[alsq]

JF and all:

Great discussion. I think I'll go for a DIY gilmore and
the dual-regulator PSU. Perhaps not the absolute best,
but likely already more than my ears can distinguish.
I'll use a 22V+22V tranny, so I'll drop the filter capacity
by 2/3 or so, I think. I may try a variant of Dimitry's rig
to test dynamic behavior, too, if I have time to waste
(yeah, sure).

My new AKG 240S just arrived, so now I need another amp for my HD580

It'll be a long while before I'm there.

JF: keep me posted on the gestation of your new baby.

 

[tangent]

Quote:

Where does Erno Borbely mesh into the fray?

New info: there is indeed a Borbely variant of the Sulzer design, and it is basically as you see it on that SB-1 page. (The pass transistor is different, and there may be a few different component values.)

I sent email to the people maintaining the SB-1 page, and while they couldn't supply a publication reference, they gave me a critical clue for tracking the origin of the Borbely variation down. At one point, Old Colony Sound Lab (the parts distribution arm of the audioXpress empire) distributed kits for this power supply, and through an old ad I was able to track down an article. In fact, it's in the same issue as the Jan Didden article (1/1987), but I overlooked it because the article was about an MC preamp, and the power supply is treated almost as an incidental in that article. Borbely says little about it other than that it's a modified Breakall design. I guess future articles didn't refer to Borbely's article because Borbely really didn't break any new ground here. The only significant changes between the two are moving from an LM329 buried zener to an LM336 IC zener, and from a zener-propped pre-regulator to a standard resistor divider configuration. Probably these two changes are a wash in terms of noise performance; the LM329 looks a lot quieter on paper, and it has better tempco, but the zener-propped preregulator probably adds noise relative to the standard configuration.

Interestingly, Borbely's current low-power supply design also uses an op-amp driving a pass transistor (a MOSFET now...big surprise) but he does away with the reference, using a trim pot instead. Apparently he's decided that getting the right absolute voltage isn't all that important, and simply having a stable relative voltage is good enough.

 

[Voodoochile]

That's cool. I love the history and evolution of some of these circuits. You should compile this bit of historical PSU discourse on your site, if only for nerd entertainment value. It seems to be quite scattered all over the place as it stands.

ALSQ: A dual 18v transformer would be plenty adequate if you intend to net approximately 16.4 volts. Just a thought. The dual 22v is going to place a fairly large load on your regs if you are going for 16.4, or thereabout. You are going to be dissipating almost as much voltage, so think big on your heatsinks.

 

[alsq]

Quote:

Originally posted by Voodoochile A dual 18v transformer would be plenty adequate if you intend to net approximately 16.4 volts. Just a thought. The dual 22v is going to place a fairly large load on your regs if you are going for 16.4, or thereabout.

Voodoo: I intend to drive Sennheisers with the
gilmore and would like to get a bit larger voltage swing,
so I am aiming at 18V + 18V. 317 has 1.5-2.5V dropout.
1963 has 100-300mV dropout. 18V should allow
to power up the DC servo opamp without the
need for a voltage reduction or a fancy opamp.

Incidentally, I have a batch of TL081 leftovers, is
this an acceptable substitution for the OPA227?
The slew rate is better, but I believe noise is
much worse, any ideas or suggestions?

Back. If I use a lower voltage tranny, I'll end up
with ridiculosly large filters, as I need at least 23 very
clean DC volts into the 317, without any allowance for
variations in the mains voltage. I thought that using
22+22 would allow me to stick with smaller filters
(e.g, -> smaller box.) I went a bit like this:
22AC->approx 29VDC -> 317 -> 20VDC -> 1963 -> 18DC
which, for 300 ma per rail, is 3 W @ 317 and
0.6 W @ 1963. In 317/TO220 the thermal resistance
is 50 degC/W, and it will toast at 125 degC, so it
will indeed need a small heat sink. The 1963 should
not. Did I do this right?

Btw, anyone has a really good feel for what is
a good multiplier to the Vpp DC downstream of
a tranny/rectifier? 1.4 is the value from RMS to
peak, but I've seen quoted differently elsewhere,
even to 1.2, too. (I know, your also have the 0.7-1.4V
drop at the rectifier, but let's ignore that). I
suppose the lower numbers account for drop
under load?

 

[tangent]

Quote:

I have a batch of TL081 leftovers, is this an acceptable substitution for the OPA227?

I believe the OPA227 (or OP27) was chosen for its excellent DC specs, since it's acting as a DC servo. The TL081 won't be horrible, but you'll probably get significantly lower DC offset on the output by going with the '27.