Bipolar discrete shunt regulators

Status
Not open for further replies.
Jump to Latest
#421
Was it 560uF? My, that's a lot. Don't remember the reasoning behind it,
Click to expand...

Actually, your memory is not that bad. The superreg applies an RC filter on the voltage ref (500 Ohms 120µF). There is however a chapter "High-Performance topologies and Noise" in part 2 of the articles (should be on Walt Jung's WEB site), and he is referring to the Pooge regulator which was also presented and which has the voltage reference connecting directly to the error amp's emitter, same as in the current circuit. If I understand him correctly the noise is reduced by bypassing the voltage reference directly, and in order to be effective at a given frequency the impedance of the bypass capacitor needs to be in the same range as the dynamic impedance of the voltage reference.
 
#422
clarification - a quote from Christer

"Regarding the interesting results for zeners, i just made a
quick check with an 8.2V zener and it had reasonable noise
values of 1.1 uV, which should be at least as good as stringing
up a corresponding number of LEDs. Hence, it seems there
are mainly the zeners just around 6V that are terribly noisy.
i suspect it has something to do with these diodes
having a mixture of zener and avalanche breakdown or
something like that."
 
#424
a later conclusion from christer

* If you can get the desired voltage drop with one single
diode or LED (excluding blue LEDs) then it seems usually
best to use that single device. Special requirements for
the current may affect the choice, though. For higher voltages
up to or just below 12V, string together a suitable number
of red, yellow or green LEDs. From 12V upwards, use a
zener diode (or several if required).

so multiple LED's are fine

sorry about that 🙂
 
#426
MRupp said:


Actually, your memory is not that bad. The superreg applies an RC filter on the voltage ref (500 Ohms 120µF). There is however a chapter "High-Performance topologies and Noise" in part 2 of the articles (should be on Walt Jung's WEB site), and he is referring to the Pooge regulator which was also presented and which has the voltage reference connecting directly to the error amp's emitter, same as in the current circuit. If I understand him correctly the noise is reduced by bypassing the voltage reference directly, and in order to be effective at a given frequency the impedance of the bypass capacitor needs to be in the same range as the dynamic impedance of the voltage reference.
Click to expand...

Right. Makes sense now.

Jan Didden
 
#428
Martin, I would not bother with the alternative ccs. It is overly complicated for little benefit; sometimes I get some crazy idea that should stay on my computer, really.

If anything, my simulations show that even a simpler ccs than the currently used cascode would work very well (below 1mOhm Zout up to 1MHz, exc. line regulation), in case Andrew's concerns about the ccs are proving to be bothersome in the real circuit. The simple scheme (attached) has been much used in other shunts already, for instance in the salas shunt regulator.

Well, you have a real circuit in your hands already, hope to hear soon about its behaviour in the real world. Meanwhile my battle with the high voltage high power shunt regulator continues.
 

Attachments

  • newsimpleccs1.jpg
    newsimpleccs1.jpg
    20.7 KB · Views: 917
#429
I wonder if a simple 317+resistor CCS would work well at the input?
It does restrict the input voltage somewhat unless some start up circuit is employed to prevent the 317 seeing full voltage at switch on.

Then there's Jung's improved CCS using the 317.
Fig13a for LV duty
Fig13c for HV duty (he quotes <=450V with a suitable cascoding FET)
 
#430
AndrewT said:
I wonder if a simple 317+resistor CCS would work well at the input?
It does restrict the input voltage somewhat unless some start up circuit is employed to prevent the 317 seeing full voltage at switch on.

Then there's Jung's improved CCS using the 317.
Fig13a for LV duty
Fig13c for HV duty (he quotes <=450V with a suitable cascoding FET)
Click to expand...


Andrew, my experience with the lm317 in a different circuit was not positive. I am not one of those people that could tell with 100% certainty that the sound changed with a different output capacitor, but even I could tell there was a different sound with the lm317 employed. I have recently shared this opinion with others in the forum in the context of the Maida regulator which I built and they agreed. Unless the other two were on the same medication (just kidding), perhaps there's something to the effect of the 317. However, I don't discourage anyone to try it and check for himself.
 
#432
janneman said:



What voltage, what power?

Jan Didden
Click to expand...

Something like 250-300V, 110-130mA, for a spud amp based on the 21lu8 tube. The Maida worked but as I said, even I could tell the difference in sound, so I'm looking for a shunt design that sounds good and is stable. salas has a nice design but I've had quite a bit of trouble with it, for which I blame myself in not getting it to work properly.
 
#434
I guess I was so impressed with salas' low voltage version that I really wanted to get that working. I've had your regulator in sight too, just haven't been convinced yet to give up on the shunt version. I'll probably build a version anyway, just because I'm curious. One thing is that I haven't found any experiences with it, build wise, and sound wise too; stability, ease of getting it to work, sensitivity to layout, etc. Lots of question signs; perhaps I should just ask you all these questions, in an appropriate thread.
 
#439
LED CCS

The current through the first LED is only about 1.2 ma with the present jfet ccs. Andrew pointed out that this is too low. If the bottom jfet is changed to a 2SK170 the total current doubles and the current through the first LED goes up to 6.8 ma. Would this be enough to solve the problem of low current through the first LED? I used LTSpice to simulate this.

Jim
 
#440
no.
The input CCS does not work at all well.

The base currents of the two lowish gain BJTs destroys the constant voltage across the first LED. Varying the LED current/voltage directly changes the voltage across the first resistor. That in turn makes the CCS not constant. It just does not work when highish currents are required. That is the inherent problem:- high current and two BJTs.
This explains why the original Cetoole had R5=1k (swamp the LED with current)

The other CCS defining the reference for the shunt works very well.

I have suggested two alternatives.
A high gain first transistor and a FET for the second. This was rejected.
A cascaded pair of 317s (W.Jung), but this too was rejected.
 
Status
Not open for further replies.
Top Bottom