AndrewT said:Hi,
try replacing the input CCS with a high voltage supply feeding a high resistance.
say 10kV through 63k giving a 159mA CCS.
Makes the simulation simpler and also finds whether it's the shunt that is oscillating or the CCS or an interaction between the two.
Andrew, thanks for the suggestion, good idea. Just tried replacing the CCS with a single resistor, 10k in value, fed by a voltage source of 3kV DC, And now, it started oscillating after a few milliseconds, on this computer that it didn't oscillate before. I'm not sure what conclusions we can draw any longer, simulation is simulation.
My interest in this regulator was the slightly better AC analysis performance from 10kHz to 200kHz, when compared to salas' design. However, besides questioning the real benefit of a few dB in that range, now there are even more issues with it, real or not. It suddenly became a time sink. Personally I am very much getting the feeling of an academic exercise, interesting puzzle, that sort of thing.
Martin, subjective things like sound are also not on my list of discussions.
Again, these are just some loose comments. Please don't take any of it as a recommendation. Each make your own mind.
This being the circuit generating these plots. So much for the good news. The downside is that it oscillates for me when running the transient analysis. It would be nice if this circuit could work in reality.
One strange thing I noticed is that a small deviation from the ratio of 1 to 1.1181 in the value of the two resistors part of the voltage divider will completely throw off all the performance.
One strange thing I noticed is that a small deviation from the ratio of 1 to 1.1181 in the value of the two resistors part of the voltage divider will completely throw off all the performance.
Attachments
I don't understand D2 in that location.
It appears that D2 forces the output to ~ Vd2+Vd1+Vbeq3+Vgsj1.
D2 must allow a variable voltage to allow the regulator to compare the voltage at Q1 base to the voltage at Q1 emitter. This may explain why the resistor ratio is critical. It becomes an oscillator as soon as the Q1 voltages stray from the enforced set points.
How about replacing D2 with a resistor or maybe a crd/Jfet/ccs?
It appears that D2 forces the output to ~ Vd2+Vd1+Vbeq3+Vgsj1.
D2 must allow a variable voltage to allow the regulator to compare the voltage at Q1 base to the voltage at Q1 emitter. This may explain why the resistor ratio is critical. It becomes an oscillator as soon as the Q1 voltages stray from the enforced set points.
How about replacing D2 with a resistor or maybe a crd/Jfet/ccs?
Unfortunately this wouldn't (doesn't) fix it. There was a resistor in that location initially. If enough current passes through it it will oscillate. Less current fixes oscillation but the output impedance goes high. Would it help if I posted the currents/voltages as I did in the past?
OK, my bad, I apologize. That was a plot of the circuit with no tweaks. 30V in, about 26V out, values as shown on the schematic. Took a bit of time to use Mr. Pardo's suggestions (tweaks to set current through various parts of the circuit). The output impedance with the tweaks is 15ohms across the spectrum
He says on his website it's the best sounding regulator he has ever listened to.
Edit: bad phrasing fixed.
He says on his website it's the best sounding regulator he has ever listened to.
Edit: bad phrasing fixed.
You didn't say what your projected load current is, but if you're looking for something that performs, may I suggest the salas' regulator. Hard to beat that one; will run circles around the superteddyreg, IMHO. You can find a number of variations and get help in the following thread
http://www.diyaudio.com/forums/showthread.php?postid=1665552#post1665552
Choose 2sk170s with an Idss in the 7.5mA range.
http://www.diyaudio.com/forums/showthread.php?postid=1665552#post1665552
Choose 2sk170s with an Idss in the 7.5mA range.
Attachments
"best low noise Reg"
Ikoflexor,
It might be easier to go directly to the "best low noise Reg" thread that you, salas, analog, etc, have developed.
" It's (may) not be the "best" reg (ever), it is low noise and it works very well" (as per Salas)
It does works extremely well, ("sounds excellent"), readily scales up/down, adjusts easily, has few problems with oscillation and is quite a simple circuit, too!
Then, the addition of John's (EC Design) Charge Transfer (raw) power supply (including the Cmultiplier) considerably improves this again (the absence of "line noise"), and the result is "quite a surprise" (!).
Simple, very clever design/engineering - thank you all.
Ikoflexor,
It might be easier to go directly to the "best low noise Reg" thread that you, salas, analog, etc, have developed.
" It's (may) not be the "best" reg (ever), it is low noise and it works very well" (as per Salas)
It does works extremely well, ("sounds excellent"), readily scales up/down, adjusts easily, has few problems with oscillation and is quite a simple circuit, too!
