The Belleson Superpower, the ultimate voltage regulator ... ?

[jackinnj]

The effect of C5 (which isn't part of the modified POOGE 5.1 used as the rail driver) is shown with this "pulse" instead of sine perturbation.



The compensation scheme for the 1N4148+BZX846V2L is probably going to be a different from that used for the LM329DZ:

[belleson]

Originally Posted by jackinnj

The effect of C5 (which isn't part of the modified POOGE 5.1 used as the rail driver) is shown with this "pulse" instead of sine perturbation.

Yes, same effect, different view. A pulse is not applicable to the PSRR discussion. I stated C5 represents capacitance on the input of the tested regulator. My 20+ years of integrated circuit design and testing of op amps, regulators, micro-machined sensors, track/holds, high accuracy A/D and D/A converters, et al have shown many times that neither the test equipment nor the device under test can be considered in isolation, but instead must be developed and measured as a system.

Originally Posted by jackinnj

The compensation scheme for the 1N4148+BZX846V2L is probably going to be a different from that used for the LM329DZ:

Yes, I used available simulation components to get the same Vout. This also does not apply to the discussion of PSRR testing with a unipolar source driver.

[belleson]

jackinnj,

After reviewing your test diagram in post #48, I see that you do NOT have an input capacitor on your test circuit so my comments about C5 relate to our test setup and not yours. That tangent about slew rate distortion of the sine wave can be considered an intellectual exercise for the interested reader :-). Nonetheless, I suggest care when measuring out to 200KHz, as it doesn't take much capacitance at that frequency to distort the input signal.

Thanks for rearranging your setup and re-testing.

Brian
Belleson, LLC

[jackinnj]

OK, I think these results are more consistent with what you found -- the DSP Analyzer has a much narrower bandpass filter than the Analog Analyzer. The A/B test in Analog Mode can't be run in Bandpass Mode without writing a macro.

I pulled out the National Semiconductor LM7805 Datasheet and ran the test with an STMicro LM7805 and a whopping 3.5V RMS -- they show LR of -82dB and my result is -78dB.

The other tests were done with 1V pk-pk. The test with the Super Regulator used an AD825 error amplifier (not decoupled) and LED biasing so the results aren't quite consistent with the 2/95 TAA article. And, after all, the Super Regulator is a lot more complex.

[belleson]

We finished our new ripple rejection test fixture yesterday and there is good news and bad news. The good is the measurements are much better, we're measuring DUT performance more accurately and not so much the characteristics of the test fixture, e.g. AC feedthrough.

The new fixture uses a DC+AC driver based on the LM3886 as seen in the attached schematic. The 3886 has wide bandwidth, sufficient output power and slew rate to deliver a high frequency sine wave on top of a DC voltage.

To see how Superpower stacks up against other commercially available regulators, 3 other regulators were tested. We did direct comparison measurements by setting up the fixture, calibrating Vin and frequency and plugging in each of the regulators in turn. Vin=16VDC+1VppAC, Vout=12V with DC load of 240 Ohms. (Vout=9V for Dx7809). Output current is 50mA except for Dx7809 where it is 37.5mA, to be fair to the regulators that cannot deliver high currents. Vin AC was calibrated after each frequency change.

Two measurement systems were used, a 24 bit differential input A/D with 20KHz bandwidth and a 16 bit 195K sample/second A/D. Both were calibrated to a 4 3/4 digit true RMS DMM and correlated to each other on 10KHz and 19KHz measurements where the ranges overlap and measurements were available. Measurements were made from 100Hz to 600KHz, changing from the 24 bit A/D to the 16 bit between 19KHz and 50KHz. The 600KHz point was made with the magic of aliasing (a.k.a. undersampling), where the "carrier" folded back into the baseband at about 15KHz.

The bad news is that you probably won't believe the results--I had to triple-check to believe it myself. Superpower PSRR is essentially in the noise of each respective measurement system below 600KHz, where it still measures a respectable -78dBV. For what it's worth, a comparison graph is attached and a spectrum at 19KHz. A 19KHz LM7812 spectrum is also attached as a baseline reference to show that indeed the test fixture does work. We will have an updated web page on our site by next week.

[CraigBuckingham]

What's with the 90MHz oscillation on the website?

Superpower Super Regulator by Belleson

[belleson]

Adding the 19KHz spectra from the 24 bit A/D for LM7812 and Superpower to show we get the same answer 2 ways, with a lower noise floor.

Adding a "sanity check" where Vin is increased to 2Vp-p, with visible supply feedthrough at approx -110dB.

To CraigBuckingham:
The 90MHz oscillation could have been RF picked up by the AD797 or some artifact of the breadboard. We were interested in output dynamic current delivery so once we got the Iout-dependent oscillation tamed, we took our measurements and moved on.

[abraxalito]

The link to the ALSWR on the page referenced by Craig is broken - here's one that works:

http://www.connect-audio.co.uk/Regulator/ALWSR.pdf

[belleson]

Originally Posted by abraxalito

The link to the ALSWR on the page referenced by Craig is broken - here's one that works:

http://www.connect-audio.co.uk/Regulator/ALWSR.pdf

Updated, thanks!

[Larsham]

Here is the patent application for the Belleson regulator. You can click the Documents tab to get the complete 23 page application:
US2010037556 VOLTAGE REGULATOR USING DEPLETION MODE PASS DRIVER AND BOOT-STRAPPED, INPUT ISOLATED FLOATING REFERENCE

Interesting design.