Yet another 24V, 48V Power supply - comments please

After many permutations involving various doubler circuits and grounding schemes I came up with this supply for powering two channels of DIY Neve mic pres and 3 LEDS. I have not yet built this exact version, but variations of it worked well and so did the simulation (minus reg section). The cap values are safe figures for >1/2 amp total draw, and are likely to change once I determine the actual load. I may add some more filtering and was thinking of using chokes for low pass, but have never done this. Anyone have advice?

Thanks!

Additional info and notes on construction:

Since I'm applying the phantom via standard (3.405k limits current to 14mA) the LM317 gets a 10k, 25 turn trimmer so that it can go up 52 or 54 volts if you have a mic that needs a couple more mA of current.

If the transformer secondary is really giving 24V RMS then the unregulated voltages end up being almost exactly 33V and 66V, and my DMM reads less than 20mV ripple. I'm going to use an Amveco TE70052 encapsulated, PCB-mounting toroid which has 115V primaries, not 120. Combine this with the fact that the power in my neighborhood is usually around 124V RMS so it ends up giving more like 32V RMS at the secondary. In either case, the voltage drop for regulators is a good number. I estimate that it could run on 110V RMS as well, but the drop across the 7824 reg would be at its minimum.

Siding with caution the 24V regulator should get a low-profile TO-220 heatsink. The phantom regulator should not need one.

rectifiers will be the Diodes, Inc. KBP series, 2A, 200V
resistors will be 1/4W rated metal film 1% and/or precision 0.1%
capacitor voltage ratings will be at least 30% more than peak in all cases to ensure a long life through beastly in-rush conditions. Haven't decided yet. For the filter section, I will be paralleling smaller values to save space, no doubt. There is also a chance the 100uFs will be lowered.

Currently I'm using Corcom IEC inlet modules with filtering, but I'll probably duplicate the circuit from scratch to save money.

Seems pretty noncontroversial topologically Jens. Your 317 won't be happy short-circuited, but you will always have those 6.81k R's going to the outside world except during construction and test, presumably.

What is the secondary voltage of the Amveco you mention? I'll take your word for the higher voltage but I don't know where it comes from besides the 4% difference in rated input voltage.

Of course the ripple you measure with the DMM is without load on the input to the 24V reg (?), and I am not sure what the Neve draws or for that matter how good its power supply rejection is.

The only flag that really goes up for me is the trimpot setting the phantom output voltage. Two issues come up: (1) the end-to-resistance of many potentiometers is not closely toleranced. Values of end-to-end resistance can be as much as +/-20% of nominal (10% is available and I guess fairly standard, now that I look---I'm used to working with el cheapo parts a lot); (2) Anytime I see a wiper I conjecture what happens when that wiper becomes open or intermittent. In this case it's not too serious: if you happened to get a particularly high end-to-end resistance pot and adjusted it to give 48V out (I get about 8.88k required) then if the wiper open-circuits the voltage goes to 63V or the input voltage minus the regulator drop, whichever is lower. Probably wouldn't break anything, but you would get tons of hum if the reg dropped out and the mic and pre didn't have perfect common-mode rejection.

OTOH if the pot end-to-end is on the low end of the range you won't get even your 48V (with a 10% low one you'd just get it, barely).

Maybe a better scheme would be to use fixed R's for your 48V setting with a vernier potentiometer across, say, the R used in place of the 240 ohm. I say in place of because you want the parallel value of vernier pot and new R to be 240, when the other new fixed R is 8.87k (a standard 1% value).

Now, if the pot wiper becomes intermittent, the change will be smaller and the effect to reduce the output voltage rathe than increase it.

Again, whether the power supply ripple/noise rejection of the Neve modules is sufficient to allow the use of the relatively noisy 24V regulator is something I can't speak to. Also, I don't know if the Neve modules have local filtering as well, which would make the regulator noise less of an issue.

You could increase the output cap on the 24V regulator for better transient response I think.

Brad

hmmm I admit i've never seen a PSU like that.. well at least floating the 48v section on top of the 24v section like that.. let me know how that works out. Larger trafos are so cheap that I would have just built a "standard" psu around a higher output trafo.. but this is interesting.

Chae H Ham: "but if you wanted to be a little more efficient you could get what you want with two less diodes and eliminating one cap off the secondary. "

By efficient I guess you mean cheaper? The advantage of the fullwave doubler topology is that the ripple frequency is doubled and the impedance is lowered---also there is no d.c. to speak of in the transformer, which you get with the halfwave version.

