Hello,
Newbie here!
I am making several DC power supplies. They are shunt power supplies.
The problem I am having is that the shunt current of each PS varies. I think that must be due to the gain of the PNP transistor (used for the shunt current) is varying from one device to the next.
I can put a pot on each one and adjust the gain individually but I think it would be better to devise some circuit that ensures the amount of current coming through the PNP transistor is the same even when the gain from transistor to transistor varies.
This must be a common issue?
Are there any circuits out there?
Thanks,
skibum
Newbie here!
I am making several DC power supplies. They are shunt power supplies.
The problem I am having is that the shunt current of each PS varies. I think that must be due to the gain of the PNP transistor (used for the shunt current) is varying from one device to the next.
I can put a pot on each one and adjust the gain individually but I think it would be better to devise some circuit that ensures the amount of current coming through the PNP transistor is the same even when the gain from transistor to transistor varies.
This must be a common issue?
Are there any circuits out there?
Thanks,
skibum
What is the variation you see? Is it important to be super accurate?
Remember that a shunt has the task to keep the voltage the same, so as the load current changes or the input voltage changes, the shunt current will adapt itself to keep the output voltage constant.
The only requirement is that the idle shunt current is high enough not to run out of current for max load current situations.
Do you have a circuit you can post? It is difficult to advise without knowing the circuit.
jan didden
Remember that a shunt has the task to keep the voltage the same, so as the load current changes or the input voltage changes, the shunt current will adapt itself to keep the output voltage constant.
The only requirement is that the idle shunt current is high enough not to run out of current for max load current situations.
Do you have a circuit you can post? It is difficult to advise without knowing the circuit.
jan didden
You mean my country flag? Seems OK to me - we can use it up until it is very short. Unlike the French
Yes the shunt needs a feedback control for settability but since there's no schematic, I didn't want to speculate.
jan
Hello Jan,
Yes, I would like to figure out how to get a super accurate shunt current. I would also like this to be consistent from one PS to the next. I made a mistake and the variation in the transistor to transistor gain does not seem to be the biggest factor in shunt current variation.
I am using a DC power design similar to this:
http://www.twistedpearaudio.com/docs/power/placid_2_0_1_schem.pdf
I would like to remove VR1 and just use a fixed resistor R1. The shunt current variation seems to be mostly coming from the variation in the JFET current source.
I replaced the JFET current source with a current limiting diode but that didn't really help. I think the part to part current variation in the current limiting diode was also to large.
What I was thinking is to somehow take the voltage across R3 (this voltage is caused by the shunt current) and use that as feedback somehow to a current source.
I found this constant current source circuit:
http://www.electronicdesign.com/files/29/18966/fig_01a.gif
Could I use the voltage across R3 as the "Vin" to the LMP7731 opamp?
Is anything like this possible or am I making things much more difficult than I should be? : )
Well the problems with these type of circuits is that they try to generate an accurate reference with D1 and D2 which have undefined voltage levels and also vary from one to the other unit.
So even when you adjust the current through them there still will be unit-to-unit variations.
That's one of the things with diy audio. There are very good engineers that developed reference diodes and shunt reference regs specifically for this purpose, but hey, someone in diy audio uses LEDs so they must be the best and everybody uses them, even if they are not the best choice.
Then if you want to explain how this really works, someone comes up and will say, but it sounds the best and that's the end of the discussion.
So, I believe when you continue to use LEDs you will continue to see this variations.
The give-away is the trimmer - a proper designed shunt doesn't need any trimmers!
Use something like an LM4040-5V as a 5V ref instead of the LEDs and you're done.
jan
So even when you adjust the current through them there still will be unit-to-unit variations.
That's one of the things with diy audio. There are very good engineers that developed reference diodes and shunt reference regs specifically for this purpose, but hey, someone in diy audio uses LEDs so they must be the best and everybody uses them, even if they are not the best choice.
Then if you want to explain how this really works, someone comes up and will say, but it sounds the best and that's the end of the discussion.
So, I believe when you continue to use LEDs you will continue to see this variations.
The give-away is the trimmer - a proper designed shunt doesn't need any trimmers!
Use something like an LM4040-5V as a 5V ref instead of the LEDs and you're done.
jan
Last edited:
In my curcuit I have replaced the two LEDs with a single 1N4448 diode. I did need that much voltage.
Should I replace just the 1N4448 with the LM4040?
OR
Replace the 1N4448, the JFET and R5 with the LM4040?
Thank you!
Well the diode is also not very accurate as a reference. That may not be bad if you're not concerned whether the Vout is 14.5 or 14.8 or 15.3 volts, then it's OK.
But a real reference like the 4040 series (and there are many more, 1.25V, 2.5V, 5V ...) is within 1% or even 0.5% or better.
So if you want say 15V, get a 5V reference, put some current through it (and that can be anything between a 100uA and a few mA) , use a divide-by-three resistor divider on the other input of the opamp and you have accurate rockstable 15V. Done.
For icing the cake you can put an RC-filter on the ref, or put a cap in parallel of the top resistor in the divider to increase the ripple gain, but basically that's all there's to it.
Like Einstein said, make it as simple as possible, but not simpler
jan
The input half is a current regulator.
The output half (the bit you call the shunt) is a voltage regulator.
Operating as a voltage regulator it does not need to pass or see a constant current. It's operation in holding voltage constant relies on it being able to pass variable current and to change that current very quickly in response to output load variations and input variations.
The output half (the bit you call the shunt) is a voltage regulator.
Operating as a voltage regulator it does not need to pass or see a constant current. It's operation in holding voltage constant relies on it being able to pass variable current and to change that current very quickly in response to output load variations and input variations.
Well the diode is also not very accurate as a reference. That may not be bad if you're not concerned whether the Vout is 14.5 or 14.8 or 15.3 volts, then it's OK.
But a real reference like the 4040 series (and there are many more, 1.25V, 2.5V, 5V ...) is within 1% or even 0.5% or better.
So if you want say 15V, get a 5V reference, put some current through it (and that can be anything between a 100uA and a few mA) , use a divide-by-three resistor divider on the other input of the opamp and you have accurate rockstable 15V. Done.
For icing the cake you can put an RC-filter on the ref, or put a cap in parallel of the top resistor in the divider to increase the ripple gain, but basically that's all there's to it.
Like Einstein said, make it as simple as possible, but not simpler
jan
I am not that concerend with the variation of Vout. (I need 3.0V and 5.0V). The Vout variation is actually very small (acceptable for me) between the power supplies that I have made.
It is the shunted current that varies all over the place. anywhere from no shunted current (and the Vout will not come up) to over 200mA+. I want to ensure that I can keep that shunted current values very tightly regulated.
I see that there is a LM4041 that is a 1.25v shunt reference.
So if I just replace my 1N4448 with a LM4041 will my shunt current amounts become consistent?
Thanks so much!
Ahhhh! I have no problems at all with the voltage regulator half. That works perfectly for me. I just want to be able to keep the current regulation very tight between one supply and the next. At the moment the current is all over the place when I use a fixed resistor for R1.
much appreciated!
Well the ref and shunt current (I guess you mean the current going into the ref, NOT the current of the 'shunt regulator') are two different things. Actually three here.
The ref takes the current from the FET CCS and develops the ref voltage. The FET CCS current value is a secondary issue here, it should work with whatever reasonable current it is given. So getting the right ref will make the regulator shunt work OK.
Then the 2nd thing is the value of the FET CCS. For some reason this thing has to do two tasks: provide current to the ref, AND provide a bias current for the pass device current source.
The connection here is that when you increase the FET CCS you increase the pass device CS and that may be too much current for the shunt to take up and then Vout rises.
With full load, the reg shunt should still have some current through it to keep regulating.
In fact you calculate backwards: assume you have a load that takes 100mA max. Then you design the reg shunt to take say 120mA for peace of mind. Then you set the pass device current to 120mA as well.
Problem here is (again) uncontrolled gain of the pass device comes into the equation and you keep on tweaking it and it drifts with temperature and it's generally a mess.
A good series CS in itself is a regulator (current regulator) and not some flaky open loop thing like a series Q with a base current pulled out of it.
Look up de data sheet of the LM317, the application as a CCS. Easy, simple, rock stable and accurate.
