EEVblog Electronics Community Forum

Electronics => Beginners => Topic started by: look4awhile on September 16, 2023, 01:23:39 am

Title: noisy linear ps design
Post by: look4awhile on September 16, 2023, 01:23:39 am
Hi. I'm having trouble with my linear ps design for tube amplifier. I'm using capacitance multiplier to filter, and TL431 to stabilize output. It seem to regulate the voltage fine, but I'm getting a lot of noise on the output. Please suggest. Attached are schematics, PCB layout, and the kind of noise i get on the rails under 650mA load.
Title: Re: noisy linear ps design
Post by: dobsonr741 on September 16, 2023, 01:47:24 am
What is the voltage in and out. You have both ripple and ringing by the look of it .
Title: Re: noisy linear ps design
Post by: DavidAlfa on September 16, 2023, 02:11:12 am
You want a power npn there to make a Darlignton, use a TIP41 or TIP43 instead the tip42, which makes no sense.
Or flip  the polarity so the emitter is at the regulator input.
C2 and C4 so large are a bad idea, will make the regulator response slow.
Title: Re: noisy linear ps design
Post by: look4awhile on September 16, 2023, 02:37:40 am
Looks like I connected TIP42 backwards :( It was supposed to be Sziklai pair.
The idea of using 2 low-pass filters I copied from https://sound-au.com/project15.htm
Title: Re: noisy linear ps design
Post by: look4awhile on September 16, 2023, 02:38:58 am
In is ~17v (from 12v transformer), out is 12.6 v. I'll post update once I fix my transitor.
Title: Re: noisy linear ps design
Post by: DavidAlfa on September 16, 2023, 03:06:44 am
Yeah that capacitance multiplier circuit maintains the base current very constant, so the load will also be powered by a pretty much constant current = low voltage variation.

I made a noisy input, 24V DC + 10V 10KHz noise. Yeah a bit crazy but see the output difference.
Then added a 5Amp load, switched at 1KHz. See how the big capacitance completely kills the regulator response.

Since you're powering an amplifier, it will also have quick load trasients. So you want a fast regulator.

Sincerely, I'd just use a LM317!
Title: Re: noisy linear ps design
Post by: look4awhile on September 16, 2023, 05:16:57 am
Fixing transistor addressed the awful noise, but it very much breaks down under load; with Your explanation I can see why.

The tube amp I'm building this for is very sensitive to power supply. Standard 7812 or LM317 circuit as is produces very noisy mains `humm`. Of several power supplies which I tried, only one from Apple laptop worked (and god knows what it has inside). So I've tried several circuits, and the only one working reasonably well was that with capacitance multiplier, since it has this strong low pass filter. Next issue was the voltage drop. When using standard 12v transformer standard linear regulators work fine, but if coupled with capacitance multiplier they don't have enough at the input. With the circuit above I was trying to avoid additional voltage drop while having regulated 12v output, so that I can keep using 12v transformer. But I guess I'll have to go with 18v (or 24v).

I've redesigned with 7812 (its more components than LM317). Does that look reasonable?
I've tried something very similar before, only it did not use szikai pair; I used TIP41 (and not enough current was going through). I've also tried replacing it with TIP120 (Darlington pair), and somehow the `humm` was back (possibly due to high B).
Title: Re: noisy linear ps design
Post by: WimWalther on September 16, 2023, 05:30:39 am
Just for drill.. try cutting C3 down to 22uF + 0.1uF film.. and then see how it behaves.
Title: Re: noisy linear ps design
Post by: DavidAlfa on September 16, 2023, 06:30:59 am
Have you tried the original circuit removing the 470uf caps?
Title: Re: noisy linear ps design
Post by: look4awhile on September 16, 2023, 06:38:27 am
Yes, I've tried removing 470 caps and just using TL431. It regulates fine, but the 'humm' is back, i.e. it no longer has low pass filters.
Title: Re: noisy linear ps design
Post by: DavidAlfa on September 16, 2023, 06:44:35 am
And if you add the 7812 after it?
Title: Re: noisy linear ps design
Post by: look4awhile on September 16, 2023, 06:50:22 am
I'll try first thing tomorrow and will post the update.
Title: Re: noisy linear ps design
Post by: Kleinstein on September 16, 2023, 09:34:55 am
The circuit has a good chance to oscillate as there is the TL431 trying to regulate the voltage and than capacitors C4 and C3 interfering with it. It it actually oscillates depends on the ESR of the capacitors and maybe parasitic inductance of the layout / exact variant of TL431. In addition the Sizlaki stage on it's own is somewhat prone to oscillation.

The initial reported noise looks very much like a mix of ripple and oscillation.

