Power Supply Tuning

[dsavitsk]

I built this NOS-DOC and used 8 chips in parallel. Initially, I hooked it up to a Velleman 1823 PS I had sitting around.* The sound was very sweet (which I like) but the dynamics and bass were lacking, and there was a bit of grain in the treble. Also, the 1A transformer was getting pretty hot.

As a replacement, I built one of Tangent's Tread supplies using a larger transformer as I thought that might be cutting into the dynamics. I built it out of parts I had around, so I used a 2200uF smoothing cap rather than the 1000uF in the schematic. Also, I replaced C7 with a 100uF electrolytic, and c8 with a 100uF Cerafine bypassed by a 0.22uF BC cap. Most of the sweetness is now gone from my DAC. It has deep, full punchy, though somewhat sloppy bass, and vocals are set way behind the speakers -- sounding like they are in a tunnel. I did manage to get rid of the grain, however.

So, does anyone have any tricks they would suggest to get the sweetness back while keeping some (but not all) of the bass? I assume I need to reduce the caps, but by how much? Shoud I get rid of the electrolytic (the cerafine) completely? I have some high quality 1uF Kimber Kaps I could use. I also have some 4.7uF Aeons, and 5.1 uF Solens.

-d



* while the picture does not show it, the Velleman uses a 2200uF smoothing cap, an 8uF electrolytic, a 1uF and a 0.1uF Wima parallel to the rails, and an lm317 for regulation.

[SY]

I'd strongly suggest you look at Errol Dietz's paper on the effects of capacitance and load on the 317 family of regulators. It was reprinted in Bob Pease's "Troubleshooting Analog Circuits," a book you ought to have. Dietz shows the effects of bypassing and regulator load on source impedance and noise spectra of the 317. And do read Pease's comments in the text. The book will cost you less than a fancy electrolytic cap.

The first conclusion that you'll come to is that the 0.22 bypass cap may not be such a good idea. The second conclusion you'll come to is that there's merit in preloading the regulator; it wastes power and generates some heat, but the regulator will perform much better.

[dsavitsk]

Originally posted by SY
Bob Pease's "Troubleshooting Analog Circuits," a book you ought to have.

Thanks for the recommendation. It has been added to my Amazon cart.

Originally posted by SY
The first conclusion that you'll come to is that the 0.22 bypass cap may not be such a good idea.

In the meantime, before the book arrives, any thoughts? Looking at th Welborne PS1, he has essentially the same design but with a 0.01uF bypass cap. Perhaps I'll try that, unless this is a bad idea too.

Originally posted by SY
The second conclusion you'll come to is that there's merit in preloading the regulator; it wastes power and generates some heat, but the regulator will perform much better.

Just a bleeder resistor before the regulator? Before or after the smoothing cap? The PS is running at about 11V, so would somethig in the range of 2K be reasonable here?

-d

[mzzj]

Originally posted by dsavitsk


Just a bleeder resistor before the regulator? Before or after the smoothing cap? The PS is running at about 11V, so would somethig in the range of 2K be reasonable here?

-d

Or after the regulator?

[SY]

After.

Dietz and Pease beat up on their own company's data sheets for not noting that the output inductance changes with load current. At higher currents, it's lower. At lower currents, it's not only higher, but has a much higher slope. That means that the output inductance will be more strongly modulated by the load.

For the 317, 100mA of preload gets you onto the sweet spot.

[dsavitsk]

Originally posted by mzzj

Or after the regulator?

I have a bleeder (and an LED) after the regulator. I always do this to drain the caps and so I know when it's on, but if there are sonic reaons for it too, all the better.

Without burn-in, replacing the 0.22uF cap with a 0.1uF cap has made a big improvement. A little smaller still might be better.

-d

[dsavitsk]

Originally posted by SY
For the 317, 100mA of preload gets you onto the sweet spot.

So my bleeder is probably not enough. More like 100-120R?

[SY]

Depends on your output voltage. Ohm's Law.

First, there's the current of the bias string that sets the output voltage. That will be on the order of 10mA with the values shown on your schematic. Let's ignore it, and it will provide a bit of margin.

OK, the preload resistor. Let's say you're running 12V output. R = V/I, or 120 ohms. Power = IV, so the resistor will dissipate 1.2W. I'd use at least a 2 watter, better yet a 3W.

To calculate the extra dissipation in the regulator, you need to specify the input voltage. The difference between input and output time 0.1A is the extra dissipation in watts of the regulator caused by the preload.

[dsavitsk]

I haven't had time to try the bleeder yet, but I did a bit of experimenting with the caps.

100uF electrolytic cerafine bypassed by 0.22uF film: As noted before, the sound was like it was in a tunnel -- terrible.

100uF electrolytic cerafine bypassed by 0.1uF film: quite a bit better, out of the tunel, but still too bassey and fat with recessed mids and highs.

100uF electrolytic cerafine bypassed by 0.01uF polyester: better still, almost right, but still a little sloppy on the bottom.

22uF Panasonic FM electrolytic bypassed by 0.01uF polyester: This is just about right. Everything is sweet with just a touch of hardness insome mids (but this is very recording dependent, and may be in the recording itself), the bass is full, dynamics are punchy. On some recordings, however, it can sound a little undynamic. I am toying with adding a 4.7uF blackgate I have around, perhaps with a switch to include it with recordings that need just a touch more. Either that or a slightly larger panasonic. Overall, however, I am happy enough to live with this for a while.

I am truly amazed how much of a difference this single part made. In fact, it made more of a difference than any output cap or opamp or any of the usual suspects for upgrades that I have heard. This may be a quirk of this circuit, but the differences were night and day.

-d