Discrete Voltage Regulator

[Samuel Groner]

Unfortunately there aren't any IC opamps available (as far as I know) which offer 100 dB PSRR within the audio frequency range. And note that you'll need to subtract the noise gain of the opamp configuration for the effective output-referred PSRR. So for a typical 60 dB PSRR at 10 kHz we end up with 20 dB for a 40 dB noise gain configuration.

Samuel

 

[JohnRoberts]

Indeed, a good reminder that opamps are 5 terminal devices and most errors referenced to input not output.

That said there are easier ways to mitigate just PS noise or impedance. A couple decades ago for a tweaky phono pre, I rolled a low noise Vreg from a TL072 and pair of discrete pass transistors. Got the job done nicely and I didn't kill many brain cells on the design.

All design is a simultaneous weighting of multiple factors. If PS noise or impedance becomes significant in the margin adjust as needed.

Keep up the good fight...

JR

 

[Samuel Groner]

I updated the linked schematic to include the additional protection resistor; I also changed my mind about fold-back current limiting.

Samuel

 

[Samuel Groner]

Was able to build a simplified version (no current limiting and simple zener as reference); worked like a charm. Output noise is about 700 nVrms in a 22 kHz BW, line and load regulation unmeasurable with the currently available methods.

Samuel

 

[JohnRoberts]

:thumb:

 

[guavatone]

[quote author="Samuel Groner"]Was able to build a simplified version (no current limiting and simple zener as reference); worked like a charm. Output noise is about 700 nVrms in a 22 kHz BW, line and load regulation unmeasurable with the currently available methods.

Samuel[/quote]

Interesting. How does that compare to the noise figure of a 317 or LT1083 in the same ckt?

 

[Samuel Groner]

If I'm not mistaken both tend to be around 50 uVrms with feedback capacitor (much more without)--almost 40 dB worse. 700 nVrms is about -120 dBu; some dBs better than the input-referred noise of the AP System One and about as good as a typical mic preamp at medium gain. Not that I believe that the sonic improvement from using my regulator is drastic for typical circuits but it's nice to have it at hand for certain situations.

Samuel

 

[Samuel Groner]

Tested the short-circuit protection today--worked as expected, screwdriver test passed.

Samuel

 

[hifizen]

Nice circuit, Samuel. I might make use of this, or something very similar.

I'm curious why you ended up with a triple Darlington output instead of the two-transistor Darlington you started with. Did you need the extra diode drop, or was it a matter of needing greater current gain?

 

[Samuel Groner]

It increases the transconductance of the second stage, which translates in a proportional reduction of low-frequency output impedance. Don't remember the exact figure of improvement, but it was substantial (one or two orders of magnitude).

Samuel

 

[recnsci]

What's the catch with D108/109?

 

[bcarso]

[quote author="Samuel Groner"]It increases the transconductance of the second stage...Samuel[/quote]

Not exactly. It increases the input impedance of that compound output stage, which reduces loading on the input stage, allowing it to have higher open-loop voltage gain, and hence the circuit to have higher overall loop gain.

Darlingtons or other compound same-polarity e-followers actually have lower gm than a single device (remembering that gm is amps per volt, delta Ic for delta Vbe). But we rarely drive grounded-emitter transistors with perfect voltage sources!

 

[Samuel Groner]

What's the catch with D108/D109?

Just a tiny gadget to reduce secondary current mirror errors (makes the current mirror transistors operate at roughly equal Vce). Will work fine if replaced with a short.

Not exactly.

Ah right, thanks for pointing out. I guess what I actually wanted to say is that it increases the transconductance of the entire amplifier (which is a transconductance amplifier).

Samuel

 

[bcarso]

Yes it certainly does that!

Don't feel too bad---I caught Norm Thagard on that same issue. He, by the way, is a most gracious gentleman.

 

[recnsci]

[quote author="Samuel Groner"]

What's the catch with D108/D109?

Just a tiny gadget to reduce secondary current mirror errors (makes the current mirror transistors operate at roughly equal Vce). Will work fine if replaced with a short.
[/quote]

It looked like that, but whats the point of reducing Early contribution
when you have unmatched and thermally uncoupled LTP?

cheerz
urosh

 

[Samuel Groner]

I don't remember whether it actually had an influence on ripple rejection or not. However from opamp design I know that using a similar improved mirror definitely enhances CMRR (and hence usually PSRR), even with a JFET frontend where mismatch is much more pronounced. In any case I think it ain't particularly reasonable to spend too much thinking about this--with the time it has cost us to discuss this we could have bought a thousand diodes. :grin:

Samuel

 

[olafmatt]

Samuel, I built a version of your regulator. Thought I was clever and used a LM329 (6.9V zener) for D105 and a 6.2V zener for D101. The thing works, but "sometimes" (tm) starts to oscillate (about 3mV of HF crap). Any idea what I should be looking for? All electrolytic caps are BC (Vishay) low ESR. Ah, and I replaced Q109 with a MJE2955.

Olaf

 

[Samuel Groner]

3 mV is not much--at which frequency? I'd try a base resistor for Q109 first, say 10-100 Ohm. The layout is fine, with good grounding arrangement?

Samuel

 

[olafmatt]

Can't really say at which frequency it oscillates. On my 20MHz scope I can't really get it to show up correctly. Could also be a more noise-like mix of a lot of frequencies.

I added 51 Ohm resistors in the base of Q109 and now it doesn't seem to oscillate anymore. However, with a load attached (drawing 50mA) the oscillation starts again and increases to about 50mV! (This also happens without the resistors, I just never tried it before.) I also tried with a relay as the load (to be sure it's not stuff coming from the circuit I was powering with it) and the same problem shows up.

Here is a picture of my layout. There is a solid ground trace dividing the PCB in the middle. One half is a mirror image of the regulator to get a negative supply voltage. It shows the same oscillation problem.

Olaf

 

[Samuel Groner]

If it's above 20 MHz then it's definitely some parasitic instability, and not nyquist instability of the feedback loop. Probably in the output devices, as the benefit from the addition of the base resistor showed. Unfortunately I don't have much idea what to try next. Perhaps adding a small series resistor (0.1 Ohm as a starter) for C104, or a substantially lower value for C104 (220 uF). Otherwise I'd try replacing the power transistor with something closer to the specified part.

Is the negative regulator a complementary implementation (i.e. with a NPN pass transistor)?

Samuel