On-Card regulation --- routing

[Freq Band]

I'm building several mic pre's in one case, each powered at  +18v...(each with one THAT1510, and 1646 per board.)
The power transformer will be in the same box.

I've never wired for "on-card" regulators, so I was wondering...
First of all....are there any real benefits ?

If so, should I....

1) Pre-regulate...Use a transformer, and a regulator circuit with high enough volts/amps to account for the dropout of the smaller regulators that will be "on card" ?

2) Use only a  transformer/rectifier/filter caps (accounting for dropout downstream)...and send that unregulated DC to each card's regulator(s) ?

3) Use a method such as this.....?

=FB=

 

[Ian MacGregor]

Hmm...

I'm not a big fan of option 2, as then you've got DC with a bunch of AC (noisy) on it routing around your audio circuits. That hum could build up...think tracking drums with your pres and every channel has the same tiny bit of hum in it. Becomes not so tiny...

I'm a fan of the "pre-regulate" option and I am starting to experiment with using a SMPS to get a DC voltage a couple volts above my desired rail, and then having onboard LDO (low drop out) regulators doing the rest. Less transformer cost and heat dissapated by the linear regulators (which are less efficient than switching). I think Jim Williams had a paper on using linear regulators to sub-regulate switch-mode stuff.

I suppose you have to ask yourself whether you really *need* to regulate each pre card individually. Can you just have a PS PCB that connects to all the cards in parallel with a ribbon cable ala Seventh Circle pres (and a gazillion of commercial stuff)? I'm also curious as to the pros/cons.

I suppose if all the pres draw an appreciable amount of current, and that current is multiplied by successive resistive loss in the connecting cable (where the pre with shortest power cable has least resistance and longest cable has greatest) - then the farther the pres get from the supply, the lower the supply voltage. But by how much do we care? Even if pre #1 has +/- 18.0V and pre #8 has +/- 17.0V, the difference in headroom is roughly -0.5dB (correct)??

I suppose there may be crosstalk issues by having all the pres using the same power bus, but now I'm sorta guessing...

ian

 

[Samuel Groner]

Option 1 is my way to go. Option 2 has been covered by the previous post and option 3 may result in interaction which will counteract any potential advantages of local regulation.

Are there any real benefits?

It's all a function of the PSRR of the signal circuit. The THAT1510 datasheet I have at hand does not show meaningful PSRR figures (at high frequencies) so it's a bit of guessing.

The farther the pres get from the supply, the lower the supply voltage.

Correct in theory, but have you tried to estimate the order of magnitude we're looking at? A THAT1510 draws 6 mA typical. Say the cable has 100 mOhm (likely a pessimistic value)--so we end up with 600 uV difference. Your handheld DMM will not even be able to measure the supply voltage to this accuracy.

Samuel

 

[abbey road d enfer]

I'm building several mic pre's in one case, each powered at  +18v...(each with one THAT1510, and 1646 per board.)
The power transformer will be in the same box.

I've never wired for "on-card" regulators, so I was wondering...
First of all....are there any real benefits ?

If so, should I....

1) Pre-regulate...Use a transformer, and a regulator circuit with high enough volts/amps to account for the dropout of the smaller regulators that will be "on card" ?

2) Use only a  transformer/rectifier/filter caps (accounting for dropout downstream)...and send that unregulated DC to each card's regulator(s) ?

3) Use a method such as this.....?

=FB=

First I would NOT go for on-card regs. The big problem is current noise injection; even with good decoupling caps these currents can induce longitudinal noise. At system level, a single centralised PSU with good grounding concept is always cleaner.

 

[Freq Band]

I'm leaning toward "simpler is better"...and probably will just build an ample regulated 18v supply, and routed to each card....no on-board regs...(although the idea is tempting ) Plus, I haven't heard anyone mention any benefits that might win awards.
And...the psu will be only 8-10 inches away from the preamp boards.
I assume on-board regs would be useful if they were some distance away from the transformer, or if there were circuits that "kicked-in" with large power drains,  or if there was a need to isolate digital circuits, from analog circuits.
But it's good to know how to do it. Thanks everyone.

=FB=

 

[JohnRoberts]

In a large physical structure requiring high reliability, independent regulators on each card can isolate and prevent local faults from taking down the whole system. For a modest sized project a single regulator with good local PS filtering seems more logical.

I have made large consoles with 3 term regulators on each module. (your option #1). It worked but I did have to deal with regulator noise corrupting the ground, from an unusually noisy batch of regulators.

JR

 

[Samuel Groner]

I did have to deal with regulator noise corrupting the ground.

Just to make sure I understand correctly: the noise current of the ground/adjust pin of the regulator was enough to cause troubles?

Samuel

 

[JohnRoberts]

I did have to deal with regulator noise corrupting the ground.

Just to make sure I understand correctly: the noise current of the ground/adjust pin of the regulator was enough to cause troubles?

Samuel

No. This was a few decades ago and and didn't pursue it beyond, black balling that one brand of 3 terminal regulators when they refused to provide or meet a noise spec (six-sigma my ass, it's easy to pass the spec when there is no spec). The specific noise source was on the output of one of the two, three-terminal (78/79xx) regulators per channel PCB. This noise from say 30-50 regulators, got dumped into the ground by all the PS decoupling caps on the channel circuit boards. I am not sure of the exact vector that this power supply ground noise managed to get into my bus noise floor, but there was over 100 feeds to the two-mix in my largest configuration so it was "corruptible".

