Reading through the DIYaudio threads and a little own experience it looks as if shunt regulators really excel in ripple rejection and ‘subjective’ music quality. There are some proven designs, like Salas SSHV, still these don’t do ‘high power’ demanded by output stages. With the availability of high powered tubes for little money (like 40W GU50 for some USD) I am willing to experiment with ‘high power’ shunt regulators.
For example a PS capable of 300V at 220mA, with 160mA to feed 4 EL84 output tubes and a spare 60mA through the shunt tube, a triode strapped GU50. For CCS I am thinking about Gary Pimm’s CCS. I understand that if there is no load on this PS the GU50 will burn 220mA * 300V = 66W, which is more than max plate dissipation, but seems this tube can take this (even more) for short whiles. As such I know that shunt regulators produce additional heat, but that is nice for the winters when I am back in Europe.
I attach a proposed schematic for this regulator, and this is really the part I need assistance. A differential pair fed from a ‘negative’ regulated negative supply (since I read 4th edition of VA I already know how I will make this one) supplies the differential pair. Reference voltage is taken from a potential divider (R1 and R2) and control voltage is developed over R3. R4, R5 and P1 sample the output voltage, R6 is tail resistor (is a CCS, maybe LM334, also needed if it is about DC?).
If B+ voltage goes up, V on base of T2 increases, T2 conducts more, thereby T1 conducts less, less voltage is developed over R3, grid of shunt element comes closer to zero, shunt element conducts more, reducing B+ voltage.
Questions:
- Do you think it will work?
- What about using some HV bipolar transistors (MPSA 42, if I am not mistaken). How does ‘base current’ impact performance? Or better use a Vref (like TL431) instead of R1?
- Some decoupling needed, for better performance and or stability? Speed-up capacitor from B+ output to gate of T2?
For example a PS capable of 300V at 220mA, with 160mA to feed 4 EL84 output tubes and a spare 60mA through the shunt tube, a triode strapped GU50. For CCS I am thinking about Gary Pimm’s CCS. I understand that if there is no load on this PS the GU50 will burn 220mA * 300V = 66W, which is more than max plate dissipation, but seems this tube can take this (even more) for short whiles. As such I know that shunt regulators produce additional heat, but that is nice for the winters when I am back in Europe.
I attach a proposed schematic for this regulator, and this is really the part I need assistance. A differential pair fed from a ‘negative’ regulated negative supply (since I read 4th edition of VA I already know how I will make this one) supplies the differential pair. Reference voltage is taken from a potential divider (R1 and R2) and control voltage is developed over R3. R4, R5 and P1 sample the output voltage, R6 is tail resistor (is a CCS, maybe LM334, also needed if it is about DC?).
If B+ voltage goes up, V on base of T2 increases, T2 conducts more, thereby T1 conducts less, less voltage is developed over R3, grid of shunt element comes closer to zero, shunt element conducts more, reducing B+ voltage.
Questions:
- Do you think it will work?
- What about using some HV bipolar transistors (MPSA 42, if I am not mistaken). How does ‘base current’ impact performance? Or better use a Vref (like TL431) instead of R1?
- Some decoupling needed, for better performance and or stability? Speed-up capacitor from B+ output to gate of T2?
Attachments
Do you need 90V swing on the grid?
50V will reduce plate current to <5mA in a GU50
Changing R1 / R2 value will change the feedback ratio to a lower number...Will probably help response time.
i.e. You are using a 10V reference right now. Your feedback ratio is 39:1. [(300+90):10]
and your grid swing is close to 90V
If you change the R1 / R2 to 30K and 70K (30Vreference) then the feedback ratio is 12.33:1. [(300+70)/30] and your grid swing is 70V...still enough to drive the plate current low.
This would also reduce the power dissipation of T1, T2...can't be a bad thing.
Select a nominal operating point for T1 and T2 that gives them similar power dissipation for similar operating temperature. (play with R3 value)
IMHO
Some things to think about.
50V will reduce plate current to <5mA in a GU50
Changing R1 / R2 value will change the feedback ratio to a lower number...Will probably help response time.
i.e. You are using a 10V reference right now. Your feedback ratio is 39:1. [(300+90):10]
and your grid swing is close to 90V
If you change the R1 / R2 to 30K and 70K (30Vreference) then the feedback ratio is 12.33:1. [(300+70)/30] and your grid swing is 70V...still enough to drive the plate current low.
This would also reduce the power dissipation of T1, T2...can't be a bad thing.
Select a nominal operating point for T1 and T2 that gives them similar power dissipation for similar operating temperature. (play with R3 value)
IMHO
Some things to think about.
Many thanks for the reply! You are right that -90V is a lot, specially as I also mention the possible use of a 6P15P (for maybe a driver stage) that will cut off with less than -15V. This is more a concept, I think, and values of resistors can be changed according to final application. But I had overlooked the aspect of feedback ratio, which is a good one, thanks!
Another thing, based on tubecad: I would probably test with a 22k (or larger) resistor in series with the grid, and AC connect B+ to the grid with a cap, so that the tube handles AC that is coming from the amp part, while decoupling gate of T2 with an elco.
Another thing, based on tubecad: I would probably test with a 22k (or larger) resistor in series with the grid, and AC connect B+ to the grid with a cap, so that the tube handles AC that is coming from the amp part, while decoupling gate of T2 with an elco.
Last edited:
that is indeed the idea. The negative 100V, or less (in case less voltage is enough) will be provided by a shunt regulator, as demonstrated in Valve Amplifiers 4th edition, by Morgan Jones. I don't think, however, that it is right to publish that regulator here? But it is a very quiet one!
Hi Erik,
Yes. differential bipolar transistor error amplifiers are the way to go.
If the grid current of the pentode is say <10uA, you can run the differential pair at 500uA. This in turn will keep the base current low. The tail can remain a resistor if the -100V is regulated - it will be constant current anyway, and CMR is not a real problem.
As for transistors, please try the excellent NXP PBHV series of 500V parts in SOT23 and 1W SOT223:
PBHV8540Z :: NXP Semiconductors
A "lead" cap can be used from B+ to FB, but it may be necessary to compensate the diff pair with a cap, also.
Yes. differential bipolar transistor error amplifiers are the way to go.
If the grid current of the pentode is say <10uA, you can run the differential pair at 500uA. This in turn will keep the base current low. The tail can remain a resistor if the -100V is regulated - it will be constant current anyway, and CMR is not a real problem.
As for transistors, please try the excellent NXP PBHV series of 500V parts in SOT23 and 1W SOT223:
PBHV8540Z :: NXP Semiconductors
A "lead" cap can be used from B+ to FB, but it may be necessary to compensate the diff pair with a cap, also.
Hi Ian!
Compared to op-amp error servo:
1. no need for separate power supply, or risk of damage to opamp from HV.
2. open-loop gain and bandwidth can be finely controlled, so you can tailor the transient response to perfection. All stages can be separately frequency compensated.
3. cost is almost zero.
Hi Rod, thanks !
So, you would run a shunt regulator based on the NXP transitors across the HV power supply outputs without using an additional minus supply ?
I see in the above referenced tubecad article Broskie's well known 431 / tube hybrid shunt reg. and if I remember rightly, I think you said once that the 431 wasn't exactly the chip of your dreams (my paraphrase) so now I am wondering about the inverse. If you were going to use the BJT pair as the error amplifier, would you put them under a tube or would you go for a high voltage BJT, MOSFET etc. ?
So, you would run a shunt regulator based on the NXP transitors across the HV power supply outputs without using an additional minus supply ?
I see in the above referenced tubecad article Broskie's well known 431 / tube hybrid shunt reg. and if I remember rightly, I think you said once that the 431 wasn't exactly the chip of your dreams (my paraphrase) so now I am wondering about the inverse. If you were going to use the BJT pair as the error amplifier, would you put them under a tube or would you go for a high voltage BJT, MOSFET etc. ?
I use BJTs and enhancement-mode FETs for my shunt regulation, but if you use a triode, the grid supply must (of course) be negative (compared to its cathode anyway).
431 can work OK, but the noise generated by their internal bandgap reference is not helpful. It varies from one manufacturer to the next, which may make for individual testing (yuk).
I think the best path is to start with one of the existing public circuits for a SS shunt regulator, and see if there is anything wrong with it, for your application.
I do prefer an error amplifier though (but not op-amps) - for transient control over voltage - and I don't like noisy references!
431 can work OK, but the noise generated by their internal bandgap reference is not helpful. It varies from one manufacturer to the next, which may make for individual testing (yuk).
I think the best path is to start with one of the existing public circuits for a SS shunt regulator, and see if there is anything wrong with it, for your application.
I do prefer an error amplifier though (but not op-amps) - for transient control over voltage - and I don't like noisy references!
Thanks Rod,
Not looking for a design, I was just trying to learn what I could from your perspective. I'm inching towards a shunt reg that will regulate something just under 500VDC and while I have a couple of regultaor tubes that can stand the voltage my inclination is toward a simple transistor circuit if I can arrive at one that results in the sound I'm after. Just fishing for educational thoughts.
Not looking for a design, I was just trying to learn what I could from your perspective. I'm inching towards a shunt reg that will regulate something just under 500VDC and while I have a couple of regultaor tubes that can stand the voltage my inclination is toward a simple transistor circuit if I can arrive at one that results in the sound I'm after. Just fishing for educational thoughts.