Brad

Finally, some feedback! :thumb: Excellent comments, guys! I do not have the kind of perspective or experience to pick up on these things. Allow me to address:

[quote author="bcarso"]Your 317 won't be happy short-circuited, but you will always have those 6.81k R's going to the outside world except during construction and test, presumably.[/quote]

What is the secondary voltage of the Amveco you mention?

Click to expand...

Its supposed to have two 12V secondaries, which I've wired in series for 24V. I was surprised to get as much as I did but 31V was nominal after construction. Same thing happened with a couple of other 115V:12V trafos.

Of course the ripple you measure with the DMM is without load on the input to the 24V reg (?), and I am not sure what the Neve draws or for that matter how good its power supply rejection is.

Click to expand...

Yeah, unloaded. I believe they draw 120mA max per channel, plus an LED = 260mA. I've measured about 65mA per channel with no i/o connected and the bias trim for the 3055 turned down lower than usual. For the 48V supply, total draw is <70mA.

(2) Anytime I see a wiper I conjecture what happens when that wiper becomes open or intermittent.

Click to expand...

Good point. I didn't think of this. The tolerance and reliability of the Bourns trimmers I was using was sufficient, but better play it safe (and save $3!). Fixed resistors it is!

Again, whether the power supply ripple/noise rejection of the Neve modules is sufficient to allow the use of the relatively noisy 24V regulator is something I can't speak to. Also, I don't know if the Neve modules have local filtering as well, which would make the regulator noise less of an issue.

Click to expand...

So you're saying the 7824 is noisier than the 317? I wasn't aware of this. Funny, my original design used two 317s but I wanted to reduce the number of components. The 24V is applied to the modules through single stage RC filters and so is the phantom, but I'd rather have redundancy and a sharper slope. Again, I originally had some hefty RC filtering on both outputs, but wanted to reduce heat, cost, and number of components. Oh well, no free lunch!

You could increase the output cap on the 24V regulator for better transient response I think.

Click to expand...

Couldn't figure out the formula for this so I just used the recommended value from the application notes.

That was a nice lesson for me. Thanks a lot, Brad! :guinness:

[quote author="Svart"]hmmm I admit i've never seen a PSU like that.. well at least floating the 48v section on top of the 24v section like that.. let me know how that works out. Larger trafos are so cheap that I would have just built a "standard" psu around a higher output trafo.. but this is interesting.[/quote]
I admit it is kinda wacky and a little dangerous (see Brad's comments about shorting the LM317 above), but I wanted both outputs to share the same 0V reference and could not effectively achieve this with all of the normal configurations I tried.

[quote author="Chae H Ham"]Your supply will work as it is, but if you wanted to be a little more efficient you could get what you want with two less diodes and eliminating one cap off the secondary. I'll echo the above comments as well. Also, depending on the total current draw, you might want to up your input and output cap values. I usually opt for a higher value electrolytic than a lower value tant at the output of the regulator but that's just personal preference.[/quote]
Chae,

I assume you're talking about a Villard voltage doubler? That was my original plan but it didn't perform well and the difference in ripple freq created problems when trying to tie the commons together. Two caps and a bridge is plenty efficient for me! I wasn't going to use a tant anywhere in this circuit, but will definately be putting some larger electros back in there.

Thank you very much for your comments and for helping to make this great forum possible! :guinness:

Jens, do you really mean 70mA of phantom? That's a bunch! That may challenge even your fullwave doubler as far as low enough ripple. BTW I've used that design plenty and it works great. I hadn't seen it before and then in this last week saw it twice, your schematic being one of those. Wear safety glasses if you think you might have the capacitor polarities reversed the first time you turn it on---but that's good practice for nearly all power supplies!

The three term regs are all pretty noisy, but then maybe they are good enough. There's a ref somewhere about regulator noise and I'll post the link after I dig it up.

Three-term regs are so convenient and often quite adequate, but it would be nice to have something a little better without it running into a million parts. Maybe I'll give that some attention.

I need to do some simming (or even measurements) on the Neve designs, but regardless of their PS rejection there is the susceptibility of the mics to power supply ripple/noise. This is usually not too great because of the low inpedance of the mic output and the balanced operation, but I wonder if anyone has actually measured the sensitivity to ripple on the 48V for some popular mic.

