Regulator output impedance

[jesper]

Hi,

How low does the regulator output impedance need to be? I know it’s a tricky question which depends on what you’re feeding, but say I have an op-amp based circuit which class AB outputs. There’s a large amount of op’s so I’d rate the current demand as fairly dynamic.

I’ve been lurking around the internet for a good PSU and the Jung Super regulator is interesting, but right now my gut feeling tells me to look at something simpler without feedback. I’m thinking of a simple series transistor regulator with filtered zener reference followed by a large capacitor bank to lower the output impedance. Before the regulator there is a shottky diode rectifier bridge and a RCRC combination (or maybe LCRC) to filter out line ripple and noise.
The output impedance of the regulator itself is fairly high (some 500mohm) but the output capacitor bank lowers this significantly from some 500hz and up depending on the size of the capacitor bank.

But what about frequencies below 500hz? The PSRR of the op’s is fairly good in this area (min 75dB), but is it good enough.
Does anyone have any real experience?

/Jesper

[hjelm]

This is line level opamps?
I guess you can estimate the required supply impedance if you know how much voltage error you can cope with on the output of the opamp. Count backwards with the PSRR of the opamp and you will know how low inpedance you have to have at the given frequency.

[jesper]

Hi,
Unfortunately I have lost my simulationfiles for the preamp and active filter, so I can't easily estimate the dynamic current consumption.

But let's do a quick calculation, say each opamp drives 10k, with a worstcase swing of 2V. This gives 200uA of dynamic current for each opamp. There's 12 opamps in each filter, and 4 in the preamp, per channel that is, giving a total of 32 opamp, adding some extra for furter stuff so lets say 40.
Giving a total of 8mA, say 10mA. 1ohm output impedanse gives 10mV of PSU ripple. With 75dB PSRR this translates to about 2uV.
LSB in 16bit data with 2V swing is about 30uV so I guess it's ok.

/Jesper

[jackinnj]

Refreshing this -- Janneman showed that the remote sensing of the Jung-Didden regulator facilitated much lower output impedance. Well, I've just ordered some more of the J-D boards from Old Colony, and I have some of Peranders as well. I thought it would be interesting to compare the output impedance to the new "ultra-quiet" TPS7A4901 regulator from Texas Instruments:



The chart depicts the a.c. voltage across the load for 9.7mA a.c. current. DC load is approximately 150mA (which is about the limit of the TI regulator.)

Note that the SLA battery is much higher impedance than any of the regulators.

This test does not employ the remote sensing which is available on PerAnders SJO regulator, or the Jung Didden. The results seem to be consistent with Janneman's 3/95 Audio Amateur article. Remote sensing is next. (Remote sensing isn't available on the Invisus regulators.)

This is the setup I used, with apologies to Fred:

[Pano]

Cool stuff, Jack. Thanks.
Interesting to see that the Invisus and SJO are basically the same. The battery has the same curve, just higher up. Due to the rig?

Originally Posted by jackinnj

Note that the SLA battery is much higher impedance than any of the regulators.

I keep try to tell folks that, they won't believe me.

[jackinnj]

I've measured battery impedance from 100 milliohms to 1.2 ohms -- using the rig described above, with a Jensen transformer as the current injector, with the AP Complex Impedance method.

The fascinating thing is that you can get the impedance into tens of microOhms with the load sense leads separate.

[nigelwright7557]

One very important thing with regulators is decoupling.
You can have the best reg but without decoupling it is useless.

I had a few problems with a class d amp howling and I had missed off the output 100nf capacitor.

[abraxalito]

Originally Posted by jackinnj

Note that the SLA battery is much higher impedance than any of the regulators.

Interesting to see this, thanks Do you have any lithium ferrophosphate cells to compare this with? They have the reputation of being low impedance but I haven't seen any audio band measurements anywhere.

[jackinnj]

Decoupling a super regulator (of the Jung-Didden variety) can lead to oscillation. This is really a case where "bigger is better" and "more is better" ain't. Class D is supposed to howl!

I've asked around to borrow a LiFePO4 battery and will see what I can come up with. I can purchase a 12V 7AH unit for around $99, but am content to use the SLA batteries I have for test purposes. They power the regulator tests in the big Danish cookie tin.

[martin clark]

Choosing the right decoupling for such fast regs is certainly a non-trivial problem. usually works better with a very small series R to define some damping, but choice of R and C may well require measurement and re-measurement...

BTW I've a spare 90Ah SLA (Bosch Silver, about 20Kg) under the kitchen bench, and it doesn't measure as well as Jackinnj's SLA example. I'd be very surprised if a small LiPo cell comes anywhere close, given the huge differences in plate area - H2SO4 as an electrolyte conducts rather well...