: you may want to try moving the base of T1 from the gain node to the output (or a derivative). This will reduce loading on the gain node, and probably give lower distortion as a result. But whether you move the base of T1 or leave it where it is presently at, you probably want to add a bit of level-shifting to give T4 room to operate.

Personally, I would also try feeding C7 from the collector junction of T4/T5 (rather than from the output), and create a local nested feedback loop. IME, stability and RFI immunity frequently go up when you do this.

If you must use resistors RE and RF instead of active current sources, you may want to exchange D1/D2 for a precision voltage reference. The Vf of a normal diode is too easily affected by a variety of factors, and this may result in unwanted current fluctuation (which will degrade the overall performance of the circuit).

You can also improve the distortion of the output stage if you drop the two current sources and instead run a current mirror on either voltage rail, so that the input of the current mirror is fed from the collector of the output device and the output of the current mirror forms the active load for the junction of the drive transistor's emittor and the output transistor's base. To keep the current mirror going, add one of the active current sources that you removed back across the collector of T10 to the collector of T9.

No need for DC offset control?

regards and hth, jonathan carr

[nelsonvandal]

Thank you for your suggestions, I'll look in to them and run some simulations. I'm somewhat limited because of board layout and available space, but I'll take the experience and your input with me to forthcoming projects.

Could you please explain what level shifting is and how I could implement it. I'm not so clever.

What voltage reference could be used instead of the two diodes? In one amp I use a 4.7 V zener, and that amp sounds better, but then it can't be run effectivly on batteries because of the reduced voltage swing. I couldn't find a 1.2 V Zener.

The DC is in the need of some trimming by paralleling R1 or R2, but the offset is "only" 20 - 30 mA untrimmed. All transistors except the output pair are closely matched. I'm done using pot's for offset trimming because of high price, board space and noise (I don't think the ones I've bought at ELFA from BI Technologies are of good quality).

The amp runs perfectly well without the transistor T1, but sounds better with it. I did the mods from AD829 to AD797 in stages - 1. the floating CM 2. the decomp cap 3. transistor T1.

DC at collector of T5 is 38 mV and of T6 is 22 mV, almost exactly the same as simulated. Is that a problem?

Yes I've considered cascoding the input pair. I have a couple of SK246 for future experimenting. My favourite though is CFP input with JFET/bipolar. It does wonders to the sound, more than any other tweak. I suppose both CFP and cascoding à la Lavardin/peufeu is even better. My next amp will definately use CFP input.

[nelsonvandal]

Distortion at output. 1 V, 1 kHz, 300R load.

[nelsonvandal]

Distortion at CM.

[dimitri]

Hi Jonathan
Glad to see you

Still alive and active

Happy Holidays!

[nelsonvandal]

Distortion from the cloned AD829's output with the same biasing currents and 300R load at 1 kHz.

[nelsonvandal]

And from the CM of the "AD829".

[dimitri]

Are these disto graphs measured or simulated?