> In the case of the simplified pre-preamp shown by John in post 67475, if a floating current source with sufficient voltage compliance is used in place of R2 and R8, except for certain circuit impedances that are not perfectly complementary, the noise in the current source nulls at the output!
Very clever. How could I miss that .......
Patrick
Very clever. How could I miss that .......
Patrick
That is an interesting collection. But they talk about output resistance without mentioning capacitance, which dominates in a hurry at audio frequencies. With control-electrode current recycling, this capacitance can be made very small out to fairly high frequencies.
Has anyone done a really comprehensive presentation in diyaudio about current sources, including noise and capacitances? If not, perhaps it is time. I know Walt Jung did a lot of work on the subject of such sources, and most of it is probably on his site.
OK, my question wasn't correctly formulated. What I wanted to ask is whether it is of paramount importance to do anything possible to reduce the 2nd to very low levels in John's folded casode topology or is there some threshold below which the further reduction is not important anymore. -60dB, -80dB?
I could work extra hard to make a quiet enough current source, but it does not come easy. Almost ANY forward gain in the active device will add extra noise. Resistive degeneration in mandatory, and you ultimately wind up with just a resistor, for the most part.
I will keep with the resistor, and as I did, 30 years ago, make an extremely quiet regulator, (not just a cap multiplier) to keep out external power supply noise. It still works today.
What we need is an extremely high Idss, low noise, and low Gm part. Any suggestions? '-)
I will keep with the resistor, and as I did, 30 years ago, make an extremely quiet regulator, (not just a cap multiplier) to keep out external power supply noise. It still works today.
What we need is an extremely high Idss, low noise, and low Gm part. Any suggestions? '-)
What would be helpful here is to understand the relations between a nonlinearity that produces some specified amount of harmonic distortion at a given level, and how much IM distortion that entails. Then we could use a pair of tones and determine how much difference and sum signals are produced, and where they are both in level and frequency.
One of the bothersome things relates to IM that falls on or close to a musical note that is "in the chord" as it were, and when it is jarring and potentially unmusical. Since the near-universal acceptance of equal temperament, there are few cases where the overtone series of harmonics of a given tone are precisely the same as the nearest equal-tempered note.
When one factors in the various ways that other material masks the perception of small distortion byproducts, it gets hairy. But since it is fairly easy to make distortion small, one may as well do it IMO. It doesn't save much money, for one, to let it go by.
> What we need is an extremely high Idss, low noise, and low Gm part.
J111.
We used it as cascode in a BF862 JFET Buffer.
http://www.diyaudio.com/forums/digi...ng-new-ess-vout-dac-es9022-7.html#post2867664
http://www.diyaudio.com/forums/digi...ng-new-ess-vout-dac-es9022-8.html#post3040691
Patrick
J111.
We used it as cascode in a BF862 JFET Buffer.
http://www.diyaudio.com/forums/digi...ng-new-ess-vout-dac-es9022-7.html#post2867664
http://www.diyaudio.com/forums/digi...ng-new-ess-vout-dac-es9022-8.html#post3040691
Patrick
Wayne C. kindly sent me a couple samples of a huge part from Interfet. I believe Pass Labs had to order quite a few of them to induce the manufacturer to make a batch, and they are no longer in the catalog. So I'm afraid, for the dedicated few, it's going to require "massive parallelism" to roll some composite device that is comparable.
I don't recall the doubtless-high gm spec, but I seem to recall the e sub n as less than 300pV/sq rt Hz, which would allow one to deduce the gm. I was tempted to saw the top of the metal case off to see how big the chip is! I wish I had downloaded the datasheet.
Afaik they did not find their way into a Pass Labs product. The motivation for them to begin with was large-area nuclear particle detector preamps, where the detector crystal and preamp front end would be cooled to probably about 100 Kelvin. This both enhances gm and thus lowers noise, but more importantly slashes gate leakage as long as you run at sufficiently low drain-source voltage. EDIT: Oh I see JC wanted low gm. Yes if that is the criterion the process 51 parts are o.k.
Well, I experimented this question of the linearity of feedback ratio in the audio bandwidth, long time ago and on a very subjective way (listening). I stay convinced that an amp sound better when open loop is flat at least up to 10KHz than the same amp with more open loop gain at low frequencies and lower FC.
And when slew rate allow *a lot* of margin.
Just my 'opinion'. It is like distortion numbers ;-)
Section on designs of current source design for low noise ---
https://books.google.com/books?id=0...noise and low voltage current sources&f=false
THx-RNMarsh
https://books.google.com/books?id=0...noise and low voltage current sources&f=false
THx-RNMarsh
Last edited:
I bought that book and it was very disappointing. Walt Jung was less than pleased and returned his copy!
Harrison, Current Sources and Voltage References: A Design Reference for Electronics Engineers
Softcover, ISBN 075067752X
Publisher: Newnes, 2005
But as I say, not recommended. Not as bad as another book though about switched-capacitor systems, which had NOTHING in it about noise at all!
- Status
- Not open for further replies.