bench/lab/ audio power supply

[martingriffith]

My 25+ year old home made supply just died, dried caps mainly, I dug out the original circuit, and re draughted the PCB layout.
The original had amazingly good specs, for noise, I never did a transient response test, since it was only used for line level audio stuff.

http://uk.geocities.com/martingriffith/labpsupcb.pdf
http://uk.geocities.com/martingriffith/labpsusch.pdf

It still needs a bit of work on the layout, so any comments befor I get a prototype board made?

martin

 

[bcarso]

The values of R25 and R26 at 100 ohms each seem way too small---the divider current would be 25mA. Perhaps those values are misprints. You don't state the unreg voltage, but it clearly isn't much more than 40V since that's the bulk cap rating, and with 40V the available total current to the TL1431 and everything it feeds is about 15mA.

I can't comment on grounding and routing because it isn't complete yet, but as you probably know those details are often the secret of really low ripple and good regulation.

Is the current limit when engaged stable against oscillation?

 

[martingriffith]

Ah, well spotted the 431 resistor values. Its just a case of getting the schematic correct and then work out the values, while th PCB is being made. I was going to make 2 up, for + and - supplies. The current limit works fine, never seen it oscillate in the old one.
If I want to get more than 30V out, I put 2 in series or parallel for more current

As you say the grounding is most important, so I'll thicken up the power traces to 3mm or more if possible

 

[Svart]

C7-C10 we've discussed the merits of these before and are likely not needed. I'd think about changing these over to power Schottky and loosing the caps, but that might be overkill..

There are some extensive discussions on PSUs in the drawing board, do a search.

Tl1431..groan.


:green:

Also looking at this.. Grounding is unclear.. by negative rail do you mean ground? It looks as if the "negative" rail is floating.. but not by much through that 0R3 resistor..?

 

[bcarso]

The trick with power supplies and power amplifiers for that matter is to look where the big currents flow and make sure you are not assuming that different portions of a current-carrying trace are at the same potential. Hence the vast improvement with remote sensing of the output teminals, and making sure that the voltage reference is similarly referenced to the same "ground." Bigger and thicker traces are fine but the geometry/topology is vastly more significant.

The simplistic doctrine that embodies a good deal of this advice is called "star grounding."

To Chris's question: the current sensing R is in the low side of the unreg supply, and "ground," in the sense of output common, is slightly less negative depending on current. The important reference points are the low sides of the TL1431 and divider/filter/voltage adjust pot, and the low side of the resistive divider on the first 741. As well the top of the 741 divider R should be routed to the output positive terminal for the lowest output impedance.

Note that the auxiliary negative supply gives the op amps something to work with so their input common-mode ranges are not exceeded. An alternative might use ground-sensing op amps like the LM124 series, with their neg supply terminals tied to the low side of the bulk cap, although those particular ones are quite slow.

 

[Svart]

ok I looked at the schematic for a bit longer.. it makes *more* sense now but the way it's drawn somewhat confuses.. or is just me?

I usually draw and have almost always seen the current sense R directly under the TL1431.. same for current sensing PWM ICs that I use, the current sense R is in series with the (mosfets for me) to ground with a tap between the mosfet and the R.

 

[bcarso]

It wasn't immediately obvious what was going on to me either. But this is typical for the designer's draft of a schematic when he/she already knows well how the circuit works, unless instruction of others be the intended use of the document. And making something that one would be comfortable teaching from is a lot of work, or at least requires considerable skill.

OTOH, I've handed manually scribbled schematics to a board layout person---before I had a virtual gun held to my head and was forced to learn a schematic capture program---and what I got back as a machine schemo looked so bizarre that I was at first certain that it couldn't be correct :shock: