[kevin gilmore]

since this picture is a bit big (1.3mb) i will post it as
a link instead of a picture.

http://gilmore.chem.nwu.edu/ultrareg.gif

any comments???

 

[aos]

Looks good!

I have two points (aside from the obvious fact that some stuff is missing on the picture, like protection diodes etc., because this is just a principal diagram):

1. since you're not forcing your buffer into class A, I assume you're driving class A amps so that the p/s is always working in class A?

2. Why the unregulated 48V if the regulated is only 24? LM regulators are going to have to dissipate all that difference and they're going to get VERY hot.

 

[hsnam]

hi kevin,
quick question: what kinda TI opamp do you recommend for this PS?
will TI OPA2/37 handle this?

by the way, the KG amp i built a short a while back is working very well. thanks!!

 

[kevin gilmore]

Yes this is to drive my class A dynamic headphone amp.
The standard stereo version draws 120 milliamps, definitely
keeping the power supply in class A. The dual monoblock
bridge version draws 250 ma.

Obviously all the protection stuff etc is missing. Drawing
time less than 1 minute. Then scanning it in, another minute.
Much better than the hour or two to make it look nice.
Afterall you all DO get the idea right???

Why 48 volts?

1) I have a bunch of transformers laying around

2) The extra heat on a cold winter night....

If you are going to use the ti opamp its an op541

If you are nuts, you can use the apex microtech pa19
(stupid amounts of money, but i have a dozen or so pieces)

 

[Bob_G]

Just one comment Kevin,

Since you (wisely) included adjustment pots for the regulators, the money spent on 0.05% accurate resistors is wasted. You could use 20% resistors and still get all the performance out of the regulators, just tweak the pot till you get your voltage.

Secondly, I've used Apex stuff in the past and one simple way of ensuring very low distortion out of these power amps is to include them in the loop with a "good" op-amp on the front end.

 

[aos]

Bog_G, there's another regulator in the schematics, REF02, that gives 5V. This is (should be) a high precision low noise regulator. 0.05% resistors are setting the output voltage in relation to this one, not the one generated by the LM regulators, so their precision is not wasted. It would be a waste of money otherwise, indeed

.

On the other hand, I don't recall seeing any resistors of such precision. Even the most expensive stuff I've seen and have - Caddock and Vishay - are still "only" 1% or 0.5% at best. I'd like to know what brand is Kevin planning on using here - maybe tantalums go that low?

 

[Bob_G]

But the REF-02 is *not* a regulator but a precision voltage
reference, normally used as the Vref for ADCs.

As such the REF-02 needs no peripheral components, it's
completely self-contained.

Also, I didn't (originally) look closely enough at his BOM, but I now realize that he's specifying the 0.05% resistors in the feedback path of the op-amps! These op-amps just do not the sort of accuracy that would warrant using such precision.

0.05% *can* be had. They're normally sold in an array to manufacturers of test equipment (such as Agilent Technologies), but Kevin's are not "normal" values.

So exactly what Kevin has in mind is still a mystery.....

 

[aos]

Exactly, the precision resistors are in feedback path of the opamps, and they are the ones that set the output voltage -

Vi = Vo * R5 / (R5 + R6)

i.e.

10 / (10 + 22.8) * 16.4 = 5

which is the voltage reference generating on the other input terminal of the opamp.

Of course, if the input drift of the opamp is too high then super precise resistors are not warranted. I myself plan to use a very similar schematics, but with discrete complementary output stage driven with something like OPA277 which has extremely low drift, pA input currents and low noise (I think). Or maybe a wideband opamp.

 

[ppl]

A discrette Component output stage for sutch a Reg. Could well have enough DC offset and thermal drift to require a large feedback factor. Like almost operating the Opamps in the unity gain mode or at best no more than a gail of +5 @ -5, otherwise the Opamp may not beable to compleatly control the output stage. I think the Power opamp of Opamp/Buffer IC is maby practical route in this application.

 

[kevin gilmore]

A lot of you are not going to be happy with my resistor response,
but its true.

In the early 1960's the US goverment gave northwestern
university $1.2M (then) in ultra precision .1 and .05% titanium film
resistors. I have no idea who makes them. I have enough
of them to design and build stuff FOREVER. They are
ultra low noise. If i had to buy them today, i would not
have even a clue where to get them.

I see no reason not to use them. I know it has nothing to
do with reality, but then the whole university is surreal anyway.
Thats why i stay at NWU.

Obviously you should substitute the best parts you can get
or are willing to pay for.