LM317 and LM337 regulators

[regiregi22]

Hello guys

I am struggling myself with a PSU design I'm implementing. It will have a positive rail, negative rail and gorund. Nothing new. Just a plain LM317 implementation.
But I was wondering about the regulator selection for both positive and negative rail. The positive is a no brainer, LM317. But for the negative rail I have noticed to have two options:

-I can use an LM337, wich is designed to regulate a negative rail. And implement it just like the LM317 for the positive. After joining both grounds, you're done. Like this one:
http://www.massmind.org/images/www/h...if/ckt21_1.gif

or

-I can use another LM317 for the negative regulator. I could achieve that by joining the ground from the positive rail with the out from the negative rail. Then, the negative rail ground will be the negative out. Like this example:
http://myweb.tiscali.co.uk/nuukspot/...ub.reg.psu.png



Is there any difference between both topologies? Because, if there isn't, why does exist a negative version of the regulator, if in any case you could get away with an LM317 in almost an situation?

After looking extensively through the web, I haven't found any argument to found my choice.
Thanks for all, guys

[AndrewT]

Some regulators perform better in either the -ve version or the +ve version.
A centre tapped transformer can use the +ve & -ve versions to create a dual polarity supply.

If you want to use twin similar regulators for dual polarity you should use a dual secondary transformer.

[Speedskater]

In 2007 Walt Jung wrote in one of his AudioXpress magazine articles:

audioXpress - Articles and Addenda

That using two LM317 regulators might be better.

You might have to do some searching as this was not the main topic of the articles.

[woody]

Walter Jung did a power supply regulator aimed at the Pass A40 back in 1978 or 79.
In this he did touch on why it was best to use a + and a - regulator instead of using
2 positive regulators. My memory isn't all that good but I think he discussed that
using a positive regulator for the negative rail would work but the transient responce
suffers. Perhaps he said more but I don't rember anything else. That article was in The Audio Amateur which has changed it's name to AudioXpress.

[AndrewT]

most of Walt J's papers are on his website.

[Speedskater]

From Walt J's article (jung2779.pdf):

A detail worth noting at this point: If complementary source and sink circuits are needed for an application, it is actually better performance-wise to use a pair of LM317s, than it would be to use an LM317 and an LM337.

[Gordy]

When building distributed systems (maybe not audio) it is often necessary to have second (/third, etc.) sets of positive / negative regulators distant from the main supply regulators, and for this it is easy to use a simple pair of (positive LM317 and negative LM337) regulators.

Also, some designs need just a negative supply, hence it is logical to just have a simple LM337.

When building a simple +/- supply it would seem rational to consider the 2 x LM317 / 'regulated ground' solution. However fault finding on a standard LM317 / LM337 pair would be easier.



[Yes, I know it's not a true regulated ground, but I didn't know the right name for it! I'm just a simple country boy...]

[wintermute]

I've been going through the same exercise I put up a blog post with my design last night (only simulated at this stage). It is a dual rail +- 10V supply using two LM317's (I'm going to use LT317A's as on paper they look better than a standard LM317).

Tony.

[mikesnowdon]

Ho Tony.

Is this PSU for something audio related? (sorry, haven't looked at your blog post yet..)

If so then have a read of this from Martin Clark:

Inveterate meddling #377: Quick notes on using 3-Pin Voltage regulators

[wintermute]

Hi Mike, yes it is for audio. I'm going to use it for my active crossover. Very nice article, I hadn't seen it, thanks . My circuit is based on one that Fred DieckMann posted, I suspect that the transistor is actually having the same effect as the zenner in the article you posted. Certainly the ripple rejection is improved by a fair amount. I was originally thinking of using an LM329 as per the article, since there was an implementation in the datasheet which gave 10V (what I'm after) with that config, but after I came across Fred's circuit I decided to give it a try, and certainly in the sim it seems to be an improvement over the standard circuit.

It doesn't do anything for the impedance rise with freq (but does help at low freqencies a bit), but since in the sim, the max output impedance is 0.11 ohms at about 1Mhz I'm not that fussed... it's close to 0 ohms at 100Hz where the primary ripple will be so I think it should be ok.

The main disadvantage of my circuit (and this could just be because I don't understand) is that I can't work out a formula to set the output voltage. The standard one doesn't work because of the changes to the circuit, and I just had to adjust values till I got something that gave me the voltage I wanted. I could put a pot in for the lower resistor in the divider I guess and just have it be adjustable

Tony.