LM317 CCS in DC Power supply - Does HF performance matter?

[skibum]

A couple weeks or so ago I had some questions about a CCS for a DC power supply.

Janneman helped me out an I am currently using a LM317 for a CCS in my DC Power supply.

One of the things I have been reading (in this forum and others), is that a CCS using a LM317 (or any LMxxx), is "not very good at high frequency".

If the LM317 is being used as a CCS in a DC power supply does this matter?

I am using a LM317 for a CCS in a DC Power supply for a DAC. The DAC has oscillators at 20+ MHz and BCK frequencies above 10MHz, ect. Plenty of high frequency stuff.

So I am wondering if it matters or not since in this case the LM317 is supplying DC current in a DC power supply?

Thanks!

[jan.didden]

You again!

Regulators 101: ALL regulators are filters (just like all amps are filters), meaning they all lose performance above their roll-off frequency. The 317, or any other regulator, integrated or discrete, is no exception. So, the hf performance of ALL regulators deteriorates at hf.
That means that ripple rejection, output Z etc gets less at higher frequencies.
That's the reason why you see all these graphs bending upwards (or downwards, depending on what the data is) from a certain frequency.

The bummer is that, because all amps are filters too, the Power Supply Rejection Ratio of amplifiers also deteriorates above their Fc. So you get hit doubly: the supply gets worse, and the amp's ability to reject supply junk also gets worse.

There's no sure-fire solution to this, except that with smart circuit design you try to juggle supply performance, turn-over frequency and amplifier PSRR to get the best possible performance as a system.
This is the area where the good designer earns his next installment on the Lexus

jan

[skibum]

Yep, it's me again!

They LM317 worked great as far as consistent current output, but it just did not sound as good.

I tried using a cascode DN2540 with the LM317 that I saw in Walter Jungs CCS articles. The problem I had with that is that my supply voltage is 9.5 volts. Walters circuits were all done using 18 volts. I could only get about 1.9 volts maximum for the (Vo-Vin) of the LM317 with the DN2540. The minimum Vo-Vin for the LM317 is about 3 volts.

Even with this low Vo-Vin it sounded much better than just the LM317 on its own.

I have been reading about current limiting diodes and might try to feed the LM317 with one of those. According to the data sheet they have a pretty high impedance up to at least 200KHz.

Any other ideas would of course be welcome!

[AndrewT]

Performance of any regulator is very different for a continuous non varying current demand cf. a step varying current demand.

Any step in current demand whether from 0% to 100%, or from 100% to 0%, or from 53% to 55% is effectively an HF regulation problem looking for an HF solution.