Is there any benefit to adding r and c after the 317/337 pair or is that just an invitation for oscillation?
The regulators must be decoupled as per the data sheet recommendation to preserve stability.
Adding any 'R' after the regulator totally negates the whole point of having a regulator in the first place because now the output voltage is modulated by load current.
Also adding large amounts of 'C' after the regulator destroys its transient response to changing load currents.
Adding any 'R' after the regulator totally negates the whole point of having a regulator in the first place because now the output voltage is modulated by load current.
Also adding large amounts of 'C' after the regulator destroys its transient response to changing load currents.
So i had crcrc filter from the output of 317/337 into an opamp circuit, the filter values 10r and 1000uf. I have to admit despite the technical objection the circuit sounds better with the filter than without. The 317/337 regulator circuit by itself is very robust and stable, btw.
That's fine in itself for the listener (you in this case 🙂) but it only stands up as a statement if you could then reliably pick that change out in a controlled test where you have no knowledge which you are listening to.
This is something we see a lot when changes are made to a circuit, you want it to be better and often that is how it seems... and that is perfectly OK.
Oh, its been done before in commercial audio - Naim traditionally used R-C after the LM317s for ..reasons I explain here: likely made good sense 40yrs ago when conceived, esp since their generic circuits in -use where essentially constant-current, though low in PSRR - a combination not so much dependant on 'low-Z' supplies:
This page wot I Rote long ago:
acoustica.org.uk: modifying Naim circuits
This page wot I Rote long ago:
acoustica.org.uk: modifying Naim circuits
ETA:
The LM317 and 337 do NOT need any output capacitance for stability -read the datasheet; and it's true from experimentation.
The input side cap is more important than stated, measurement suggests that c.100uF is better than 10uF is better than 0.1uf stated minimum. There are really, really good reasons for that also - the Errol Dietz paper, originally, that demonstrates that output loads are reflected at c -20dB to the input side capacitance. If you have freedom to design that layout, the '0v' end of that input cap - decopuled from raw PSU by a dab of L or R - can then be returned to 0v near the load - and then good things happen, because you've closed Kirchoff's loop.
tl;dr:
Much more important that worrying about parts and values - is understanding the effects of the exact layout. The map (schematic) - is not the Territory. Where you select to terminate the '0v' ends of the various bypass caps you want to use really, really matters - the difference between 'brilliant' and '...got a bad case of hum I don't understand'. 80dB or more with poor vs good choices!
HTH 🙂
The LM317 and 337 do NOT need any output capacitance for stability -read the datasheet; and it's true from experimentation.
The input side cap is more important than stated, measurement suggests that c.100uF is better than 10uF is better than 0.1uf stated minimum. There are really, really good reasons for that also - the Errol Dietz paper, originally, that demonstrates that output loads are reflected at c -20dB to the input side capacitance. If you have freedom to design that layout, the '0v' end of that input cap - decopuled from raw PSU by a dab of L or R - can then be returned to 0v near the load - and then good things happen, because you've closed Kirchoff's loop.
tl;dr:
Much more important that worrying about parts and values - is understanding the effects of the exact layout. The map (schematic) - is not the Territory. Where you select to terminate the '0v' ends of the various bypass caps you want to use really, really matters - the difference between 'brilliant' and '...got a bad case of hum I don't understand'. 80dB or more with poor vs good choices!
HTH 🙂
Last edited:
There is also a minimum value for ESR in the output capacitor to ensure the regulators inductive output Z is at least critically damped in the resulting potential series LRC circuit. Use low ESR output cap and risk noise peaking. A subsequent RC filter could reduce this noise but best to deal with problem at source.
John
My understanding is the input cap is more vital when the regulator isn't close to their power supply caps. Output caps having low ESR sounds worse and if larger caps (than 0.1uF) on the input then a diode benefits from input to output.
My experience is thin though and interested in peoples experience
My experience is thin though and interested in peoples experience
Please realize I find it that interesting that I'll be up all night reading this excellent information. Just what I've been needing after spending all week building my first power supplies 🙂
Very much appreciated, thank you
Very much appreciated, thank you
Well thats a relief knowing that what im doing isnt completely wrong.
Come to think of it ive seen such rc filter arrangement in a composite amp setups whichs where i probably got the idea from.
Come to think of it ive seen such rc filter arrangement in a composite amp setups whichs where i probably got the idea from.
Some manufacturers LM317/LM337 are extremely noisy, lots of low frequency flicker noise.
I see some designs stuff a large say 1,500uF capacitor on the output of the vreg to mitigate this and it does cushion some of it. I can't see a RC useful.
Stability requirements for some LDO's need R added but not the 317/337 as they are not LDO's which have a PNP pass-transistor.
I see some designs stuff a large say 1,500uF capacitor on the output of the vreg to mitigate this and it does cushion some of it. I can't see a RC useful.
Stability requirements for some LDO's need R added but not the 317/337 as they are not LDO's which have a PNP pass-transistor.