The output cap on the 7824 is not a critical value---it needs to be at least a tenth mic usually for other than wacky transient response, but if it is too large the reg will current-limit a bit on turn on (no biggy), and then backflow the stored charge on turnoff if there's another load pulling down the unregulated voltage. But you've got a diode there to protect against that being damaging, and no other load pulling down, so you are ok. I'd use maybe 47uF and the 100nF in parallel, but it's not critical.

Here's the link to the articles about regulator noise. A lot of good data and remarks imo: http://www.tnt-audio.com/clinica/regulators_noise1_e.html

I use BYV27 diodes, then I can skip the extra caps over the diodes. Adding caps over these diodes will increase the spikes instead.

[quote author="Tekay"]I use BYV27 diodes, then I can skip the extra caps over the diodes. Adding caps over these diodes will increase the spikes instead.[/quote]
In other words, ultra fast diodes don't need to be bypassed? Why is this? I can't find BYV27 at mouser or digikey (Im in the U.S.) but a number of cross references come up. I was hoping to use bridges in the KBP or SIP-4 package. I'll have to shop around for something with the same kind of specs as a BYV27. Thanks, Tekay!

Bcarso,

This article is very interesting sofar. Thanks a lot! I've only played with a discrete regulator once and it wasn't very good. My goal is still low cost and a minimal amount of parts, but fidelity is much more important than that so perhaps a reevaluation of my regulator section is needed. I'll get back to you on that.

The 70mA for the phantom assumes two mics at 14mA maximum and two LEDS at 20mA. Many power supplies rely only on a shunt regulator relying on zener diodes to set output for the phantom supplies, which isn't very reliable in my experience. Also, all of the designs I see around here are LM317 based so I got the impression it was more than sufficient. Would be interesting to know for sure. I'll go do some homework. Thanks again!

[quote author="Family Hoof"]The 70mA for the phantom assumes two mics at 14mA maximum and two LEDS at 20mA.>>

You can save some current capacity by running the LEDs at 5mA apiece or less. The current vogue for glaringly-bright LEDs mystifies me; I have a couple sitting here in the workroom that might as well be nightlights.

Peace,
Paul

Ackk! A 48V supply for a 3.6V (max for blue even) LED! Acckk!

One would like to keep switching regs the hell away from mics, but that's almost tempting considering the waste in power.

Amen to Paul S on getting more effiicient ones at least.

Jens, see also the article posted by tmbg in the Drawing Room thread on voltage doublers, about active filtering of the reg outputs (wait I have the link here: http://www.techlib.com/electronics/finesse.html

Brad

PS: re fast diodes: they don't per se eliminate reverse recovery spikes but there are some "soft-recovery" ones that make the transition have less way-high-freq energy that tends to leap off the board and get into everything. The effect of shunting diodes with caps only is to shift the ringing energy to lower frequencies. It does have the unfortunate side effect of making the diodes pass more hf energy coming in on the transformer, so it's a mixed blessing.

Thanks Brad, that's a very interesting article. Im surprised this very simple Zener + Emitter Follower circuit on page 3 acutally works quite well. I used a very similar circuit for a phantom power supply. I always meant to replace it by a TL783 circuit, but now I think I better stick with it.

I may be wrong about this, but I'm pretty sure that's no ordinary zener, it's a 3 term regulator of some kind.

The BZX79 mentioned in the article is a normal zener AFAIK. Or do you mean TL783 ? That's of course a regulator. I meant to replace the whole 48V circuit, not just swap out parts.

Rossi, are you perhaps referring to the other article on voltage regulator noise? The one just above is for an active filter for noise.

[quote author="bcarso"]Ackk! A 48V supply for a 3.6V (max for blue even) LED! Acckk!

One would like to keep switching regs the hell away from mics, but that's almost tempting considering the waste in power.

Amen to Paul S on getting more effiicient ones at least.[/quote]
I don't understand. Each channel gets its own SPST switch followed by an LED in series with a 10k resistor and the 2x6.81k w/local bypass cap configuration. What is wrong with this? Isn't that how everyone does it? I guess it is wasteful but that's something I wasn't even thinking about. If ripple is less without LEDs I will gladfully ditch them.

Anyway Brad, all of your responses and linked articles have caused me to doubt relying on IC regs to do all the work. Now I'm starting to think maybe I should learn more about power supplies and really design something good. Thanks for your help!

[quote author="bcarso"]Rossi, are you perhaps referring to the other article on voltage regulator noise? The one just above is for an active filter for noise.[/quote]

Yes, I was referring to the other article. This thread is too fast for me, I guess :roll: