Belleson Superpower regulator, here's the schematic

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Member
Joined 2011
Paid Member
I just found it on the website of the United States Patent and Trademark Office , where they call it application number 20100327834 . I was delighted to see that the very first word-page begins with the phrase "FEDERALLY SPONSORED RESEARCH". Gub'munt money put to good use.

The circuit is pretty cute. A conventional series pass element (NPN emitter follower) delivers current to the load. Its base is driven by an Nchannel JFET source follower, and the gate of this JFET is controlled by the error amplifier. The error amplifier and the reference voltage are bootstrapped from the output. Since the output is presumably much less noisy than the input, this creates a virtuous cycle of low noise goodness, similat to the Jung Super regulator and others. Another cuteness is that the error amplifier's supply voltage (VCC - VEE) is less than the output voltage; he "floats" the negative supply to the opamp. This gives the option of using fast, rail-to-rail, CMOS 5 volt opamps while providing a 15 volt (or 24 volt) output. Sweet.

I've attached the circuit schematic (.png file) and the entire patent application (.pdf files). Enjoy!

Grrr, I had to divide the application into 5 pieces to get each file size down below the byte limit for posting. Sorry about that. Just download em and assemble them together using Acrobat on your 'puter.
 

Attachments

  • Belleson_schematic.png
    Belleson_schematic.png
    19.9 KB · Views: 2,923
  • part1.pdf
    115.1 KB · Views: 883
  • part2.pdf
    115 KB · Views: 498
  • part3.pdf
    110.8 KB · Views: 425
  • part4.pdf
    107.5 KB · Views: 423
  • part5.pdf
    142.8 KB · Views: 480
the very first word-page begins with the phrase "FEDERALLY SPONSORED RESEARCH".

Just after that, it says "Not Applicable." The research was sponsored by curiosity and audiophile-ity :).

The circuit is pretty cute.

Thanks!

while providing a 15 volt (or 24 volt) output.

(or 450V).

Also, note that the JFET can provide a ton of drive current to the series pass device base without loading the error amp at all. And the (only) two input connections are high impedance, which provides very high ripple rejection across a very wide bandwidth.
 
Oh yes,pretty cute,but I don't get it :-(
The action of A1 makes VR9=VREF and thus IR9=VREF/R9
IR9 goes throug R8 making VR8=IR9.R8=VREF.R8/R9
VR8+VR9=VREF(1+R8/R9)=K
VR12+K=VOUT ! So the entire variation of VOUT is found on R12.
There is no regulaion of VR12 ,how is VOUT regulated ???
Mona
 
Member
Joined 2011
Paid Member
The inventor presents the relevant equations on page 3 of the patent, see "part4.pdf". Starting at paragraph 33, he derives equations for the operation of the circuit in Figures 3 & 4 of the patent. Those Figures are contained in "part1.pdf".

Or just build a SPICE simulation model and vary Vinput. Does Voutput remain constant and regulated? Vary Rload; does Voutput remain constant?
 
Last edited:
After a good night sleep,I see the light :)
The circuit can work very well.VR11 is constant too making the current in R12 only dependent of VOUT and is adjusted if that varies.So R12 determens the VOUT.
There is something missing,a capacitor between the supplypins of A1.
Mona
 
Last edited:
they were one of the lowest performance discrete regs in his testing for Linear Audio, the mag of course you have to buy to get the full article, but the colour testing results were placed for free download here. The noise was worse than all but bybee, belleson and one of the Linear regs, not sure which. they were significantly noisier than LM317... output Z was one of the worst as well
 
I'm late to the party but I'll crash anyway. Since no one has mentioned these things, allow me to say the following about

* noise
Jack's article in Linear Audio says "...in this family of regulators, in this test situation in which the source is not a low output moving coil cartridge, noise is of little importance to the rankings." In fact, unless an amplifier has passive loads (vs. current sources) and high gain such as a phono or microphone amplifier, noise of a (decent) regulator is never important. Instead the input equivalent noise dominates.

"The noise was worse than all but bybee" [qusp] You mis-read the graph. Burson has the highest noise (up to 4KHz anyway), then Belleson, NCD, Linear Tech are very similar except below 20Hz LT is better. In the listening tests Burson, LT and LCD were in the top 5, which is likely the basis for Jack's statement above.

The Superpower datasheet clearly states noise as 1 to 4 ppm of Vout (Vrms) from 20Hz to 20KHz. The results in the article appear to correlate so I'm not sure why people are surprised.

* PSRR
Why is it that nobody wants to point out Belleson's fantastic PSRR? Waaay better than Bybee, Burson, NCD, Linear Tech, LM317 and Sulzer across the full 4 frequency decades.

The article states (again p. 188 bottom) "line rejection seems to be the most important, particularly for subsonic and supersonic artifacts." This is puzzling considering that number 1 in the listen tests has the best PSRR measurement and number 2 has the worst (Burson is essentially tied for worst with Bybee if I read the graph correctly).

