Super Regulator, collecting the facts

[peranders]

Somewhere here in this thread someone claimed that it was impossible to make an universal pcb for both positive and negative voltages.

http://www.diyaudio.com/forums/showt...689#post268689

Parts which are doubled: LM317/337, the reference (not entirely neccessary) and the zener. Other polarized parts ahve to be turned around, caps and LED. The only thing you may fear is the 431 which is displaced quite a bit in the negative version but you can always choose an another reference.

Isn't this task rather easy when you see the pictures?

This is the positive regulator.

[peranders]

This is the negative regulator.

Notice IC2, IC8, DZ2 and all electrolytic caps which are turned around. All transistors are changed to the opposite polarity. The LED is also turned around but you don't see it.

[diy]

Pretty good job done there!!

Du är helt enkelt makalös !
/Martin

[jackinnj]

PA (I know you are hiking in Northern Sweden) -- have you ever tried a Darlington as the pass device?

[jackinnj]

I finished up some layouts for an SMT version of the SuperRegulator -- this is what a 3D version looks like -- the pre-regulators are SOT-223 and Nat Semi has some specific recommendations for copper plane heatsinking:

[jackinnj]

why the multiple references? -- before filtering a single LM329DZ has noise of about 30uV, two in parallel bring it down to 20uV -- a third is overkill, perhaps about 1uV effect.

[peranders]

Originally posted by jackinnj
PA (I know you are hiking in Northern Sweden) -- have you ever tried a Darlington as the pass device?

Used to hike. Long time now.... I have only been at my summer cottage on the island Orust at the west coast of Sweden, still it has been a good vacation.

Your question: No

[peranders]

Originally posted by jackinnj
why the multiple references? ...

You should of course you only one of them at the time. (I'll guess you haven't read my text?)

[jackinnj]

I read that articles on your several permutations of the super-regulator -- I have nothing against the LM431 -- it just has to be compensated or it will oscillate.

Here's the setup I used to measure the LM329's -- the devices are housed in a cookie tin, balanced outputs are fed into my r.m.s microvoltmeter via Pomona RG58 cables. the tin itself is grounded to the voltmeter. with the inputs shorted inside the container the ambient noise is less than a microvolt.

with 1 diode the noise is ~31uV, with 2 it is 27uV, with 3 it is 10uV.

if you have a bunch of LM329's however, the noise will vary -- I don't think that this device was built like an REF-01 which has highly repeatable characteristics among a "bin" of devices.

I will repeat the tests with LM431's, REF-01's.

[peranders]

I have the 431 without any decoupling because I'll believe it isn't necessary. In the datasheet there is an area with unstableness with certain combinations of capacitance and current. I haven't ever experienced any oscillations nor have I provoked it to happen. I don't know how true the graphs are in the datasheet.

If you measure the noise you should do it after the LP-filter 1k+ 100 uF. I think the value at the opamp input is the important spot.

I say that LM431 LM329, LM385 and many more may be used. I don't prefer LM431 really. I think it is sufficiently good but there are better ones.