LM317+TL431, really? - Page 3 - diyAudio
Go Back   Home > Forums > Amplifiers > Power Supplies

Please consider donating to help us continue to serve you.

Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
Reply
 
Thread Tools Search this Thread
Old 2nd January 2014, 07:47 PM   #21
diyAudio Member
 
Join Date: Oct 2004
Location: Denmark
Quote:
Originally Posted by martin clark View Post
Not to hand, but Christer's figures looked pretty good to me then and, checking, now too. Even if the figures given were a factor of 10x too small, the zener's noise will be utterly swamped by the internal noise of the 317 ... ~150uV output noise even in 'lowest' in 1.25v mode.

Easy to built up a reg and test it - have a go. You might be surprised.
LM317 output noise is typically 0.003% of the output voltage, however if you use a bypass cap on the adjust leg the output noise is reduced to 0.003% of the 1.25V reference, 0.003% of 1.25V is 37.5uV RMS output noise.

Last edited by Neutrality; 2nd January 2014 at 07:51 PM.
  Reply With Quote
Old 2nd January 2014, 08:00 PM   #22
diyAudio Member
 
Join Date: Jul 2003
Location: Bath, UK
Oh call it 3-5dB difference if you must; and I think the point of DIY is not to believe all datasheets but to actually build and to test

Last edited by martin clark; 2nd January 2014 at 08:04 PM.
  Reply With Quote
Old 2nd January 2014, 09:00 PM   #23
diyAudio Member
 
Join Date: Jun 2009
Quote:
Originally Posted by Neutrality View Post
Great answer, just what I was looking for.
You're welcome, thanks. Actually, compensation's not hard. It's just compensating to get a regulation bandwidth of more than 2kHz or so (depending on acceptable peaking) that's tricky. Most of the TL431's performance is getting thrown away but I don't know of any faster pass device implementation which can hit the 317's feature set at comparable price and board area. Chip amps are probably the cheapest 1MHz+ option with current limiting and thermal protection but clipping voltage/dropout, bias current, and higher current limits are all issues.

Quote:
Originally Posted by martin clark View Post
With the usual resistive divider, adding a bypass cap from Vadj to 0v improves performance not by doing 'something' to the LM317 or adding gain or anything like that. It is simply making this external voltage reference quieter - by reduced AC impedance.
Careful with this. Lowering the cutoff of an RC lowpass increases the amount of Johnson and excess noise absorbed by the cap due to the cap's lower impedance. But the reason 317 ripple rejection improves is the lower cutoff reduces the amount of line and load regulation error reaching the reference voltage. So in this case it happens increased passive attenuation is equivalent to increased error amp open loop gain.

Quote:
Originally Posted by tomchr View Post
The LM317 is a nice regulator, however, considerable improvements have been made since its introduction in the late 1970ies. If you don't mind spending a bit more, I suggest looking at the LT3080.
True; the differences between 1.1 and 1.5A and 36 and 40V aren't usually significant. I see this thread as more about an interesting point on the price-performance frontier---LM317+TL431 is under $1 versus $4.40 for an LT3080 (all quantity 1). In particular, TL431 pricing starts at 31 cents in SOT23 so it's at least notionally cheaper, smaller, and more performant to use a 431 to stabilize the reference at low frequency than to follow the traditional solution of using a large value cap between Vref and ground.

The 431 bias current used so far in this thread is a bit high (10mA) and, while it can be reduced to a few hundred uA, that's still rather higher than the 317's adj pin current. Traditional 317 feedback networks typically pull several mA so this probably isn't too significant, however.

Last edited by twest820; 2nd January 2014 at 09:07 PM.
  Reply With Quote
Old 2nd January 2014, 09:05 PM   #24
diyAudio Member
 
Join Date: Jul 2003
Location: Bath, UK
Quote:
Originally Posted by twest820 View Post
Lowering the cutoff of an RC lowpass increases the amount of Johnson and excess noise absorbed by the cap due to the cap's lower impedance. But the reason 317 ripple rejection improves is the lower cutoff reduces the amount of line and load regulation error reaching the reference voltage. So in this case it happens increased passive attenuation is equivalent to increased error amp open loop gain.
N, and it's not remotely a problem because the 317s internal gain and compensation is unaffected. Essentially it returns the performance at higher voltages to that of the basic 1.25v Vreg it is.

The better way is of course to use something that properly drops the DC impedance too at this point. Like a zener, or the TL431, or one of Fred's circuits.
  Reply With Quote
Old 3rd January 2014, 02:16 PM   #25
diyAudio Member
 
Join Date: May 2011
Location: Silicon Valley
Quote:
Originally Posted by tomchr View Post
I suggest looking at the LT3080. I ... have found LT's claims of 120 dB ripple rejection to hold up in real life!
I only see 75 dB ripple rejection in the datasheet you linked ... ???
Attached Images
File Type: png datasheet_LT3080.png (76.4 KB, 511 views)
  Reply With Quote
Old 4th January 2014, 07:36 PM   #26
SGK is online now SGK
diyAudio Member
 
Join Date: Dec 2012
Location: London
I'm interested in this thread as I have a modified 1A Velleman PSU based on the LM317 which I am looking to upgrade as a little learning project. I use it to power an SSD in my music server from regulated 12V input. The PSU currently has a 10uF cap on the adjustment pin as shown in the LM317 data sheet (although I note it doesn't have the diode to discharge this cap).

