Best IC voltage regulator....

DanK

Member
2001-08-30 5:05 pm
I have used the LT1085/1033's in the past,
and they seem to be well regarded (even though
they are an aging design).

Are there better performing (for audio) single
chip regulators, or is the only other option
(which I am desperately trying to avoid) to
build a discrete regulator??

Thanks in advance,


Dan
 
Better than the 78xx and 79xx regulators are the LM317T/K and LM337T/K three pin variable regulators. They can be used upto 37 volts input (HV version is 50v), deliver 1.5 amps and can be used in conjuction with transistors to boost their current output. Great for preamps and even input stages of power amplifiers. They are marginally costlier than the 78/79 types.
 

vdi_nenna

Member
Paid Member
2000-10-10 7:27 pm
PA, USA
Linear Technologies

Linear Technologies are suppose to be the undisputed champs, but I can't remember the designation...I think LT1085 is one of the two. You can get them from Digi-Key. They cost more, but are worth it.

They were used by Corie Greenberg back when he reveiwed for Stereophile mag. He put together a buffered-passive preamp that's suppose to be real nice. (The issue also has an interview w/ Nelson Pass). A verion of it it sold in kit form from Micheal Percy. If you download his (.pdf) catalog, you can also find the LT IC's. No web site...search on his name.

I believe the power supply was created by Walt Jung, he's really good w/ IC circuits and regulators.

Vince

Man, I just re-read your post, I feel silly, sorry. :eek:


[Edited by vdi_nenna on 09-05-2001 at 02:51 PM]
 
I've recently been considering power supplies much more carefully. I've come to the conclusion that mindlessly tossing the "best" IC regulator into your circuit is not always the best option. For instance, the cheap and slow 78xx and 79xx series regs may offer a certain measure of isolation to high frequency noise...

The main advantage of the expensive regs their better line/load regulation. So, if you need a precise DC voltage or a low impedence power source, they're good for that. Adjustable regs also provide a lower noise output than fixed voltage types. But, depending on your circuit's PSRR, you may not need the line/load regulation, and would be better off to focus on other aspects like providing a higher impedence between various sub-circuits to prevent interactions via the power supply. Remember: low impedence = good signal path.

The more I thought about it, the more attractive various methods of passive regulation became... not only because you can acheive a lower noise supply with passive components, but there are fewer frequency-related problems and a passive solution is also generally cheaper than complex active regulation schemes. As well, I think a large number of problems occur at higher frequencies where circuits can't adjust quickly enough to remain unaffected by the disturbance. Active regulators are just as incapable of effectively dealing with these frequencies, but passive schemes can work very, very well.

One device i'm using a lot more since I started doing lots of digital work is the ferrite bead. These little guys have basically zero impedence at DC, but will effectively block most of the high frequency junk which can cause insidious problems. Just be careful not to put too much current through them, or they will saturate. Another useful power supply device is the good 'ol pi-filter (remember your tube designs?). You don't need an inductor in there, just a resistor will do. Tack a cheapo fixed reg on there, and you've got a great supply (or preregulator).

Anyhow, when I need a good, stable DC voltage, I fish a LM317/LM337 out of the parts bin. They are 95% as good as the expensive LTs, which I no longer buy. If I'm really concerned about the performance, I'll invest the money I saved in better passive parts like larger capacitors, and I sleep soundly in the knowledge that I have a better power supply for less money...
 

ALW

Member
2001-06-12 10:15 am
UK
Reg's

Another vote for 317 / 337, I've never had these oscillate in any circuit, it would have to be very badly designed for this to happen!

The BIG advantage of 317 / 337 over fixed 78xx and 79xx series is the significantly increased ripple rejection, providing the reference voltage is decoupled (see data sheet).

They're also generally lower noise, although there is a reasonable variation between samples. Look at Linear Tech Application note 83 (AN83-1) for a scheme for testing regulator noise.

The LT parts should offer lower noise, but all of the ultra low noise parts only work up to 20V - bit of a pain when I want 24V.

Andy.
 