On should remove C4 but of cause keep C2 (possibly even enlarge it or increase R1 a little)
C3 is also a bit large - some 10-100 µF plus a film / ceramic would likely be better.
Title: Re: noisy linear ps design
Post by: iMo on September 16, 2023, 09:55:35 am
Hi. I'm having trouble with my linear ps design for tube amplifier.
Such a complex regulator for your filaments? No regulator/filter needed with filaments, afaik..
Title: Re: noisy linear ps design
Post by: look4awhile on September 16, 2023, 10:03:50 am
Not for filaments. Tubes is powered at 12v. Its modified version of http://electronics-diy.com/class-a-12au7-tube-headphone-amplifier.php (http://electronics-diy.com/class-a-12au7-tube-headphone-amplifier.php)
Title: Re: noisy linear ps design
Post by: JustMeHere on September 16, 2023, 12:08:30 pm
That bottom trace that connects to GND looks like it could be a problem.  You have a ground plane, so that trace shouldn't be necessary.  You have two ground paths.  They are right next to each other, so I wouldn't think so, but maybe. 
Title: Re: noisy linear ps design
Post by: look4awhile on September 16, 2023, 03:08:46 pm
I've reproduced on a breadboard. Same exact rining. Removed TL431 and replaced it with resistor - same thing. Its seems like its sziklai pair - gonna try on the proper board today, with 7812 in series - and with TIP41C as voltage follower in the capacitance multiplier section instead of pair.
Title: Re: noisy linear ps design
Post by: Vovk_Z on September 16, 2023, 03:24:58 pm
If you really need only 0.65A then any linear regulator (7812, 317 itself will have much better properties than any capacitance amplifier.
Title: Re: noisy linear ps design
Post by: look4awhile on September 16, 2023, 05:36:17 pm
If you really need only 0.65A then any linear regulator (7812, 317 itself will have much better properties than any capacitance amplifier.

I get fairly loud 'humm' with just linear regulator. I've tried several versions - both DIY and commercial. I think it has to do with tube amplifier.
Title: Re: noisy linear ps design
Post by: DavidAlfa on September 16, 2023, 05:37:23 pm
Maybe the tube amplifier design is just as terrible as the psu  :-DD
Title: Re: noisy linear ps design
Post by: look4awhile on September 16, 2023, 07:06:16 pm
Maybe the tube amplifier design is just as terrible as the psu  :-DD

It's this one. http://electronics-diy.com/class-a-12au7-tube-headphone-amplifier.php (http://electronics-diy.com/class-a-12au7-tube-headphone-amplifier.php)
I did some modifications so that it has slightly better noise and frequency response characteristics, but even original design was just as terrible.
Either way its a gift/joke so I doubt I'll be going very deep into tubes. I prefer 70's technology to 20's :)

Now, for the update. I've tried attached design and it works well. It does use 18 volt transformer due to voltage drop, and I'll be redesigning the board to accommodate 1 smaller and 1 bigger heatsink - LM7812 gets reasonably hot (and TIP42C gets a bit warm). Its not an award winning PSU, but no crazy noise - and, most importantly no 'humm'.

I might try schottky diodes for the bridge and possibly slightly smaller resistor values on the filters - so it may end up being 12v transformer after all.
I wonder if its possible to effectively replace sziklai pair with a mosfet. Should be less voltage drop that way as well.

I appreciate all the help and advice I got here. Thank you.

ups. its 10k and not 1k resistor.
Title: Re: noisy linear ps design
Post by: DavidAlfa on September 16, 2023, 07:42:58 pm
Have you tried a swicthing regulator?
You won't hear 60KHz+ noise.
Title: Re: noisy linear ps design
Post by: look4awhile on September 16, 2023, 08:17:50 pm
Have you tried a swicthing regulator?
You won't hear 60KHz+ noise.

The ones I've tried were producing very audible hiss. I suspect general 'white' noise level was a tad too much. Originally I wanted to just use external PSU, like anything 12V 1-2A. But turns out most of them were noisy enough. The only one which worked well of the ones I have was Apple laptop one of all things. The article recommends using battery due to low noise (duh) or "Cannon K30120 13v 1.8A portable printer power supply" which is oddly specific. So I suspect they were hitting similar issues.

So I was like 'okay, i'll design one. how hard can it be", right? :)

I'll replace LM7812 with LM317 with a voltage follower if I every need it adjustable as well as more than 1.5A. It seems very low noise, but I have not measured yet.
Title: Re: noisy linear ps design
Post by: DavidAlfa on September 16, 2023, 09:12:52 pm
Is this the amplifier?? I'm not an expert but looks crappy to me.

(http://electronics-diy.com/schematics/1221/class-a-12au7-tube-headphone-amplifier-schematic.jpg)
Title: Re: noisy linear ps design
Post by: look4awhile on September 16, 2023, 11:29:30 pm
Is this the amplifier?? I'm not an expert but looks crappy to me.

Obviously I am no expert either. With minor changes it measures fine though. (0.5% THD, -72db noise, pretty linear frequency response).
My biggest issue with it is how sensitive it is to power supply :) But then again those giant 'chokes' on most tube amp psu circuits make me wonder.

Personally I find this amp funny if anything. It uses tube, but at 12v. Its using "class A" transistor amp as 2nd stage for current gain. Its overall silly and simple enough to make a good joke\gift to someone who is a little bit too much into audio.  :-DD
Title: Re: noisy linear ps design
Post by: look4awhile on September 17, 2023, 08:04:50 am
With schottky diodes without load 12V transformer works. Once I add the load voltage drops. 18V transformer still works like a charm, but its 16.5 volts on the input of the 7812, so it gets pretty hot. I might just call it good enough, or perhaps look for another in-between transformer.
With the current setup I estimate about 3.5W goes to heat. Its not ideal but should be fine with this heatsink.

Also I might increase R1\R2. It will make for better filter and lower the voltage on the regulator.

Mandatory picture attached.