Super Regulator

@WaltJ I can not find the part two what is called "A 1V Reference with Mirrors". Is this article available in the same website?

When will be your new advanced regulator schematic available that you were working on with some diyaudio members?

Next week I will try with some other opamps like OPA1611 to see if I will get it working on.
 
@WaltJ I can not find the part two what is called "A 1V Reference with Mirrors". Is this article available in the same website?

When will be your new advanced regulator schematic available that you were working on with some diyaudio members?

Next week I will try with some other opamps like OPA1611 to see if I will get it working on.
You should have the P2 writeup now. It is not a new regulator, but a new reference for a lv regulator. If you are still having troubles, post your sch. Opa1611 looks like it wont work on 1v ref, but will on 2.5v.

wj
 
is there any advantage of using a Current Feedback type of opamp in this project ? I already have a nicely working >5A version with ADA4898-1, but think about pushing it to the absolute limit, i.e. GLED431 , string of LEDs instead of zener, bulk foil resistor on feedback. Some photos for eye joy. last picture is FFT from my LCLC filter (14x3300uF Nichicon UKW 11mR ESR) under 5A step with 0.5A/us, and what Superreg output.
 

Attachments

  • negative 1.jpg
    negative 1.jpg
    195.7 KB · Views: 46
  • positive 1.jpg
    positive 1.jpg
    235.8 KB · Views: 43
  • FFT 1.jpg
    FFT 1.jpg
    280.1 KB · Views: 44
  • LC filter.jpg
    LC filter.jpg
    283.5 KB · Views: 43
also, due to the unpredictable current demand of power amplifier output and beta droop of pass transistor, and current limitations of the op amp, i find it hard to properly program input ccs. Darlington seems to solve many problems at once, keeps op amp cool, renders slew rate irrelevant, keeps current load on level shifting diode(s) zener/led string almost constant, so their dynamic impedance characteristics becomes irrelevant, and most of all, keep everything nicely in class A operation even at 12+ Ampere spikes. Only Spice simulation suggests that regulator will be much slower, although i keep both Darlington transistors in high current state, close to their max ft, with base discharge resistor in place. Despite my best efforts, smd resistors, tight layout, hybrid polymer caps, i still see see things on ouput with square 5A load, and not much better than cable salad of prototype breadboard. Since both input CCS and pass transistor need some dropout for proper function, and I found 2.2 Volts adequate for sudden voltage drops at the input under heavy load, i use this dropout for another 100R to green LED between input and output to check for proper function. I shorted already a few transistors, and not noticed immediately when it started giving non regulated output.
 

Attachments

  • lc filter 2.jpg
    lc filter 2.jpg
    184.1 KB · Views: 32
  • lab view.jpg
    lab view.jpg
    177.7 KB · Views: 32
  • green led ok.jpg
    green led ok.jpg
    205 KB · Views: 32
is there any advantage of using a Current Feedback type of opamp in this project ? I already have a nicely working >5A version with ADA4898-1, but think about pushing it to the absolute limit, i.e. GLED431 , string of LEDs instead of zener, bulk foil resistor on feedback. Some photos for eye joy. last picture is FFT from my LCLC filter (14x3300uF Nichicon UKW 11mR ESR) under 5A step with 0.5A/us, and what Superreg output.
Several major disadvantages to CFA as error amp, high and dissimilar Ib, need for fixed Rf to keep stable, just not a good match for this application at all. Stay with low noise jfets.

pm329 with 3 gleds is best around 7v for vref, low noise and high gain for low Zo. Be sure to use low noise pnps for best performance.

Vishay S102cs great for sense Rs, but keep net z balanced. Texas Components are a great source.

darlington for pass device is good, a real necessity for 5a.

Walt Jung

Ps: see p2 of the 1Vref with rleds, and ancillary stuff. Grunf’s favorite mosfet, bss139i, works wonders as Isource for 10ma or so. Two parts!
 
for 22V output, this arrangement has enough voltage headroom for everybody>, AC analysis of Vref vs Vin
CCS output current simulated - 11mA, real measured 7,5 mA, 220R resistor over led string sends 5mA to ztx951, 2,5mA stays for LEDs. I feel i need to push the leds a bit harder, min 5mA, maybe even >10mA both branches. For leds I bought HLAMP 6000. Don`t know what can i do better.
 