Hi Rod,
thanks for your advice. I am 'in the field' now (living in Mozambique) and unable to build anything, so for the moment lots of reading, theory, and some exercises in how to tackle a Shunt regulator. But I am quite sure these circuit will be put together when I have access to my stuff again
Hi Suntechnik
Are you trying to say that I should use chokes instead of (shunt) regulators?
Reviving this old thread with some new info.
I still got this idea of building this shunt regulator, so I made some experiments today.
To start I traced some G2 driven curves of the GU50 with the utracer, and found the following set of curves, where G1 is at fixed -9V and G2 varies, very interesting for this application.
CCs would be set at 150mA max, and I think that 40mA through the GU50 during operation would be a good value, so I draw both points, to find that G2 will vary between 100 and 190V. So I made the regulator circuit, based on the LND150 depletion fets, and used 50V as reference voltage.
I can say it works, but haven't done extensive tests, yet. With G2 at 100V, which will be the case during operation, G2 pulls about 1mA of current, so the Mosfet feeding the G2 current doesn't need heatsinking. I also want to put a gridleak resistor in series with the 9V battery and a cap from B+ to grid, so the GU50 does some "Ripplerejection" work as well.
decision to change drive to G2 is mostly because it makes it possible to have everything powered from a single supply (with the exception of the -9V at G1, but well, a battery is a simple solution, otherwise three 3V zeners in serie with the cathode, and G1 to gnd).
I still got this idea of building this shunt regulator, so I made some experiments today.
To start I traced some G2 driven curves of the GU50 with the utracer, and found the following set of curves, where G1 is at fixed -9V and G2 varies, very interesting for this application.
CCs would be set at 150mA max, and I think that 40mA through the GU50 during operation would be a good value, so I draw both points, to find that G2 will vary between 100 and 190V. So I made the regulator circuit, based on the LND150 depletion fets, and used 50V as reference voltage.
I can say it works, but haven't done extensive tests, yet. With G2 at 100V, which will be the case during operation, G2 pulls about 1mA of current, so the Mosfet feeding the G2 current doesn't need heatsinking. I also want to put a gridleak resistor in series with the 9V battery and a cap from B+ to grid, so the GU50 does some "Ripplerejection" work as well.
decision to change drive to G2 is mostly because it makes it possible to have everything powered from a single supply (with the exception of the -9V at G1, but well, a battery is a simple solution, otherwise three 3V zeners in serie with the cathode, and G1 to gnd).
Attachments
Hi Ketje,
I am Dutch indeed An uncle did some research into our family name, and found both "de beste" (dutch for "the best") and "de beest" (dutch for "the beast). And then he got scared, and stopped his search
Thanks for the schematic! I have looked into something similar, well, actually I made a lot of sketches, but decided for the above circuit because it does really put all the heat on the GU50, and the SS stays cool. During operation the FET feeding G2 passes about 1.5mA at 200 to 300, so between 300mW and 450mW, which is comfortable for a TO220 (maybe I ll put a small heatsink on it).
Best, Erik
I am Dutch indeed
An uncle did some research into our family name, and found both "de beste" (dutch for "the best") and "de beest" (dutch for "the beast). And then he got scared, and stopped his search Thanks for the schematic! I have looked into something similar, well, actually I made a lot of sketches, but decided for the above circuit because it does really put all the heat on the GU50, and the SS stays cool. During operation the FET feeding G2 passes about 1.5mA at 200 to 300, so between 300mW and 450mW, which is comfortable for a TO220 (maybe I ll put a small heatsink on it).
Best, Erik
I expect if you run the numbers, you'll find that the overall system efficiency is worse than if you use a stable (or even regulated) amplifier in the first place.
That is to say: class A is inefficient to begin with, but as its bias current varies little (and could be servoed to keep stable, or even regulate the supply if desired, with a little effect on audio compression/expansion), it will do the same thing as you propose, without wasting an extra tube (and regulation overhead, and heater, and bias, and all that).
Even back in the toob days, everyone understood that pass regulation is just so much superior to shunt regulation, when any more than a few watts are at stake (i.e., whatever a 0A2 / 0D3 can handle). The overall efficiency still won't be much better, of course.
Tim
That is to say: class A is inefficient to begin with, but as its bias current varies little (and could be servoed to keep stable, or even regulate the supply if desired, with a little effect on audio compression/expansion), it will do the same thing as you propose, without wasting an extra tube (and regulation overhead, and heater, and bias, and all that).
Even back in the toob days, everyone understood that pass regulation is just so much superior to shunt regulation, when any more than a few watts are at stake (i.e., whatever a 0A2 / 0D3 can handle). The overall efficiency still won't be much better, of course.
Tim
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.