Higher values do tend to give lower low-frequency noise on the supply rails. Whether this noise reduction is audible depends on many factors, if you do hear it its like the recorded space (on acoustic recordings) has opened up wider and deeper.
The SAA7220 is a very noisy brute, I'd use extreme HF decoupling with lots of paralleled ceramics to tame it.
Last edited:
You guys might also want to check out this:
http://www.linearaudio.nl/linearaudio.nl/images/pdf/bypass_capacitors_part1.pdf
There's actually 6 parts to this - they can all be downloaded from Home at My Library|Miscellaneous.
Jan
http://www.linearaudio.nl/linearaudio.nl/images/pdf/bypass_capacitors_part1.pdf
There's actually 6 parts to this - they can all be downloaded from Home at My Library|Miscellaneous.
Jan
Thanks for the clarification. If I put LM317 regulators at each of the power pins, do you think more than 100uF will still make a difference?
The problem with LM317s being used to power opamps is their output impedance turns out to be surprisingly high at low (opamp Iq, typically below 10mA) output currents. So yes, increasing the capacitance does make a difference when the regulator presents a significant inductive impedance to the opamp supply pin.
A quick and cursory look at the LM317 DS shows a promisingly low Zout in the audio band but the small print tells us that the plot shown is at 500mA - 100X more current than the typical opamp draws.
What Andrew says - use a cap on Cadj but make it 10X bigger than the DS suggests to get lower output noise.
LM317 scored badly in the listening test. If you're going to use an OEM regulator, try the LT1963A.
OK, if you're going to use the LM317 anyway, using a much higher value of the resistor from OUT to ADJ -- see Part 1 of WJ's 1995 articles in which he uses 1K/10K.
If you use too big a cap from ADJ to GND to knock down noise, you may get inferior transient response.
Jans boards are available on the DIYAUDIO Store. Per Anders boards are available through his website.
In general LM317/337 perform great at highish output currents.
If the DAC is TDA1541 all the supplies take <40mA typically which means LM3x7 won't perform too well, especially at HF. I'd use assisted shunts (TL431 plus bipolar transistor) as shown in the TL431 DS. The additional transistor improves the output impedance and helps with dissipation.
For the SAA7220 an LM317 would be in its element as the draw is around 200mA from memory. I'd use LC filtering between the LM and the SAA.
One thing I just can't wrap my head around: I know that low ESR cap should absolutely be avoided at the output of the LM317 (and other regulators). Does this mean I should remove the ceramic bypass cap at the power pin of the IC?
edit: I think I figured it out. This post is interesting on the subject. If I get it right, either don't use low-ESR caps after LM317 or put a resistor in series with each supply pin.
edit: I think I figured it out. This post is interesting on the subject. If I get it right, either don't use low-ESR caps after LM317 or put a resistor in series with each supply pin.
Last edited:
Its a very good question - ceramic caps are indeed very low ESR. What I normally do is add some series inductance between the reg and the bypass caps (a 220uH bobbin wound choke with about 0.2ohm DCR). Also don't use a ceramic cap without a 'lytic at least 20X bigger across it for damping when there's appreciable series inductance in the circuit. Using a resistor will give you poor load regulation unless the R value is very small.
In various Philips CD players I've had yes, there were lots of supply resistors (22ohms rings a bell) between the regs and the opamps. I think they're jolly bad news for sound quality given that opamps don't operate in classA and hence create lots of load-induced supply noise. Increasing the supply impedance is not wise if satisfying sound is desired.
But they do damp ringing -- here's the rub -- the ringing goes both ways along the supply rails and can excite other circuitry into unstable condition. Reverse PSRR is what Omicron LAbs calls it. (Personally, I haven't experienced it.) Perhaps its an issue if you have mixed analog and digital.
A really well designed regulator will have impedance from low milliOhms into the microOhms -- see WJ's 1995 papers.
I'd guess we're not looking at the same ones. This is the kind of thing I meant - 20pcs Radial Inductor 220uH 221 8mm x 10mm +/ 10%-in Inductors from Electrical Equipment & Supplies on Aliexpress.com | Alibaba Group
They're 20 for about $2 i.e. 10cents each.
You can use a resistor in place of the inductor, sure. I'd go for under 2.2ohms though. Low supply impedance correlates with best SQ in my experience.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.