Regulated supply rails for amp - as big as you want.

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I'm getting this itch to build another multi-channel gainclone; something that will drive eight 10 inch woofers as subs. Now, my collection of transformers never seems to have something that will give + / - 35 volts, more like 43 volts is what will be produced by those I have a few of laying around. Ones like the one that blew the gizzard out of my ETI-480 sub amp a while ago :rolleyes: So.... what to do? Why not bite the bullet and build up a regulated supply that 1/ brings the DC voltage down to something more suitable, 2/ is totally hum-free, and 3/ annoys the "1000uF is best" crowd ;) all at the same time? A hat-trick says I.

To cut a long story short I simply turned one of my more unusual class A amp circuits into a regulated supply. The performance is really quite good. The output voltage is always nominally 1.25 volts less than the zener voltage. The output voltage stability and regulation depends almost entirely on the zener itself. As an example, I measured the voltage between the top of the zener and the fet source with 17mA preload (the 2k resistor) and the source was 1.23059 volts lower than the zener. Now stick a 24 ohm load across the output for 1.4 amps and the voltage difference rises to 1.23065, i.e. the output sags by 0.00006 volts. 60 microvolts.

That means if we had a perfect zener or other reference source the output voltage would drop from say 35 volts to 35.99994
volts. That's 0.00017%. I could live with that. In practice the zener current and therefore voltage varies a little with load because zener current drops with increasing load. The cap across it minimises this effect for sudden load changes though. This same cap also gives the supply a soft start as it charges up.

One of the things about this voltage regulator cct is it's scalability. All you need is a suitably rated N-channel mosfet and heatsink and caps to make the current or voltage or both as high as you want. :smash: This is due in part to the floating 12v supply that feeds the gate pullup resistor. I only prototyped it with 50 volts going in but there is nothing that I can see that would stop it even being used for vacuum tube psu's.

The first drawing is the basic circuit.
 

Attachments

  • single rail.png
    single rail.png
    6.4 KB · Views: 1,397
Dual rail version

This second drawing is how to use it for split rails like in an amplifier. Input 1 and input 2 *must* each be fed from their own =separate= transformer winding / rectifier / filter. Either 2 completely separate windings on the same transformer, or 2 transformers. Trying to run them both from one winding will only make a big fat short circuit.

Now, just gotta make the amp to use all that juice! :D
 

Attachments

  • dual rail.png
    dual rail.png
    12.9 KB · Views: 1,351
Graham,

Very, very clever.

However, the speed of the LM317 is a tad limited over 20KHz I believe. Have you considered using a simple diff pair with a voltage reference (could be LED, bandgap or zener) one side and a voltage divider from the output? You can use the second transistor's collector to drive the gate of the source follower.

I built one; blistering speed, a little overshoot, but really good rejection characteristics, and not too much feedback.

Cheers,

Hugh
 
Are Fast Supplies Neccesarily Better ?........

Does the regulator need high speed, and is this actually an advantage if there is sufficient decoupling capacitance on the output side of the regulator ?.
By this I mean, if the regulator response is relatively slow (only fast enough to respond to envelope average demand) and feeding relatively large capacitors, then one could efffectively have a power supply with the characteristics of a standard transformer/bridge/caps supply but controlled average voltage and without ringing/overshoot problems.
The PSRR of these gainclone chip-amps is best at lower frequencies of psu noise/ripple, so why not keep possibility of high frequency noise/ripple/overshoot right out of the equation ?.

Eric.
 
Hi iCebReakEr409. The losses on the main mosfet in watts would be the volts dropped across it times the amps through it. For an amplifier as a load, the current and so the actual watts loss would be going up and down all over the place so the heat loss would depend on the "duty cycle" of the music.

Gidday Hugh. Yeah, I just built it that way because I have this thing about LM317's. Very docile they are. I also tried a TL431 and it just wanted to oscillate and I didn't really feel like sussing it out. As far as speed is concerned, if the output cap is big enough then it's hardly an issue IMHO. Always keen to have a listen to your thoughts though! Gotta get over and see you again sometime soon too.

I did another measurement on it this morning, this time from 1.4 -> 2.8 amps and under these conditions the voltage only drops by 15 microvolts. :) Seems that with higher current the 32 amp fet I am using gets to the steeper part of it's transconductance curve where it starts to boogie.

Eric, yep, I'm with you.
 
" "1000uF is best" crowd "

I was at the NS site the other day reading some of the app notes for the various 'clone' chips.

The 470µF~1000µF caps recommended are for local bypass, they are not the main filter caps.

Where did those guys ever come up with the idea that those were the main filters?

As regards full regulation, do we really need it?

Maybe for the 'clones', because they are all on one set of supply pins, but for an output stage? I find regulation adequate for just the voltage gain stages, say 50mA, and a no feedback design to be adequate in most cases as most voltage gain stages are biased class A. Electrocompaniet uses a simple series pass transistor with a zener in the base, to me they sound better than most three terminal regulators.

Something disgusting to think about:

Virtually no filter cap ahead of the pass transistor. I was able to use only 220µF to keep the regulator from dropping out with a 60hz line. The pass transistor needed less heatsink too.
 