The noise went away with any one of the several other brands of 3 terminal regulator, so I want back to work on other projects. One of my disappointments was that I didn't get to release the second generation board set for that design, but the market was buying $4k inline consoles and I was trying to sell them $20k splits.. Wrong product for the time and market.

JR

 

[okgb]

I have  / built a quad Great River MP-4
with power supply separate case feeding
A.C. to the rectifiers & filters ,  dc to
the boards with on board regulation
No problems , it's done Classical gigs
running unbal out sometimes , no problems
[ not the prettiest layout but was crampt ]

Also a smarter friend of mine claims the 3 pin regulators have their own noise problems
and recomends discrete  [ 2 or 3 transistor ] versions

 

[JohnRoberts]

The early 3 terminal regulators (78/79xx) were based on 741 opamp level performance. Their output noise was not great (also a function of v ref), and their output impedance was not great at HF.

I found paralleling them with 1,000uF electrolytic capacitor would nicely complement and flatten out, the rising output impedance of the stock 3 terminal internal circuitry.

In general the significance of PS noise depends on the PSR of the circuitry, while there is a tendency to over engineer PS sometimes, because it's there. I did one such phono preamp, where I rolled my own PS from TL072 and some pass transistors. Much better than it needed to be.

JR

PS: It was probably 15+ years ago that I had my little skirmish with regulator noise, at the time I only found one vendor measurably worse than the pack, and unwilling to deal with it. 

PPS: Others here have been critical of my "paralleling with a cap" strategy, so I will only say it worked for me.

 

[Viitalahde]

I'm thinking this too, for a new EQ. It'll have the Sjostrom/Jung regulators on board, but it'll use an outboard power supply.

I'm leaning towards option 1 myself. A sturdy supply with LM338K pre-regulators should work just fine. The raw DC approach interested me too, but I guess I worry about the possible induced noise. Dunno.

 

[Samuel Groner]

This noise got dumped into the ground by all the PS decoupling caps on the channel circuit boards.

Thanks for the elaboration, that makes more sense.

I'm thinking this too, for a new EQ.

The real high-end solution are local shunt regulators. Gives much higher isolation if you can stand the dissipation.

Samuel

 

[abbey road d enfer]

The real high-end solution are local shunt regulators. Gives much higher isolation if you can stand the dissipation.

Samuel

Can you elaborate on that? Why should it be better than series reg? And the supply source needs be ultra clean, I think. I reckon any residual noise at the source translates in high current dumped in the 0v return.

 

[Samuel Groner]

I hope it is fine if I just redirect to what Putzeys has written on the topic (covers all questions): http://recforums.prosoundweb.com/index.php/t/14477/0/ and grimms recipes.pdf

Samuel

 

[JohnRoberts]

Shunt regulation in concert with a constant current source... 

Seems like the power supply equivalent to class A audio circuitry.  Waste all the power all the time.

In principle the 0V rail (assuming a split supply) will not be pure DC but the input current, less audio (AC) current steered elsewhere by the circuitry powered by that shunt regulator.

I don't see the benefit, but plenty of added cost when applied to a typical system requiring distributed power. Heat build up is already enough of an issue inside consoles.

JR

 

[Viitalahde]

The real high-end solution are local shunt regulators. Gives much higher isolation if you can stand the dissipation.

I don't know about shunt regulators.. I am going for separately regulated rails for both L & R, which should give a proper isolation with good grounding techniques and well-bypassed sections for each filter.

The SMPS pre-regulator supply started to interest me. What are good SMPS's to use? I've seen Lambda mentioned, and that's what I'm looking right now from Farnell.

 

[Ian MacGregor]

Check out the SMPS modules made by Triad, available from Digikey and Mouser (and other distro)... I just got a couple and while I haven't got around to using them yet, they were decently priced.

Ian

 

[abbey road d enfer]

I hope it is fine if I just redirect to what Putzeys has written on the topic (covers all questions): http://recforums.prosoundweb.com/index.php/t/14477/0/ and grimms recipes.pdf

Samuel

Interesting discussion, based on the hypothetic existence of a noiseless constant-current source...

 

[Samuel Groner]

Based on the hypothetic existence of a noiseless constant-current source...

A standard BJT CCS at 100 mA has about 20 nArms current noise in a 20 kHz BW. That's much less than what flows through a 100 uF decoupling cap after a standard series regulator (a very rough calculation leads to about 100 uArms).

I'm not advertising the shunt regulator as holy grail or a must; it's just that it will be much more effective in reducing crosstalk (which I presume is the main intention of local regulation). Whether the improvement is needed or not is another matter which cannot be answered without clear context.

Samuel

 

[blue_luke]

I have been using solution 1 for years...
Here is how I go about it...

The raw supply uses a transformers with at least the same regulated output.
Say +/- 15 volts, so a 30Vac CT.
Then It goes trough a diode bridge where each diodes is bypassed by a .01 or .1 uf to the reservoir capacitor (2000uF by Amp). This reservoir cap should be made with many smaller values in parrallel instead of a big one. Keeps the ESR lower somewhat.

Then it goes trough a capacitor multiplier using high hfe darlingtons. (TIP122/TIP125 for example)
A side benefit of this is that the power turn on is nicely ramped and takes a few seconds to reach its full voltage.

Then LM317/LM337 pairs on each card with a 100uF caps bypassed with a .1 at each cards.
Works everytime, no noise problems.
If you read Walt Jung article on super low noise PSU (the Sultzer regulator), he shows graphs of tests he's done and you will find that the LM317/LM337 pair is very good indeed.
I stay away from 78XX/79xx series of regulators.

Luc