If you don't like an IC, replace the LM317 with an NPN transistor assuming that the Vin = C, Vout = E and the adj terminal = B. That LM317, after all, is just a transistor with ambitions
jan
The ref takes the current from the FET CCS and develops the ref voltage. The FET CCS current value is a secondary issue here, it should work with whatever reasonable current it is given. So getting the right ref will make the regulator shunt work OK.
Then the 2nd thing is the value of the FET CCS. For some reason this thing has to do two tasks: provide current to the ref, AND provide a bias current for the pass device current source.
The connection here is that when you increase the FET CCS you increase the pass device CS and that may be too much current for the shunt to take up and then Vout rises.
With full load, the reg shunt should still have some current through it to keep regulating.
In fact you calculate backwards: assume you have a load that takes 100mA max. Then you design the reg shunt to take say 120mA for peace of mind. Then you set the pass device current to 120mA as well.
Problem here is (again) uncontrolled gain of the pass device comes into the equation and you keep on tweaking it and it drifts with temperature and it's generally a mess.
A good series CS in itself is a regulator (current regulator) and not some flaky open loop thing like a series Q with a base current pulled out of it.
Look up de data sheet of the LM317, the application as a CCS. Easy, simple, rock stable and accurate.
If you don't like an IC, replace the LM317 with an NPN transistor assuming that the Vin = C, Vout = E and the adj terminal = B. That LM317, after all, is just a transistor with ambitions
jan
Well the ref and shunt current (I guess you mean the current going into the ref, NOT the current of the 'shunt regulator') are two different things. Actually three here.
The ref takes the current from the FET CCS and develops the ref voltage. The FET CCS current value is a secondary issue here, it should work with whatever reasonable current it is given. So getting the right ref will make the regulator shunt work OK.
Then the 2nd thing is the value of the FET CCS. For some reason this thing has to do two tasks: provide current to the ref, AND provide a bias current for the pass device current source.
The connection here is that when you increase the FET CCS you increase the pass device CS and that may be too much current for the shunt to take up and then Vout rises.
With full load, the reg shunt should still have some current through it to keep regulating.
In fact you calculate backwards: assume you have a load that takes 100mA max. Then you design the reg shunt to take say 120mA for peace of mind. Then you set the pass device current to 120mA as well.
Problem here is (again) uncontrolled gain of the pass device comes into the equation and you keep on tweaking it and it drifts with temperature and it's generally a mess.
A good series CS in itself is a regulator (current regulator) and not some flaky open loop thing like a series Q with a base current pulled out of it.
Look up de data sheet of the LM317, the application as a CCS. Easy, simple, rock stable and accurate.
If you don't like an IC, replace the LM317 with an NPN transistor assuming that the Vin = C, Vout = E and the adj terminal = B. That LM317, after all, is just a transistor with ambitions
jan
Hi Jan,
When I am saying "shunt current" or "current regulation" I am talking about the amount of current that the supply will deliver. I am NOT talking about the Ref shunt current provided by the Jfet.
Example:
I have a load that takes 25mA. I want to set this power supply to deliver 50mA. (25mA of headroom). When I set up R1 to deliver 50mA in one power supply and then I use the SAME value for R1 in another power supply I get anywhere from 0mA to well over 200mA of current.
This is my problem.
Of course that's the problem. The current delivered by the CCS you're trying to set varies directly with the transistor gain, which is about as unpredictable as it gets. If you have a transistor with a beta of 50 and you pull 1mA out of the base you get 50mA CCS. Then in another supply you have a transistor with a beta of 120 and you get a CCS with 120mA. etcetera.
Your CCS is totally uncontrolled, that's why I gave the reference to the LM137.
jan
I will replace the Jfet with LM317. It is Incredibly simple!
Will the LM317 have noise and slew rate values in the same range as the PN4393 Jfet that I have been using? If not I might also try an NPN transistor. Can you recommend a good low noise NPN to try out? BC550?
Thanks!
No not the JFET. You replace the series pass transistor with the 317.
You then can use the whole JFET stuff, fed from the raw supply, just to set a current through the reference, with no concern about the pass current value.
Edit: I mean QP1 in the twisted pear circuit posted before. Because that causes the trouble: the base current pulled out of QP1 sets the current but because the transistor beta is not constant, the current varies all over the place.
jan
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.