Attachments

  • image_2024-05-21_070526554.png
    image_2024-05-21_070526554.png
    65.7 KB · Views: 43
for 22V output, this arrangement has enough voltage headroom for everybody>, AC analysis of Vref vs Vin
CCS output current simulated - 11mA, real measured 7,5 mA, 220R resistor over led string sends 5mA to ztx951, 2,5mA stays for LEDs. I feel i need to push the leds a bit harder, min 5mA, maybe even >10mA both branches. For leds I bought HLAMP 6000. Don`t know what can i do better.
“Don`t know what can i do better” you said. I think you flat out missed my point about the pm329, Simulacrum. Look at https://refsnregs.waltjung.org/2New_Vrefs_082715_Draft.pdf in particular fig 3 where the topology is two_stage, not one. Allows for much lower z, and is more effective.

Walt Jung
 
@baswamin - the one you chose should do ok, however it errs on the expensive side already. Some time ago, I tried many different capacitor types in superreg in prototype board, including tht mlcc, tantalum, bypassed or not with polypropylene, audiofile and non-audiofile electrolytics, yada yada, without any conclusive difference between them on regulation quality. Unless you really need the 135 Celsius working of your choosing (doubtful), you might be better off with well regarded here Panasonic FM. I myself went with hybrid polymer that are even more expensive, but hybrid polymer according to many sources should be even closer to the ideal capacitor, in all regards. I chose this model for input cap and for Voltage Reference RC filter> https://www.mouser.pl/ProductDetail/Panasonic/EEH-ZU1V471P?qs=iLbezkQI%2BshMS2dpa58Rhw==

for all other places in this schematic, i decided for now on these ceramics> https://www.mouser.pl/ProductDetail...s/CL31B106KBHNNNE?qs=T0XSgvH75d4ZN8WrMi8xmQ== , but after reading about all failure possibilities of mlcc`s i might revert to use some smaller hybrid polymers everywhere.
 
  • Like
Reactions: 1 user
“Don`t know what can i do better” you said. I think you flat out missed my point about the pm329, Simulacrum. Look at https://refsnregs.waltjung.org/2New_Vrefs_082715_Draft.pdf in particular fig 3 where the topology is two_stage, not one. Allows for much lower z, and is more effective.

Walt Jung
lower z on voltage reference is nice in an universal voltage reference, i.e. with sparse input filtering and dynamic load. One more jelly bean npn bjt will not bankrupt anybody, but I just thought, that when supplied with a a super duper high output impedance ccs, and used specifically in super regulator circuit, i.e. loaded with another constant high z node of op-amp input, further lowering in dynamic impedance shouldn`t do any appreciable difference. If you insist that this is an upgrade even in Super Regulator schematic, i will add BC550C and resistor.
 
Thanks, Peranders.

Fortunately, Mr Jung and others have completed the design and tests and here we are only using their design. There is no change to the design. It is all about implementing it. Reducing the current slightly, etc, is about tweaking, not redesigning.

People have given their different opinions on how to implement it. Some of them did not originate from Mr. Jung. I think some of the advices from posters may be wrong.

Examples are:

1) Remote sensing cancels the inductance of the wires - this tells only half of the picture because remote sensing does help at audio frequencies but it still degrades output impedance at frequencies around 1MHz comparing to having the regulator right at the load.

2) Remove all or most your capacitors at the load - this seems to be very wrong with my simulations. Removing any high Q film / ceramic capacitors is the absolute right thing to do, but not necessarily electrolytic capacitors. In my simulations, electrolytic capacitors at the load help lower impedance and improve on the phase margin a lot and they are absolutely necessary. My subjective listening test also agreed with this. Simulations show that you can place multiple electrolytic caps there and the capacitance can be above 1000uF in total without problem.

3) Avoid using low ESR caps - This tells only half of the picture. My latest simulations (now uses AD825 and D44H11) showed me that the cap right at the output of the Jung regulator needs to be high ESR. 0.5R should work, 1 or 2R may be better. However, electrolytic capacitors at the load should have relatively lower ESR. Low ESR caps help improves phase margin and lower the output impedance in the MHz region.

Of course, I / my simulations can be wrong.
I'm afraid this comment removes all or most of my wishes to build a SR. This and the fact that half of the original parts are absolete and replacements need careful matching and testing with a scope. But SR isn't for amateurs, I guess.