I think it comes down to sonics/dynamics tuning regarding the 1000 uF idea.
The 1000 uF supplies cause throttling down of these GC chip-amps, and the PSRR (or lack of) interacts with light speakers to give a 'bouncy' bass, that seems well regarded by some, and is probably reasonably appropriate at the power levels of these chip-amps.

Grahams regulator idea is mainly to enable usage of whatever power transformers he has lying around.
There had been talk of GC amps changing sound according to supply voltage, with a small range regarded as optimum.
As I see it, Grahams regulator could be used as a means of tuning the amplifier sound by varying the supply voltages - also useful to reduce supply voltage when driving 4 ohm etc loads.
For a practice guitar amp it could be used to set max power/distortion characteristic.

Eric.
 
Eric, Graham,

Thanks for your thoughts on the use of a filter cap to scotch any speed issues. I'll certainly think on this POV more; it makes a lot of sense.

Recently I was faced with designing a +5V power supply for a DAC. Looking at the clock speed, and the on/off nature of digital current demand, I felt speed was necessary, if only to lessen the load on the output cap, since electros probably won't be adequate. That was the thinking behind this; analog was far from my mind.

Graham, I recently built a 1.5W SET using the 6EM7 out of simple curiousity. I'd encourage you to have a listen; it certainly doesn't do it all well, and the AKSA clearly beats it on dynamics and detail, but there is something beguiling about the female vocals in particular which typifies just about all SETs, and I'd love your opinion.

I've got a couple of SV811-3 Svetlanas lying around unusued, and am considering taking the plunge, as the lovely vocals and sound stage has me entranced. Tube lineup will be 6BQ7A SRPP, 12HG7 (pentode!!), and SV811-3, cap coupled all the way, fixed bias, running 600V and 75mA. I will incorporate the secondary into the cathode circuit, should get around 15W, should really boogie....

I'm also working on a tube/PP SS hybrid which might interest you, very trick.

Cheers,

Hugh
 
Slow And Steady Wins The Race............

Hi Hugh, my thoughts are that the regulator should be as slow as possible, like only a few hertz response speed.
5 Hz response presumably would compensate the envelope power demands of music running at 300 BPM !!!, which I don't think exists.

At envelope repetition rates less than this, the supply should effectively work like a standard rect/cap supply with sufficient capacity, especially if it were to have a CLC filter at the output of the regulator, except no rectifier noise/ringing etc.

There would be no 100Hz short duty cycle current pulses to excite electrical resonances in caps and circuit stages, and may go close to duplicating a good battery supply (batteries are relatively slow).

I find op-amps sound good (best ?) with a WELL filtered but non regulated supply, and other constructors say the same about battery supplies.

Regulators are after all power amplifiers, and prone to the standard stability considerations when driving reactive loads, and this can be evident as HF noise on the rails, and op-amp (and most circuits) PSRR is worst at HF.

I think a slow response regulator with CLC output filtering whilst not providing perfect regulation ought to be the quietest way of providing AC sourced power.

Eric.

Be careful of those tube things - I hear they are dangerously addictive.
 
No batteries. They are just to represent a floating supply where the negative end goes unmistakably to a certain place other than ground. I drew it at work and so I was unable to create a suitable symbol.

Also, I wonder if the "1000uF is best" people have ever put a scope on the supply rails when their 'clones are really pumping. I mean, 1000uF (at 100 Hz rectified) at 1 amp load makes for 10v p/p ripple. :dead:
 
millwood said:
how much of that ripple would show up on the speaker terminal?
I gather your inference is that it would be small enough to neglect completely. ***

If assume the power supply rejection ration (PSRR) of the amp chip is 80dB then the ripple voltage would be reduced by a factor of 10,000 i.e. to 1mV. If the ripple is equally in opposite directions when looking at both rails at the same time it should cancel completely I think. Of course, if we started off with lower ripple in the first place then it would end up proportionally lower again. I know what I would rather have.

*** This would be an excellent question to ask the Magic Power Cord worshipers. Ding! -> Checkmate! :smash:
 
Circlotron said:

I gather your inference is that it would be small enough to neglect completely. ***

*** This would be an excellent question to ask the Magic Power Cord worshipers. Ding! -> Checkmate! :smash:

Yes, it would seem that a power lead ought not to make a difference, but in practise I find that they can make an extraordinary difference.

Just yesterday I tried a 'magic' power lead on a friend's system - Sansui CD-X711, Sansui AU-X911 and Yamaha NS-1200 speakers.
By any standards this is a very nice system, but the improvement in dynamics, clarity and musicality, loudness and sense of power was dramatic and unmistakable.

Eric.
 
djk said:
" "1000uF is best" crowd "

I was at the NS site the other day reading some of the app notes for the various 'clone' chips.

The 470µF~1000µF caps recommended are for local bypass, they are not the main filter caps.

Where did those guys ever come up with the idea that those were the main filters?

Yes, National talks exactly about the local bypass caps but National talks about these caps right after the claim that 3875/3886 have excellent PSRR and do not require a regulated supply. National here uses copula “however”, thus suggesting this might be the complete PS consideration.

But honestly I’ll rather not be much scholastic about the National’s datasheets here, I anyway know that these chips do like the regulated supplies (be they “required” or not).

Pedja
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.