LM317 - value for C-adj.

[Harpo]

Hallo out there,

maybe someone can proove this concerning the cap from adjust to ground to improve ripple-rejection at about 15 dB.

All LM317-datasheets I read (Vishay,National,Fairchild,SGS-Thomson)state this to be 10 uF as a type of constant, but all have a 240 Ohm resistor for R1 and receive power from a 60Hz line.

This R-series / C to ground seems to be a low-pass for 120Hz behind rectifier, set for 3dB at 60Hz (+12dB for 120Hz giving the desired 15db rejection).
For C = 1 / (2*pi * R * F) will get 11,05uF, this will be close to the stated 10uF (constant?).
In case you lower R1 to maybe 120 Ohm and have european 50Hz supply, this value would increase to 26,5uF for this 15dB loss at 100Hz.

Couldn't find this in the meta and have no instruments to proove myself.
I hope, I did'nt mix things up and better have asked for brain-adjustment.

All the best
Harpo

 

[PRR]

I think the cap bypasses the resistor to ground, which is about 3K, so 10uFd is plenty big enough for any power frequency and most reasonable values of resistor. It could probably be smaller in many cases, but cost would not drop very much.

I think the "15dB" is because, without the bypass, the regulator sees only a small fraction of the output voltage and ripple. That fraction is roughly 1.25V/Vout. For output voltage of 8.28V the fraction is 15dB, a little more at higher output voltages. AC-bypassing the resistor to ground makes the reference see the full AC ripple, and should give about 15dB improvement.

 

[Harpo]

Many thanks for your reply, PRR.

makes sense to me. I was only gathering, because Vishay's datasheet was giving specs of ripple-rejection at 120Hz.

I'll now get a screwdriver to adjust my brain for least nonsense.

All the best

Harpo

 

[PRR]

It may not be that simple. Look at the top of page 7 in National's data sheet http://cache.national.com/ds/LM/LM117.pdf

You may get a dB less ripple at 100Hz if the cap is bigger than 10uFd.

Mid-page, the Output Impedance with 10uFd is already rising at 120Hz, suggesting a bigger cap might give slightly better result.

Obviously if you scale the resistors lower, you need to scale the cap bigger.

But read the top of page 9. If you short the output, the ADJ cap discharges through an input, which could blow. The ADJ terminal is nominally protected for caps up to 10uFd, past that they want you to add diodes.

How much ripple rejection do you really need? You have 60dB anyway. If you need more, you may need LOTS more, which suggests some passive filtering or a pre-regulator. The few-dB you gain or lose with 100 versus 120Hz or cap tweaks is not a big deal.

 

[Harpo]

Thanks for your reply.
I'm just racking up a pair of TAB V372/1D (output-stage driven other than on kubi's site made by LH0002CN). I checked, wether I could use a 2x20V CT toroid laying around with a delon multiplier for 48Vdc - may give 54V at Reg.In after diode loss - not a big margine if line-voltage is lower than suggested and ripple become more important. I came thru JLM's site, who's 48V stage has a 120R and 10uF cap at 60Hz line-freq downunder and thought, if there might be an improvement in increasing cap-size to 22uF. I guess, I'll be better at a 2x22 or 2x24V CT toroid for this purpose.