3 term. regulators...something better??

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

Freq Band

Well-known member
GDIY Supporter
Joined
Jan 5, 2006
Messages
608
Location
Electra City
All of the PSU's I've built are using basic 3 terminal regs, 317/337, 79/79xx, etc.

I want to try to improve in this area.
(+15 - 17v territory) 1 amp or so.

Not that there's anything wrong with these psu circuits.....but I'd like to learn/experement a bit further with different (and tighter tolerance) PSU designs
Mostly for Headphone amps, Line Drivers, HiFi preamps....all op amp based.
--------------
Here is what I've come across so far. If you can add to these choices, or recommend one...:

**1970's National PSU design handbook (circuits for adding (dated) op amps along with 3term regs, to create dual tracking supplies.

**CA3130 and CA3140 datasheet schems.
http://www.intersil.com/data/fn/fn817.pdf
http://www.intersil.com/data/fn/fn957.pdf

**Sulzer, Borbely, Jung designs.
http://tangentsoft.net/elec/opamp-linreg.html

**Tweaking this to give +15v
http://www.djgardner.com/headphone/gilmore/dynahi/schematics/dynahi_psu_20041128c.gif

** ??


=FB=
 
I've built several projects using the Sulzer, and have gotten very good sound out of all of them. Haven't done an actual A/B against 317/337-based regs, so I can't definitively say it's better, but for sure the numbers are way better in all respects, and the sound certainly is good.

On a few occasions folks have built a Sulzer-design regulator on a circuit board I designed, and it has oscillated; the same reg on the same board didn't oscillate for me. I conclude that different batches of 5534s from different makers have different stability characteristics. With enough ceramic discs even the worst oscillators were tamed.

Haven't gone beyond the Sulzer to the Jung, so can't say about that, but the attention to detail in the Jung design is impressive.

Peace,
Paul
 
I guess you've seen the link at the bottom of one of the pages you linked to Jung's Site: www.waltjung.org/Regs.html

IMO linear PSUs have more important problems which one should solve before improving the regulators. I'd first think about snubbers for the rectifier (a simple 10 nF is not a proper solution) and reducing stray field of the mains transformer (e.g. by shielding or resistive peak current limiting). Next thing to consider is your PCB layout--it's not as uncritical as one might think. I've measured wast differences between almost identical circuits, probably due to bad grounding in the weak performers. Only once I got that right I'd start to worry about regulators.

Samuel
 
It's pretty hard to do a lot better than the 317/337. The beauty of these is that they are fast and stable and the temperature tracking due to everything being on one chip helps too. Don't forget the current overlaod feature also.
It seems almost everyone is using LDO, low dropout, regulators today, but most of the time we don't worry about that quality, since we are working with higher voltages to start with. You can make a mostly discreet regulator, and it will probably be almost as good as the packaged 317/337 chip, at least that's been my experience. Good luck! I have a book with hundreds of regulator designs in case you need some inspiration, and I'll be more than happy so send you some of the better one if you like, but you may have to substitute for parts no longer commonly available.
 
[quote author="Samuel Groner"]I'd first think about snubbers for the rectifier (a simple 10 nF is not a proper solution) [/quote]
What would be a good solution ?

Regards,

Peter
 
I see what many of you are saying.
.....a more complicated circuit means more areas of potential problems....and I'm sure I can still improve on the three term layout, I'm sure.*
One could say, " a thing is never mastered, only improved upon".
I needed to be reeled back in. After all, I could always start to build a more discrete regulator and experiment.....but for casing up a current project...stick to the working solution.

*I guess things like moving the regs closer to the circuit, learn proper snubberizing, and ground paths....(jeez, those could take years alone).
:cool:

What about dual bridges, and seperate regulators per opamp?

=FB=
 
It seems like we had a thread here about this several months ago. While you can roll your own that may slightly improve on one or two parameters, it's hard to beat the off the shelf modern 3-terminals for covering a lot of bases well.

Yes, any regulator circuit can be improved. You must decide if that's the best use of your time and money. Focus on the direct audio path will usually bear more fruit unless you have already identified a shortcoming in PS.

JR
 
What would be a good solution?
Oh, a simple RC snubber does the job very well--see e.g. snubber.pdf.

There are some very good application notes by www.linear.com on the topic. The use of one of their regulators would probably be the simplest solution for upgrading the LM317/LM337. Noise, precision and transient performance might be much better. Again, if it matters or not is another question.

Samuel
 
Good snubbing is highly recommended*, and look carefully as well at the peak charging current spikes that may be momentarily saturating your transformer, a not-uncommon occurrence especially with toroids. When this happens they radiate like crazy, their self-shielding properties null and void. These effects are dealt with in various ways, and also motivate putting the transformer as far away from the circuits as possible.

But then as well: make your layouts of the signal-processing circuitry itself less susceptible to magnetic radiation---the prime directive there being to minimize loop areas.

A lot of circuitry, particularly those using the better modern opamps, is fairly insensitive to PS rail noise per se. But some minimalist designs are not, and these may benefit from better regulation and lower noise. Whether the minimalism is worth pursuing to begin with is another subject, discussions of which usually generate more heat than light.


*Simple cap bypassing helps by lowering the resonant frequency of the reverse-recovery-spike-induced ringing, so it doesn't propagate as readily. It's often enough for EMC but is not optimal. RC snubbing actually damps the ringing as well. Soft-recovery diodes can be helpful, or if you can tolerate the voltage drop and the cost, the recent silicon carbide parts.
 
Back
Top