Spin-off topic - Constant Current Sources

[TomWaterman]

Another thought, could make small little L.E.D hats for the audiofiles to use over their CCS lights!!

Here comes my fortune...who wants in? :razz:
-Tom

 

[bcarso]

I know someone who mentions that light sensitivity thing whenever the LED bias stuff comes up.

It is a VERY tiny effect. It is real, because LEDs can be used as photodiodes. But they are lousy for this purpose, and the d.c. and a.c. photocurrent generated under direct fluorescent illumination is VERY small compared to a milliampere. With incandescents the a.c. component is even smaller since the filament time constant smooths the 100 or 120 Hz variation. Regular junction diodes in glass packages are much better photodiodes, still not very good though. Of course one can use diode-connected transistors like in per anders circuit with opaque cases.

There was a story I heard from Rich May about when he worked at a broadcast studio. The boss would test the engineers by having them find problems he created in pieces of equipment. One case was a small but audible hum out of some rack mounted piece. It was generated by a cover left off of a neon pilot lamp that allowed light to shine on a diode in the circuit.

 

[TomWaterman]

Nice story. That boss sounds like a hardcore nut! Good at keeping his crew on their toe's I expect.

Did they find it?

-Tom

 

[pstamler]

[quote author="TomWaterman"]Paul, how does the load calculation change when the opamp is inverting? The phones amp is non-inv but the mono summing system is virtual earth and inverted amps are used.[/quote]

Instead of Rinput + Rfeedback, that term is simply Rfeedback. Again, compute that in parallel with the other loads, whatever they may be.

Oh, and I usually use MPS-A06's for current source duty. They seem to work fine, and they're cheap.

Peace,
Paul

 

[TomWaterman]

Thanks Paul,

I'll try and get some schemos up when I'm done.

-Tom

 

[featherpillow]

OMG, what a great thread, I totally missed it because I've been working on that buffer circuit.
(Do I sound like a sorority girl?)

I'm convinced--I'm abandoning my JFET cascode CCS... When I have time (next project), I'll prototype some CCS stuff using the LED/NPN example you've provided. Thanks for the great information!

 

[TomWaterman]

Haha,

I actually think I managed to get my head around the BJT/LED cascode...can't wait to try the headphone amp out with these onboard.

I also read somewhere else that if you take the resistor from the anode/base junction, (the one which goes to ground or +V) and divide into two halves (two equal resistors) and take a 100uF cap from the mid-point to ground, the PSRR and noise of the CCS can improve 6dB or so.

-Tom

 

[bcarso]

If it is noise on the power rail, with the initial setup the best you can do is make that resistor as high Z as possible (another current source in other words) so that the LED impedance dominates the emitter resistor-to-base voltage.

However, if you split the resistor to the LED as you say, you can have that cap between the split point and the rail (not ground or the opposite rail), and get effectively a low-noise source for the LED. The alternative would be a much larger cap directly across the LED, which would also reduce any LED noise.

 

[TomWaterman]

[quote author="bcarso"] The alternative would be a much larger cap directly across the LED, which would also reduce any LED noise.[/quote]

Yeah I heard about that too. Worth doing in your opinion?

-Tom

 

[bcarso]

No ;-). Not to damp your excursions, but you are going to be SO dominated by the amp itself w.r.t. noise that most of this is academic.

However---for the future, it is nice to know this stuff when a CCS is used where its noise, high output Z, stability with temperature and large signals, etc. is really critical! And there are plenty of those situations that arise.

 

[kiira]

[quote author="featherpillow"]OMG, what a great thread, I totally missed it because I've been working on that buffer circuit.
(Do I sound like a sorority girl?)
[/quote]

You're getting there but it would actually be

OMG! Like what a great thread, I tottaly missed it because I've been working on the um buffer thingy.

good to know I I can answer some questions around here :roll:

Kiira

 

[CJ]

Sorry to dredge this "I source" thread up from the depths, but I have a question, and I know you have cell phones.

My question is:

1) Why is this considered a low impedance current source
(thats what the patent blurb says, at least)

and

2) Why does it matter?

It feeds a gain cell, by the way. The I source is the box at the top.
Thanks!
cj

 

[bcarso]

Actually I don't have a cell phone.

I think it is a misnomer, and to some extent an oxymoron---it's a current mirror with a trim adjust, but the impedance at the right-hand collector is not particularly low.

It was probably written by a patent attorney and the inventor got tired of correcting the errors after a while---I know I went through this myself.

 

[CJ]

Here is another current source circuit that I don't understand. Supposedly this one supplies equal currents out of both of those left hand wires, until one gets unbalanced. If one wire draws more current, the other wire draws less to keep the sum of the currents equal. Right?

 

[bcarso]

Without any current in the lower left-hand wire the collector of the transistor acts like a constant-current sink (Iout about = (Vz - Vbe)/R80 ). The collector current is only slightly dependent on the collector votage as long as Vce is more than about 0.5V or so.

The impedance at the emitter is ~low, about 1/gm, ~25-30 ohms for a mA of collector current. So it is certainly not a "current source" at that node, in the sense of being high-Z.

If however you think of it as being an input for a current, any such current that flows adds or subtracts from the current at the collector. As long as you don't pull away all of the current of R80 (positive current into the node) the collector output is the difference between Io and the inputted current. If the input current is negative it adds to the magnitude of the collector current. That is, this is a common-base unity-gain current amp: current in, current out.

As far as the circuit's function, I don't know how accurate the patent description is without seeing more of the circuit and the context.

 

[CJ]

OK, here are some patents on the early VCA's. It is fun to read them in order, to see how things developed.

General link:
(scroll down a bit)
(good to print for easy reading)

http://www.audioannals.com/dbx.htm

And the circuit I am concerned with in particular, Adam's DBX 2001 VCA:

http://www.audioannals.com/4331931-Adams.djvu

And thanks for the explanations! These patent notes don't read too well, at least with my brain.

cj

 

[bcarso]

What is the djvu extension about? I can't view it as such.

Patent language is often pretty opaque stuff, a garbled translation of what the inventor tells the patent attorney. I got quite frustrated working with one guy as I had to keep rewriting his stuff. It turned out not to matter much anyway because H*rman abandoned them eventually in a effort to reduce legal expenses :evil:

 

[CJ]

I was worried about that browser thing.
It is a really cool system. Free download link below. Files are much smaller than adobe:

http://www.djvuzone.org/download/index.html

If you don't want the hassle, I can convert that stuff.