Improving the LM3x7 regulator circuit

AX tech editor
Joined 2002
Paid Member
Since I am waiting out a blizzard -- was looking at this thread and had the TAA 2/95 article sitting in my lap:
[snipped pics]

Jack,

Did you look at lowering the series R in the bypass leg below 4 ohms?
Its only there for damping the resonance so 1 or 2 ohms might do that enough while lowering the hf Zout.
At any rate this should be verified/tuned in real life since track L and other parasitics obviously have an impact here.

jan didden
 
Just another Moderator
Joined 2003
Paid Member
Since I am waiting out a blizzard -- was looking at this thread and had the TAA 2/95 article sitting in my lap:

I spent an inordinate amount of time simulating LM317 mods based on the post 34 in this thread... the results are in my blog under YARPS if interested. Note I still haven't actually ordered all the parts to build it yet though :(

The transistor really does seem to make a difference (in the sim at least) worth giving that a try as well I think!

138d1270183005-yarps-yet-another-regulated-power-supply-updated-03-04-2010-impedance_compare1.png


The graph was the comparison of the output impeadance between standard datasheet implementation (blue) and with fred's transistor mod (green). change mV to milli ohms and you have it :) Note 0.1 ohms seems to have been enough to damp the resonance in the sim, whether it will be enough in real life however waits to be seen (that is with a 1000uF electro as the output cap however not a small film cap)..... edit again, just saw Jan's comment about track inductance, probably cutting it a bit fine with .1 ohms in addition to the caps inherent ESR.... I might add some slightly higher value resistors to my (forever) pending order...

edit: ripple comparison once again blue standard, green with transistor. :)
137d1270183005-yarps-yet-another-regulated-power-supply-updated-03-04-2010-ripple_compare.png


Tony.
 
Last edited:
Just another Moderator
Joined 2003
Paid Member
Neither have I! I've built it now on prototype borad. I'm using a 10K resistor and 50K pot which allows me to adjust over a reasonable range (but my target voltage is 10V)

You can download the spice circuit from my blog entry http://www.diyaudio.com/forums/blog...egulated-power-supply-updated-2011-02-10.html . That should at least allow you to get some ballpark figures, but if you want to go with discrete resistors I'd suggest using a pot to start with find the value and then substitute in the correct value resistor, I wouldn't rely 100% on the sim for accuracy of the voltage.. Note that the cap values are way overkill ;)

Another thing I've noticed, if the voltage is critical this modification might not be ideal, as it does drift a bit with temperature, from cold to warm it might vary by a few hundred mV. This might be better when properly laid out.

Also for some measurements and comparisons (taking in to account I have a problem with 50Hz noise pickup in my measuring) have a look at my LM317 measurements and experiments thread :)

Tony.
 
Last edited:
I had seen your thread!

Only that it seemed that you didn't get to actual conclusions, as the test circuit seems to be on a protoboard.

I've been gathering some information on regulators, as I will use two for my next power amplifier. If I don't get a better type, with lower wide-bandwidth impedance and noise, I will rely on the old 317, which is cheap, reliable and everybody uses it.

There are some tricks here:

Simple Voltage Regulators Part 2: Output Impedance

and here:

Using 3-pin regulators off-piste: part 1

They show how to lower impedance with several tricks. Particularly useful are:

1) Forcing the 317 to work at more than 100mA, by loading it with the proper output resistor.

2) Bypassing the adjustment terminal with a medium size capacitor, 220uF seems ideal.

4) Not using low-ESR capacitors as ouput bypass. Suggested ESR is not lower than 100mOhms, or there would be a resonant peak. Go higher than 100mOhms and there won't be proper impedance lowering at high frequencies. What I wonder is which capacitor brand and type we should use?...
 
Just another Moderator
Joined 2003
Paid Member
yes It stalled when I realised that the 50Hz interference was messing things up, I was doing the measurements on protoboard to allow me to easily swap components but from the feedback I got some of the measurements could be quite different if properly laid out and soldered. Still have to fix that 50 Hz problem. I was going to buy a die-cast box yesterday but ended up spending the entire day (more than 12 hours) recovering photo's off my father-in-laws pc that was wiped by a virus :(

1) I think Sy posted a great link about this in my measurements thread.
2) This seems (based on the measurements I've done so far on the protoboard) to make little difference with the modified circuit, it does make a much bigger difference on the standard circuit.
4) You could use the trick I am. I was originally going to use 100mOhms but I think the sims showed that I potentially needed more than that so I went with 0.33 ohm resistors in series with the output cap. I can always parallel some to lower the resistance if 330 mohms seems too much. I chose SMD resistors for this, the currents are tiny so they don't need to be able to handle much power.

Tony.
 
You could use the trick I am. I was originally going to use 100mOhms but I think the sims showed that I potentially needed more than that so I went with 0.33 ohm resistors in series with the output cap. I can always parallel some to lower the resistance if 330 mohms seems too much. I chose SMD resistors for this, the currents are tiny so they don't need to be able to handle much power.

I think we are talking about different things.

I didn't mean the resistor in series with the output capacitor. I mean using capacitors which are not low ESR types.

Maybe one option would be to use a low ESR capacitor and put a 0.33ohm resistor in series to increase the ESR.
 
Just another Moderator
Joined 2003
Paid Member
yes that is the point :) you put the low ohms resistor in to effectively increase the ESR of the cap. This damps the resonance. That way you can use low esr caps without worrying. I know it seems a bit silly, but generally high esr caps are not necessarily the best quality. This way you can use a low ESR good quality cap and damp the resonance that would otherwise occur, you get the best of both worlds :)

Tony.
 
The improvements stand. And there is no problem to using a 3x7 at voltages above 37v in the datasheet. That '37v' is maximum differential between input and output voltages allowable (subject to thermal constraints)

The trick is ensuring the LM317/337 is not exposed to more than about 30v differential from IN to OUT during startup and normal running.

The easiest way to do this is to use a large zener diode , say 15 or 18v , min 1.3W or 5w rating preferrd, in place of th reverse-bias diode between IN and OUT pins. Then, when the PSU starts-up, the zener avalanches until this voltage differential diminshes to something the regulator can withstand. At swtch-off, it behaves like the recommended diode. The zener only needs a voltage rating of a volt or two greater than the expected input-output differential so that it is 'out of circuit' during normal operation.

If you need to stand-off very high voltages the extended datasheet contains several suggestions - up to 2000V!
 
Hi
Yes it is.
intermediate version is stable curves (bond will not be published)
and bottom of today (srb-Snoise) :)
 

Attachments

  • supershunt tinadiagRB.jpg
    supershunt tinadiagRB.jpg
    61.8 KB · Views: 959
  • supershunttinadiagNoise.jpg
    supershunttinadiagNoise.jpg
    40.8 KB · Views: 324
  • s-rb.jpg
    s-rb.jpg
    56.2 KB · Views: 196
  • snoise.jpg
    snoise.jpg
    55.5 KB · Views: 186
Hi
Yes it is.
intermediate version is stable curves (bond will not be published)
and bottom of today (srb-Snoise) :)


Are these curves for real or simulations?

The impedance curve gets to improve on the Jung Super Regulators?

They were 100 microohms at a 10K frequency:

http://waltjung.org/PDFs/Regs_for_High_Perf_Audio_2_B.pdf

How can that be possible without the massive feedback in Jung's regulator?