DIYHiFi.org

Fustrated with common base emmiter non-nfb BJT I/v stages. I mean the best you are going to do dumping a i-out into a BJT is about -80db 2H, any performance above this is from subsequent cancelation tricks that generally start sounding like opamps.

Now passive I/V and prefiltering has an advantage in the the impedance seen by the DAC is never going to get higher than the I/V resistor, no matter what frequency, at 22.5792 MHZ the passive I/V is still the same input impedance because as all the active devices following it go up in impedance, it doesn't matter 'cause thats in parallel, plus the 2nd order filtering (caps to ground) probably help (no inductors!.)

This is the best PCM1704K I can come up with, and it was just out reading this forum for years and of shear fustration with oscillation, opamps, you name other stuff I tried had issues.

All it is : PCM1704x2 (parallel) - 30 ohms passive I/V - cap to ground - Jfet gain stage with NFB and another cap to ground - and a diamond buffer output. This gives a pretty high voltage compliance, but for some reason it sounds good. Separate Mofset shunt power regs for everything. Shitty layout/groundplane. The gain stage is fed +-30V, higher rails and more gain would allow a smaller I/V but I don't know now if it is worth pursuing.

And the distortion is basically negligable, in steep contrast to the zero FB BJT attempts I've made. It just seems to fly in the face of everything we know about the PCM1704. I'm using a Legato-spdif-DIR9001-PDM100 (no reclocking).

My postulation is the big I/V resistor and the high voltage compliance just shuts down the 1704's BPO and getting that crappy servo out of the picture is good for SQ, I can't explain it but maybe someone can explain why this sounds good and measures good:

Hi,



Funny, years ago I arrived for parallel 1704 at the following:

33R & 47nF IV

"Gomes Type" Tube stage (this is very linear if executed right without NFB).



Don't ask me, but yes, contrary to what I have read everywhere I found my results also rather good...

Ciao T

As to why it measures good, I'd guess that your 30 ohm passive I/V is provoking relatively minimal distortion with the PCM1704 current output, while your FET gain stage utilizing NFB is, naturally enough, also contributing low distortion.

As to why it sounds good to you, that is a matter of personal perception, so who can exactly say? Perhaps, it is simply a matter of having implemented a simple passive I/V and L.P. filter stage, providing a band-limited input signal to a simple, (I assume) modest feedback, discrete FET gain stage. So, while the 1704's BPO servo should still be attempting to "correct" the signal voltage appearing at the current output pin, this effect should be greatly minimized for frequencies within the audio band by the servo's L.P. corner frequency.

Just my two cents worth.

Hi,



Just two data points for those who are interested. A friend send me (strictly sub rosa - so don't ask for a copy, not even in private e-mail) a discrete active (feedback) I/V stage that is used with PCM1704 in a commercial product.

According to him it not sounds excellent and measures well. It has an input impedance off 100 Ohm!!!

So we may conclude that as much as 100 Ohm resistive load does not cause trouble for the PCM1704 as such.

As for the DC Servo and BPO Servo in the PCM1704, their dominant pole capacitors are external and hence may be changed to slow these servos down. I believe this came from here, but I'm not sure, I have saved on my HDD this picture as to how the 1704 internally works:



Ciao T

Hi,

sorry to see Your disappointment about the non-nfb Grounded-Base-stages, but Your solution vertainly will work well.

My GB-stages work without fancy cancellation tricks apart from a compensation to keep input impedance constant t into the MHzs.
With Buffer output stage included in measurements they now settle at -98dB H2 at 0dBfs supplied by low voltage supplies of +-12V.

Because of parasitic effects of layout and parts the impedance will hardly be the same at HF as at LF. Passive circuitry doesn´t automatically mean constant.
But Yes, on the other hand active circuitry certainly requires counter measures to keep the input impedance constant over a wide enough frequency range.
This is imho one of the basic fads of most OpAmp circuitry.

jauu
Calvin

Thanks for the info, I would like to try a non-nfb common base bjt i/v that performed like that. I know sharing is taboo these days but if you can shoot me a PM I would like to try to compare.


Seriously though as much as I like the sound I know that slowing down the servos is like cheating (its not like the thing is 24 bit anyway), I am still NFB phobic and although this jfet gain only has about 27db nfb, I would love to here a no nfb stage that gives -98 2H at 0dbs. I think the mosfet pass D1 gets close to thought but at the cost of expensive power.

Hi,

I was talking about the simple Jocko-Type grounded Base circuit, followed by a JFET-Source-follower, with JFET-CCS-load.
The Vbe of the input transistor is cancelled with a second bipolar, of same type, connected as diode (actually I use matched duals here).
A small cap from the bases to gnd modifies input impedance. Nothing special with regard to circuit schematics.
For THD finding good transistors and the choice of working points and voltages is the key.

jauu
Calvin