Ok, you had me with this one. 15 cents for a 2SK170, that's a really good price.
You have a unique opportunity to share with your fellow DIYers something valuable: your source/dealer, I am sure everybody would appreciate buying their 2SK170 at this price. If it's in quantity, I'm also sure somebody will organize a group buy.
For the rest of the technical details/reasoning above, I suppose you need to learn and grow yourself.
I'm not sure what's gm having to do in a JFET constant current source, I thought it's about Idss
Anyway, then your Norton source is going to be noisy as hell, intrinsically much noisier than LM329. You can decouple/filter, but so can you the LM329. OTOH, LM329 will have under 1ohm dynamic impedance (current independent), while your Norton thing will have probably the output impedance in the kOhm range. And your tempco will be as worse as the JFET (that is, very poor). Not to mention the precision, as good as the JFET Idss, like 6-30mA. And how about your "no compromise" approach?
Game over, I've already mentioned why arguing with you is pointless.
There's a nice book by Linden T. Harrison "Current sources and voltage references." I quote:
"In summary, the lower the drain current and the lower the output conductance, then the tighter the regulation will be. The closer the source resistor load-line intersects the zero TC point, the less the current source will drift over the intended temperature range, and the more stable and reliable your circuit will be."
"In summary, the lower the drain current and the lower the output conductance, then the tighter the regulation will be. The closer the source resistor load-line intersects the zero TC point, the less the current source will drift over the intended temperature range, and the more stable and reliable your circuit will be."
You are attempting to teach here history rather then audio design. Audio history is very interesting, though.
LM329 has a temperature coefficient of 0.01%/°C and 0.002% long term stability. Beat that with a JFET current source.
Relying on the J201 zero tempco point is an error. There's no spec and no guarantee for that (I'm not even sure if there is any, for this particular model). To stay into the field of audio, there's an old AES paper about, which, among others, shows how difficult (or even impossible) is to use the zero Tc point in JFETs:
FET Temperature Characteristics
Author: Sinclair, John C.
Affiliation: Zenith Radio Corporation, Chicago, IL
AES Convention:34 (April 1968) Paper Number: 557
Last edited:
Well, that is good, IF I needed it. Usually, the analog circuits that I make are not very dependent on the precise working voltage. They will work happily between several volts of DC power supply change. Simplicity, low noise, and no overshoot, are my main goals. Has been for 30 years. Seems to be important, still.
I'm not taking sides at all; it would be foolish of me, since I lack your experience, guys. There's probably no chance for any jfet solution to compete with a buried zener as far as TC is concerned. But I wonder how much is the subsonic variation of a current source based on a jfet. I'm tempted to say that I prefer a quieter current source that is sensitive to temperature changes (subsonic frequencies) than a noisier one with exceptional TC.
A LM329 reference is:
- Simpler than a JFET Norton equivalent reference. Just the LM329 and a resistor.
- Has lower noise and automatically free better performance (output impedance, thermal stability, long term stability).
- Has of course no overshoot.
The only problem that LM329 has is related to it's availability (or lack, thereof) 40 years ago.
I don't know what "subsonic variation of a current source" is.
A post-regulator based on a LM329 reference can be set to have overall an equal or lower noise than a post regulator with a J201 Norton equivalent reference. Think about the under 1 ohm output impedance and how this can help to remove the residual noise. Compare to the kOhm output impedance for the J201 Norton.
Seriously, there is one difficulty with the LM329, it's a 6.9V fixed reference. A JFET Norton equivalent can be adjusted to whatever to value. With a noise penalty, of course.
Oops, my bad English here. What I meant is unstable voltage with a variation below any audible frequency, which is the only kind of variation that could happen due to environment temperature variation. If the temperature of the jfet varies at higher frequencies then we've got bigger problems than any lm329 can fix.
Do you mean that the noise can be lowered with an RC filter?
This is all next to useless. First, I am working below 10nV/rt Hz over most of the audio bandwidth. Second, the j203's that I use cost me 15 cents in quantity. Three, to do the same thing as I am doing, takes 3 LM329's in series, that gives increased noise and higher cost. Forth, both designs take at least one large cap to be really quiet. The LM329 takes an added resistor as well, and gives that same impedance at DC as the Norton equivalent source. Work it out, those who can.
No. You need to read the datasheet.
Otherwise, LM329 is obviously not for you. It's for those guys that design regulators with 1milliohm output impedance up to 10MHz, no (zero %) overshoot, 20nV/rtHz noise, feeding a MC gain stage having at least 60dB PSRR.
Definitely not the case with your 30 years old Vendetta circuit.
- Status
- Not open for further replies.