diyAudio

diyAudio (http://www.diyaudio.com/forums/)
-   Power Supplies (http://www.diyaudio.com/forums/power-supplies/)
-   -   oups! linear regulators of the same type may behave pretty different noisewise! (http://www.diyaudio.com/forums/power-supplies/56363-oups-linear-regulators-same-type-may-behave-pretty-different-noisewise.html)

Dominique 29th April 2005 10:22 PM

oups! linear regulators of the same type may behave pretty different noisewise!
 
Hi,

I wanted to listen to my phono preamp power supply for some time now! Finally I did and was pretty surprised:
(I connected the supply via a capacitor and mixer to headphones to have an idea of how that unwanted power supply ripple and noise would sound.)

Specs: +-12V supply with two full bridge rectifiers, then 470uF cap per channel, 7812/7912 regulators, finally 100uF per channel. I was drawing about 25mA per channel.

I couldn't hear any ripple after the regulators and adding any type of capacitor before the regulator didn't change anything.
But the negative channel was pretty noisy, much more than the positive! I was surprised, then searched for the datasheets and well, those Fairchild MC7912 are specified with 200uV noise while my 7812 is from Onsemi and has a specified noise of 10uV! What a difference!!
The negative regulators seem mostly more noisy than the positive equivalents, but still, the 7912 from Onsemi is rated at 75uV which is way lower than the Fairchild one!

On the fly I added several caps behind the noisy regulator and about 2000uF in all resulted in a nearly as low noise as with the 7812, while a 2.2uF polyester cap changed nothing. ( I had imagined that the high frequency noise would better be lessened by that one, but I couldn't notice anything).

After all, I learned to ask from which manufacturer the ICs I' am going to purchase are!

Next time i'm going to do the same with a 317/337. Here, the different manufacturer's datasheets all seem to have the same noise spec: 0,003% (which should be 360uV at 12V - pooh! Well, at least with the bypass cap at the adjust pin you can divide the noise by ten... err, or could you just increase ripple rejection that way?)

Hope this was of interest for anyone! :)
Cheers!
Dominique

PS. If you do the same: I added two schottky diodes in the line to the mixer between the signal and ground in that channel where I was going to add caps and changing the point of taking the voltage sample (wanted to listen to the ripple before the regulator too to have a reference) to cut out those nasty peaks I experienced when plugging something while listening through the headphones - Ugh! After doing so the peaks were tamed!

sek 30th April 2005 12:24 AM

Hi,

for a high quality linear supply that still uses monolithic regulators, the usual approach is to use two identical positive regulators with a dual-secondaries transformer, because negative regulators are known for their higher noise (or worse regulation at higher frequencies).

Additionally, the use of adjustable regulators still has additional potential to reduce noise/hiss while still keeping the regulators monolithic, because the fixed regulators have built-in gain set resistors that can compromise noise performance for most lower power supplies (selecting the most quiet resistor scale for a certain voltage divider ratio can reduce noise a little).

That said, 200uV is not bad for a fixed negative Fairchild chip. ;)
Cool that you did that survey. How did you select your dc blocking capacitors?

Cheers,
Sebastian.

hermanv 30th April 2005 12:38 AM

IC vendors
 
Each IC manufacturer has to design their own equivalent circuit for a given function the margins are just too small to allow for licensing. Besides some parameters are process dependent and they don't all use identical processes.

So if you have a critical application read the data sheet carfully. The same type number does NOT, as you have seen, get you the same device or performance:whazzat:.

Very true of Op-amps and linear devices in general.

For low power applications the shunt regulator type of device might well be quieter.

Dominique 30th April 2005 10:11 AM

Re: IC vendors
 
Quote:

Originally posted by sek
Hi,

for a high quality linear supply that still uses monolithic regulators, the usual approach is to use two identical positive regulators with a dual-secondaries transformer, because negative regulators are known for their higher noise (or worse regulation at higher frequencies).

...

How did you select your dc blocking capacitors?

Cheers,
Sebastian.

That's the next thing I wanted to post!
That PSU I tested was the first I built myself and all the schematics ( even for audio) I had seen then used positive and negative regulators.
But I already had thought about using two positive regs and wondered if that wouldn't imply any disadvantages. Because if if didn't I wonder why not all the audio supplies with two transformer secondaries already used that. In addition to the lower noise and better regulation (my 7812 has an output of 12,03V and the 7912 -12,3V) they're cheaper too. Maybe many designers are simply as lame as I am! :)
In fact there must be another 7812 lying around here so I should easily be able to modify my PSU.

Regarding the dc blocking cap:
The input impedance of my mixer is at least 10k Ohms (probably 47k), so I used 2,2uF polyester caps. (There's a formula to calculate the -3dB point depending on the capacitance and the input resistance of the next stage) I also use that one in my mic preamp.


Quote:

Originally posted by hermanv

For low power applications the shunt regulator type of device might well be quieter.

Ok, I'll get some info on this.

Thank you both for your comments and help!
Dominique

analog_sa 30th April 2005 12:16 PM

Quote:

Maybe many designers are simply as lame as I am!
Most commercial designers are really cheapskates and wouldn't want to pay the extra for extra four diodes and transformer with separate windings. Ok, it may be cents but it all adds up :)

jackinnj 30th April 2005 01:34 PM

2 Attachment(s)
when you measure regulator or power supply noise you should bandwidth limit taking the measurement over time -- this is one reason I hang onto an ancient Tektronix D13 storage scope, with a bandwidth adjustable differential amplifier -- (5A22N)
of course, which bandwidth ? -- Texas Instruments uses a 10 second sample from 100 milliHz to 10 Hz, Linear Tech uses 10Hz to 100kHz !
(I use 10Hz to 100kHz).

Here's the setup which Linear Tech uses -- (fwiw I use an "OREO" cookie tin, but the folks in California prefer Danish butter cookies):

http://www.tech-diy.com/LDO_Noise_Measurement.jpg

probing is important too -- using the ground clip of your scope probe is out of the question when you are measuring in the low microvolts . take a look at Linear Tech's Application Note 83 for more hints.

occasionally you can find very good noise measurement amplifiers on EBay -- there is an outboard differential amplifier from Tektronix, the ADA400A which will work with their last line of analog scopes (like the 2465). I was going to have some circuit boards cut for a noise measurement amplifier which would fit into the HP465A chasis -- haven't gotten around to it yet although the prototype works very nicely. This setup uses the SSM2019 -- a very low noise differential amplifier from Analog. I need someone to motivate me to have the boards done.


All times are GMT. The time now is 08:56 AM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright 1999-2014 diyAudio


Content Relevant URLs by vBSEO 3.3.2