LT versus LM

Has somebody ever compared the LM317 (ST) against the LT317 (Linear Technology) soundwise?.
The LTs are afaik higher rated, but maybe also more prone to a tendency of oscillation. I personally never had any troubles. I have got mine quite a time ago as spare parts from Exposure, but I do not know, if they are available any more.

greetings
Klaus
 

ALW

Member
2001-06-12 10:15 am
UK
Good lead

Klaus,

I've just checked the data sheet spec for the LT317 - it has a significantly lower noise spec. than the LM parts. I could kick myself for not realising this before.

I need to get some now - I'll let you know if I find a source.

Andy.

P.S. The noise spec is also lower than LT1086 etc.
 
C Simpson, in addition to the ferrite beads you mentioned I had very good results replacing the decoupling resistors feeding a commercial DAC with small inductors having the same DC resistance. These are readily available and will not affect the DC conditions intended by the original designer. However the reduction in “trash” on the power supply was remarkable.

Cheers,

Pete
 
hey, no prob.

The only thing you'll want to watch for with inductors is that you don't form an L-C resonant circuit. I like to put something resistive in there which will dissipate unwanted energy...

What's so nice about ferrites is that their impedance is nil at DC, but can go up to 90 ohms or more at the high frequencies of interest. And, ferrite beads are very lossy, so that impedence will be almost purely resistive. In combination with a cap, this is enough to effectively block anything in the 10MHz+ range, depending of course on the bead size and composition.
 
I have used all these types of IC regulators and what i like to do with them is use the 78XXX and the 79XXX as Preregulators. then follow these with the Three termanal adjustable Type Like The LM-317 and the LM337 or the LT-1085 and LT-1083. types. I have noticed that most Circuits showing the adjustable regulator Applications are just taken right out of the data sheet. While this works it is not the way to get the Best performance out of them. First off replace the resistor fron the Adj. term to ground with a Zener diode or a voltage reference of your Choice. Next replace the Resistor from the Output to the Adj termanal with a current source to drive the Voltage reference. On the Negitive regulators Place a small resistor about 1-2 ohms in series with the Input termanal and the power supply. make sure the Bypass cap on the regulators input is still directly connected to the reg's input Pin. Ferrite beeds are on the output of the Preregulators as well as the Input of the Preregulators. The concept hear is to get all the bandwidth and speed you can out of the Adjustible reg's in Addition to improving the Rejection of the regulators with the current source. Note the Adj. termanal is at a high Impedance so using a voltage reference will Provide this Termanal with a low impedance to ground and using the Current source will provide High Impedance isolation from the Output of the regulator. The regulators output can be quite noisy because of the Transients the Load places on it and you dont want this to get feed into the Adjust termanal, The low value resistor recomended in the data sheet allows this to happen. The ferite Beeds are as stated in an above post to block Line Noise. If the Best performance obtainable is what you want then the Classic three termanal regulator is not going to provide that. Then eather a discreet component regulator or a shunt type made from a Voltage reference and a Current source is what is needed. Beware most people take the circuit for the shunt regulator right out of the data sheet and use a resistor to Drop the voltage this can be improved alot by using a current source in place of the resistor.The resistor i am refering to is not the voltage devider used to set the Output Voltage but the resistor in series with the Powersupply and the shunt regulator. Regarding the voltage devider, a Voltage reference and current source can replace this as was done in the Three termanal Adjustable regulator i talked about above.
 

ALW

Member
2001-06-12 10:15 am
UK
For Lohk

Re:LT317AT's

I managed to secure some smaples of these, and they measure better than the NS LM317AT's I'd been using up until now.

Unfortunately I can't log into my website at present, but have a JPG of the FFT analyses I've done if your interested. Let me have an email and I'll send it.

I'd try and get some to play with, they're good.

Andy.
 
Andy,

from where did you get these ? - I recently tried to find new ones, but in vain...
Where can I find a data spec ?

Klaus


To all:

Is there a significant difference (in resultant sound quality) between the positive and the negative regulator ?
Have you already tried f.e. LM337T and reverse the poles for a single side supply (so that the "ground" is reguated and the positive is connected to the rectifier/transformer) ?