I take it that using a TL431 will yield a better result than the 10uF cap configuration? Am I right to summarise that the conclusion is that shifting the TL431 cathode resistor (R3 in the original post) to the output of the LM317 increases complexity significantly because of the required compensation and so I should stick to the input configuration? Is D6 necessary? Lastly, is there any benefit to using an Onsemi NCP431 which I believe is an upgrade to the TL431? (I have searched for TL431 with ED suffix to no avail so far having found just A and B versions.)

(I will have to figure out how to make the mods given the existing PCB traces but I guess that shouldn't be too hard.)
  Reply With Quote
Old 4th January 2014, 08:07 PM   #27
diyAudio Member
 
Join Date: May 2011
Location: Silicon Valley
Quote:
Originally Posted by SGK View Post
I take it that using a TL431 will yield a better result than the 10uF cap configuration?
One way to analyze this is to plot impedance vs. frequency, for a TL431 and also for a 10uF capacitor, on the same graph. If one of the two curves is lower than the other at all frequencies, it's a very easy decision. If not, (i.e. if the curves cross each other), then there's probably room for subjective human judgment to decide "which of these do I like best"?
  Reply With Quote
Old 4th January 2014, 08:08 PM   #28
diyAudio Member
 
Join Date: Oct 2004
Location: Denmark
Quote:
Originally Posted by SGK View Post
I'm interested in this thread as I have a modified 1A Velleman PSU based on the LM317 which I am looking to upgrade as a little learning project. I use it to power an SSD in my music server from regulated 12V input. The PSU currently has a 10uF cap on the adjustment pin as shown in the LM317 data sheet (although I note it doesn't have the diode to discharge this cap).

I take it that using a TL431 will yield a better result than the 10uF cap configuration? Am I right to summarise that the conclusion is that shifting the TL431 cathode resistor (R3 in the original post) to the output of the LM317 increases complexity significantly because of the required compensation and so I should stick to the input configuration? Is D6 necessary? Lastly, is there any benefit to using an Onsemi NCP431 which I believe is an upgrade to the TL431? (I have searched for TL431 with ED suffix to no avail so far having found just A and B versions.)

(I will have to figure out how to make the mods given the existing PCB traces but I guess that shouldn't be too hard.)
I would stick with the input version, much simpler and safer to use.

But put a cap between Cathode and Vref, 1-10uF is a good value, it helps with noise, especially low frequency noise.

TL431ED is the model I have used in my simulation, it is an improved version that follows the TI TL431 datasheet better than other available models.

You could go with the Exar SPX2431AM, it is similar to the TL431 but with significantly reduced noise compared to the standard TL431

Standard TL431 is around 120nV/hZ at 1K and the SPX2431AM is 40nV/hZ at 1K.

The NCP431 looks to have around 600nV/hZ at 1K according to the dataheet, that is downright terrible. Surely that must be an error?

D6 is there in case you short the input to the regulator, if that happens the output capacitor will discharge into to output of the regulator, potentially destroying it.

Last edited by Neutrality; 4th January 2014 at 08:32 PM.
  Reply With Quote
Old 4th January 2014, 08:40 PM   #29
SGK is online now SGK
diyAudio Member
 
Join Date: Dec 2012
Location: London
Quote:
Originally Posted by Neutrality View Post
I would stick with the input version, much simpler and safer to use.

But put a cap between Cathode and Vref, 1-10uF is a good value....

I would go you could go with the Exar SPX2431AM....

D6 is there in case you short the input to the regulator...
Thanks. By chance have you modeled this versus the LM317 + 10uF cap to ground config? (A cheating response to Mark's comment else I would to wait to get home to London, find a model for the TL431 and try out my Spice modeling again.)

Just to clarify the 2nd point (I'm a newbie to this stuff) you mean a cap between cathode and the LM317 adj pin, right?

Thx

Another thing missing on the Velleman module... Btw it started life with a diode rectifier. I removed 3 of the diodes and jumpered one to ground because I have DC input leaving just one LN4007 diode in series with the +ve input. I understand this diode is limited to 1A while the Lm317 can handle 1.5A or more and so the diode is a bottleneck. Can you recommend a better replacement for this? (I'm sure an SSD doesn't use much current but I may want to power two.)
  Reply With Quote
Old 4th January 2014, 08:47 PM   #30
tomchr is offline tomchr  United States
diyAudio Member
 
tomchr's Avatar
 
Join Date: Feb 2009
Location: Greater Seattle Area
The 1N540x-series handles 3 A. 1N5408 is the 1 kV, 3 A version. It comes in a DO-201AD package which is quite a bit bigger than the package used for a 1N400x. You can usually make it work if space isn't too tight, though.

~Tom
__________________
Modulus-86: Composite amp achieving 0.00018 % THD. Damn Good 300B, Novar Spud, 21st Century Maida Reg., Filament Reg., etc.
Neurochrome : : Audio - http://www.neurochrome.com/audio - Engineering : : Done : : Right
  Reply With Quote

Reply


Hide this!Advertise here!
Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off


Similar Threads
Thread Thread Starter Forum Replies Last Post
What's really going on in that TL431 jackinnj Power Supplies 6 1st December 2012 11:53 AM
TL431 comparison borges Power Supplies 6 27th September 2010 01:05 PM
help understand the TL431 jarthel Parts 6 1st October 2006 10:09 AM
Tl431 sparkle Digital Source 11 22nd February 2004 10:10 PM
TL431 sucks? Bas Horneman Parts 2 4th November 2003 12:09 PM


New To Site? Need Help?

All times are GMT. The time now is 01:18 PM.


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