Then, the addition of John's (EC Design) Charge Transfer (raw) power supply (including the Cmultiplier) considerably improves this again (the absence of "line noise"), and the result is "quite a surprise" (!).
Simple, very clever design/engineering - thank you all.
Hi Iko,
I knocked up the Shunt Reg, as per Salas' design, on some "veroboard" and it worked straight away - cranked it up to feed EUVL's Taylored-DOA follower headamp (250mA for amp'r, plus another 250mA "down the Shunt" = 500mA @ +/- 24volts) (unfortunately, I just couldn't get the toole shunt reg to give anything like these results).
Fed this via a simple C-R-C + Cmultiplier (as per F3 amp) - very pleased with the result and as yet, haven't finished the pcb.
I have a rough assembled simple CT supply for both channels so far (just ready to hand components) and it was a bit weird really, as I thought I'd killed the whole thing, it was so silent.
It seemed to elimenate a whole lot of background noise that I didn't really notice, until it wasn't there. And for some unknown reason, it seems to be more dynamic but I'm not sure about this - it's very easy to imagine things, even with the k701s (AKG h/phones) .
And then some un-noticed difficiencies in the NOS Dac became more obvious and, unfortunately, I'm still stuck there!
So I haven't actually finished a properly spec'd, balanced CT supply yet, as this is included in the developing Shunt reg pcb, but I have no doubt at all that even the basic version is worth the extra bit of circuitry and getting the right parts and things.
No doubt, Mercurio is probably wondering why it's taking so long for this remarkable power supply development to get established!
There's the "Never Connected" system that's produced in England that appears to do the same/similar thing - another company there also installs these regs in audio gear at a pretty steep cost - they seem very confident about the results, so ....
I knocked up the Shunt Reg, as per Salas' design, on some "veroboard" and it worked straight away - cranked it up to feed EUVL's Taylored-DOA follower headamp (250mA for amp'r, plus another 250mA "down the Shunt" = 500mA @ +/- 24volts) (unfortunately, I just couldn't get the toole shunt reg to give anything like these results).
Fed this via a simple C-R-C + Cmultiplier (as per F3 amp) - very pleased with the result and as yet, haven't finished the pcb.
I have a rough assembled simple CT supply for both channels so far (just ready to hand components) and it was a bit weird really, as I thought I'd killed the whole thing, it was so silent.
It seemed to elimenate a whole lot of background noise that I didn't really notice, until it wasn't there. And for some unknown reason, it seems to be more dynamic but I'm not sure about this - it's very easy to imagine things, even with the k701s (AKG h/phones) .
And then some un-noticed difficiencies in the NOS Dac became more obvious and, unfortunately, I'm still stuck there!
So I haven't actually finished a properly spec'd, balanced CT supply yet, as this is included in the developing Shunt reg pcb, but I have no doubt at all that even the basic version is worth the extra bit of circuitry and getting the right parts and things.
No doubt, Mercurio is probably wondering why it's taking so long for this remarkable power supply development to get established!
There's the "Never Connected" system that's produced in England that appears to do the same/similar thing - another company there also installs these regs in audio gear at a pretty steep cost - they seem very confident about the results, so ....
jameshillj said:And for some unknown reason, it seems to be more dynamic but I'm not sure about this - it's very easy to imagine things, even with the k701s (AKG h/phones) .
You are not imagining things.
A fancy pcb is in the works (or was) by somebody else, but I'm thinking of putting one not so fancy together for my own use as well; I'll post it somewhere around here soon, I think, or perhaps I'll open a thread for the salas regulator and variations where we can all chip in with various things.
However, given that it's such a simple circuit, perf board and point to point is even recommended.
Here's the output impedance of some latest tweaks that I'm trying, but not built yet; will be soon testing it in reality. Couldn't get the Zout lower than what's shown in the high frequency, but I'm not sure I should bother with that anyway.
However, given that it's such a simple circuit, perf board and point to point is even recommended.
Here's the output impedance of some latest tweaks that I'm trying, but not built yet; will be soon testing it in reality. Couldn't get the Zout lower than what's shown in the high frequency, but I'm not sure I should bother with that anyway.
Attachments
And this would be the ripple rejection bode plot. Let me remind you though, this is just simulation. In reality things may be different. For something true and tried I'd suggest the original salas regulator (see link a few posts earlier).
Attachments
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
- Home
- Amplifiers
- Solid State
- Bipolar discrete shunt regulators