* Zout
"output Z was one of the worst as well" [qusp] Superpower measures relatively close to the majority of regulators tested. Ten milliOhms is very good and Belleson is virtually flat to 100KHz in the article's measurements, unlike any other tested device. We plan to do some tests, but it's likely that the 10mOhm value is more due to the small PCB and its connectors rather than the circuit itself.

* "similar to the Jung Super regulator" [transistormarkj]
Yes, our design uses a JFET's negative Vgs as a level shifter instead of the zener diode in series with the error amp output of Jung's design. This puts no load on the error amp and allows voltage control of the output rather than current control. The JFET also can supply a large amount of drive current to the output transistor, allowing 2+ amps out. Our design uses a total of 10.010000022uF and fits in a TO-220 socket, the Jung uses 480uF total and does not.

* "how about the load-current transient response speed and accuracy?" [gootee]
Thanks for asking, this is where the Superpower truly shines. Our web site (Superpower Super Regulator by Belleson) has many oscillograms of transient step response for "small" (a few hundred milliAmps) and large (1A+) step response. Some of the article's tested regulators do poorly in transient response tests even though they measure well in other tests. For example the SuperTeddy has fabulously low noise but, in our measurements, does *not regulate Vout* as load current changes.

* Listening tests
It's clear that listening tests show what people like and little more. The Burson regulator as 2nd best pick in the article's listening test makes this obvious, considering that it has comparatively high noise, poorest PSRR and highest Zout. Our tests also show it to have poor output transient response.

The goal at Belleson is accurate signal reproduction, not "signal embellishment" :). We have many happy customers (evidence on our Customer Comments page) and about 70% of our sales is repeat business. We're not interested in subjective discussions about what sounds best, like the "hollow vs. solid" state arguments. When asked if Superpower will make a system "sound better" we tell people they have to decide for themselves what sounds better. We can only offer improved electrical performance.

* "now how about a BOM?" [cynikal]
1 - transformer
1 - rectifier
1 - filter capacitor
1 - Belleson Superpower :)

* Our claim to "greatness"
Perfection is a worthy goal; shy of that we seek the best compromise. Belleson Superpower offers the following combination of characteristics that cannot be found in any other regulator we know of:

- Fantastic PSRR
- Fantastic transient load response
- High output current (2+ amps, 60W max)
- Voltage output from 1.2V to 450V
- Really low Zout
- Low noise (not absolute silence)
- Small size (a plug-in replacement for TO-220 monolithics)
- A fully specified datasheet
- Great customer service
 
Member
Joined 2011
Paid Member
Oops, it seems I failed to communicate accurately. When I wrote
The error amplifier and the reference voltage are bootstrapped from the output. Since the output is presumably much less noisy than the input, this creates a virtuous cycle of low noise goodness, similar to the Jung Super regulator and others

I was trying to say that both (Belleson Superpower) and (Jung Super Regulator) bootstrap their voltage reference and their error amplifier from their output.

Hey Mark, in what way is the Belleson Superpower similar to the Jung Super Regulator? Ans: both of them power their error amplifier from the regulated output, and both of them bias their voltage reference from the regulated output.

Sorry that this wasn't clear.
 
Jack, injecting a signal on the output gives an interesting figure of merit, it's great of you to make those measurements. I put a bench setup together to do my own and connected the coupling capacitor backward. After a few minutes there was an ugly noise and I got nervous that the signal generator output was damaged. After some time I figured out it was the capacitor that popped, but that made me nervous anyway about damaging the signal generator.

So I ran the generator to our ripple rejection test setup as a high current low impedance driver to a 10 Ohm resistor in series with ten parallel 22uF BIPOLAR ;-) caps that then go to the regulator output. A true RMS meter was put between the driver and the 10 Ohms and the sine wave amplitude was adjusted so the meter reads 50mA rms.

Our original SPJ17 at 12V shows a fundamental at about -54dBV, not as good as your AP measurement of -59dB (although there could be a difference in input signal). However, the 2nd harmonic is at about -100dBV which is over 5 times better than the -86dB you saw. I'm not sure what could be the difference, perhaps the load affects the AP output?

We're about to announce a redesigned product with lower noise (you saw it here first) and it looks better than the original SPJ17, at -57dB fundamental and -106dB 2nd harmonic. An FFT of the buffered generator signal taken at the driver/10 Ohm resistor junction is included as a baseline reference. The sig gen is not an Audio Precision, just a venerable HP200CD that has significant harmonic content.

As a sanity check I plugged in a LM7812, a LT3080@12V, a LM317@12V and some of the other non-monolithic ("boutique") ones you checked, which show similar qualitative results although the numbers don't correlate exactly.
 

Attachments

  • SPJ17 12V (orig) 50mA rms.png
    SPJ17 12V (orig) 50mA rms.png
    15.1 KB · Views: 1,288
  • SPJ78 12V (LN) 50mA rms.png
    SPJ78 12V (LN) 50mA rms.png
    13.2 KB · Views: 1,271
  • HP osc output at 50mA rms load.png
    HP osc output at 50mA rms load.png
    16.6 KB · Views: 1,283
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.