[da_burl]

Guess I'll go ahead and ask a more or less theoretical question. I have built a sub-Gilmore, and I am using a regulated adjustable power supply. The one I am using now uses LM317/337's (was one of the relatives of PS **** posts!). Although I haven't done it yet, I also plan to build a not_a_PS1 clone, with LT1085/1033 regulators, and also try out some of ppl's PS tweaks (the FET CCS, and series resistors). But, my inquiring mind is wandering on subject.

The official Gilmore power supply uses pre-regulation, then a voltage tracking section.

So, I got to thinking. The purpose of regulation is to maintain the voltage constantly vs. current demands. If it is doing its job right, what are the advantages of also having tracking? Is it just to make both the + and - rails exactly the same? Or are the IC regulators not capable of maintaining the voltage 100% correct? Or is it operating in the "current domain" also?

Just wondered, I have seen the flood of recent Gilmore threads, and there are people using regular regulated supplies, dual mono supplies, and recently a variation of the tracking supply, with a different circuit than the original. I wonder if there is a concensus on whether it is true what Kevin G said, that he even wondered if anyone could hear the difference?

Just haven't posted any dumb questions lately, and I do need to get my post count above 100!!

TIA

da_burl

 

[jamont]

Using the LM317/LM337. a resistor network is used to fix the voltage, so V+ and V- will differ somewhat if the resistors are not exactly matched. Using an opamp should make V+ and V- track very closely, which the regulators alone don't necessarily do.

 

[Voodoochile]

Weird thing- the first PS1 knock-off I made was with fixed resistors, carefully matched, and the neg rail was about 150mV higher than the pos rail. I decided to swap both resistors from the pos side to the neg side and vice-versa. The voltage did not change more than maybe 2 millivolts. I am thinking that with a given set of resistors (assumed PERFECT matching

), that the neg reg is just a little different than the pos reg. I have seen this effect on three different 317/337 PSUs, and on two different PSUs using the Linear regs from the original PS1 design. Same in every case.

I think what I want to do is whip up a little stokes-erix type of PS, and socket the trimpots. I can then trim the PS to proper output, remove and measure the trimpot value. Not as a tool to make PSUs, but maybe to make an adjustment to the formula that will work right for the negative reg. Or, at least post what adjusted values might make the neg rail hit closer, but they may not exist, might have to parallel some odd values.

 

[jamont]

In my PS, V+ and V- differ by .02V (resolution of my meter), a little less than yours. I would expect this to vary somewhat with the particular devices used.

Not that I need any more power supplies, but I would like to try something like the one in this thread:

http://www4.head-fi.org/forums/showt...threadid=28063

He has a trimpot in the + side and an opamp in the - side, which looks like a nice way to do it. Of course, with a 15V supply, even 150 mV is only a 1% difference, I doubt that is audible in use.

 

[Voodoochile]

My difference was .015 volts. I have my terminology screwed up, that's 15 millivolts. Idiot! Either way, the trimpots are a nice touch, and they are only about 2.50 each, so it's an extra 4.00 perhaps in the power supply. I'm sure it isn't audible, but knowing it is different ruins my day for some foolish reason. Not a very good quality in me.

I like to be able to fiddle with different levels for testing purposes, but you all know that by now about me!

One thing I have noticed is that once set, the voltage maintains itself very tightly, even under stress. Having the amp in class A makes the job much easier for the PS, I'm sure. I note very little fluctuation, even three digits out.

 

[jamont]

So we have essentially the same difference in tracking. I bet that's a well-known limitation of the regulator chips. Any real EE types please jump in here...

 

[Voodoochile]

That's what I'm thinking.

I see the formula with the positive regs typically, so I imagine it is that the neg regs just run about that much higher. I'm sure you could get another resistor pair that would get closer...

 

[jamont]

I'm still considering a PCB design. I have in mind something that would include:

1) discrete diode bridge with bypass caps, trying to allow for the use of Schottky or whatever as well as 1N400X types

2) enough room for boutique caps

3) able to use the LT regulators as well as LM317/LM337

4) trimpot or resistor to set + output voltage

5) opamp tracking of - voltage

6) allow the use of Molex connectors as well as hard-wired connections for input and output

7) a design that could be made by ExpressPCB or by DIY means

AFAIK 3) is not a problem, aren't they pin-compatible? 5) places some constraints, probably Vmax not to exceed 18V, but that's fine for headphone amps. Any comments?

 

[Voodoochile]

#3 is true. Everything but #5 should be on the file you have, with possible exception of #2, depending what you have in mind. I think you could jiggle a few things around with that file and be on your way.

Not too hard to make a board for a one-off, prior to ordering something expensive from Express. I'd be interested in that artwork if you do it

.

 

[jamont]

I think I would like to do my own layout, for the experience. Maybe over the weekend I might get started. I'll certainly make the artwork available, and be hoping for comments and corrections...

 

[mcnowinski]

I posted a dual-voltage, tracking power supply PCB layout in a previous thread "Another mini Gilmore" that works well. It will support up to ~+/-30V with the proper transformer input AC.

Would be glad to post schematic as well, if there is interest.

Just an idea to get you started...

 

[jamont]

Your design is what I had in mind in my post above concerning that thread. As I recall, you posted the layout but not the schematic. If you have a schematic, I'd be happy to have a copy.

 

[Voodoochile]

I'll second that!

Thanks very much

.

 

[mcnowinski]

jamont,

Read your posts, but did not connect your link with my thread. ooops.

The schematic was posted to the original thread - let me know if I can answer any questions.

 

[jamont]

Sorry, I didn't mean to be obscure. Thanks for the schematic!