Anything horribly wrong with this (LTP w/ LM317 as CCS)

[rongon]

I read about this in an AudioXpress article (I think that's where I saw it) by Walt Jung about current sinks/sources. Mentioned LM317 as being pretty good, but limited by decreasing effectiveness with rising frequency. OK, so it's not the very best CCS, but it sure is easy to install.



My questions:

There is no negative supply, but there is 7V between the 5687 cathodes and ground. My understanding is that the LM317 needs at least 2.5 to 3 volts across it, which the above has. So it should work pretty well, right?

Would a negative supply of about -20V improve the performance of the LM317 in this case, or would it be about the same?

I know that there are discrete FET-based CCS circuits that will work better, but this was just soooo easy. Had to try it.

Thanks.

[SY]

But the discrete MOSFET CCS has only two transistors and three resistors, with performance many orders of magnitude better. The 317 is a very, very mediocre CCS, and at high frequencies, it's downright awful.

In any case, unless you have a need to keep the plates at the same DC voltage, you can lose the trimmer. At that value, it's not doing very much anyway.

[sycorax]

I tried this with a russian 6N6 with 22kOhm plate resistors and 75Ohm in the CCS. The output is within 1 dB up to 60kHz.
The right triode has a little bit more distortion but when you use it for example to drive a EL34 or a KT88 this distortion is lower than 0.1%

[Yvesm]

Originally Posted by SY

But the discrete MOSFET CCS has only two transistors and three resistors, with performance many orders of magnitude better. The 317 is a very, very mediocre CCS, and at high frequencies, it's downright awful.

I was told that all 317 are not equals and some use different mask/design ! ?

In any case, unless you have a need to keep the plates at the same DC voltage, you can lose the trimmer. At that value, it's not doing very much anyway.

Agreed 100%, don't care about static plate voltage, this does not affect the symmetry of the output signals, no more than tube mismatch.
The weakest tube just limits the output swing, the other follows !
The only important thing is to use equal plate load, don't forget to include the grid leak resistor of the next stage.

Yves.

[rongon]

Thanks for the replies.

I'm going to try the FET CCS soon. I did slap the 317 in there (it's actually in the tail of a 6N30P pair DC-coupled to a pair of 2A3's), and thought I noticed an improvement. I just wanted validation, I guess. Now the next thing to do is to improve it some more by using a *real* CCS down there.

I'd be glad to lose the pot if I don't need it. The DC voltages can vary a little without causing problems.

I haven't been able to find a 'cookbook' step-by-step for designing a proper FET CCS. I think I have a schematic saved, but the last time I looked at it, I got lost trying to find the correct value for the current-setting resistor. Is there a site that goes step by step how to find correct values?

[SY]

It's easy. Do a MOSFET cascode like the one in my phono preamp article. Adjust the source resistor to get the correct current by driving a fixed resistor from the CCS and measuring the voltage across it.

[rongon]

Thanks SY. I'll have a look at your phono pre article...

Opinions on IXP10M45S? As in this?:
Current Source

[rongon]

Hi again...

On page 10, "Design Integration and Details", there's the schematic for the audio circuit. In the cathode follower (output buffer), there's a CCS of two DN2540 in the cathode circuit.

Is R18 the current set resistor (the one whose value would be changed to vary the current output)?

In the first (input) stage, there's a 1k ohm resistor (R3) placed below the CCS. (R5 would be the current set resistor, correct?) This is part of the load for the current source, correct?

If yes, then one could concoct a test circuit by using just that part of the circuit. Connect R3 to ground, and feed the "top" of the CCS with a smaller voltage, say +15V. Adjust the value of R5 for the desired current.

Correct?

Thanks again...

[SY]

R18 is indeed the current-setting resistor. The other two are gate-stoppers. The nice thing about this sort of CCS (besides simplicity and high performance) is that it's a two terminal device so can be used as source or sink.

Yes, you can adjust it exactly that way- a 1k dummy load has the advantage of having 1v/mA of drop, so (for example) a 5 volt drop corresponds to 5 mA.

And the CCS will want at least 5-10 volts across it. That means you'll probably want to return it to a negative rail. Fortunately, because of the stunningly good power supply rejection that a high performance CCS brings to the table, the supply can be very simple and its exact voltage doesn't matter.

[Sch3mat1c]

FYI, if you want a high compliance CCS with a low voltage budget (i.e., avoiding a negative supply), a bipolar current mirror will do well. Without emitter resistors, it is possible to have constant current down to 0.1V or so. However, the Early effect is exaggerated, so the current will be noticably higher at, say, 10V (it's not a very constant current). For any higher accuracy, you have to sacrifice more voltage drop, e.g. 0.7V for a Wilson mirror, or a "ring of two" source.

An op-amp servoed circuit like this

can work to arbitrarily low voltages, but requires an additional rail (maybe +10V for the op-amp, though it's low current), and the arbitrariness of the saturation voltage depends on the op-amp gain and offset. Bandwidth is limited by op-amp and Cgs, which can be fairly high; above that, the output appears largely capacitive due to miller effect.

FYI, LM317 input is an open collector, so it's not a terrible CCS. LDOs are much worse (emitter/source to Vin). What are the actual numbers on its input impedance at HF, does someone have a plot? I forget if the datasheet does, I know I've seen PSRR before but that's not the same as dynamic impedance.

Tim