Hello all,
As impressed as I was from the subjective performance of my recent low voltage shunts for my NJFET RIAA (link), I developed high voltage ones for my tube gear too. The first test has been done powering the Steve Bench cascode 12AY7 parallel & 6DJ8 RIAA. A phono circuit is the most sensitive one so to fully test a reg PSU IMHO. The previous reg was a Mosfet Maida. In brief: 1. Lower noise than with the Maida, in CDP background silence territory now. 2. Bass definition and slam, image focus, image size, mid and high smoothness, one class higher with the Mosfet shunt. Maybe a 25% overall subjective performance lift above the Mosfet Maida.
Read the high voltage safety sticky thread if not familiar with tubes and tube supplies practices. A useful reminder for experts even:
Safety practices for high voltage
Updates not to search in a long thread:
SSHV1 Standard
SSHV2
Troubleshooting example
As impressed as I was from the subjective performance of my recent low voltage shunts for my NJFET RIAA (link), I developed high voltage ones for my tube gear too. The first test has been done powering the Steve Bench cascode 12AY7 parallel & 6DJ8 RIAA. A phono circuit is the most sensitive one so to fully test a reg PSU IMHO. The previous reg was a Mosfet Maida. In brief: 1. Lower noise than with the Maida, in CDP background silence territory now. 2. Bass definition and slam, image focus, image size, mid and high smoothness, one class higher with the Mosfet shunt. Maybe a 25% overall subjective performance lift above the Mosfet Maida.
Read the high voltage safety sticky thread if not familiar with tubes and tube supplies practices. A useful reminder for experts even:
Safety practices for high voltage
Updates not to search in a long thread:
SSHV1 Standard
SSHV2
Troubleshooting example
Attachments
Last edited:
Keep the constant current source running at least on double the mA that your application circuit needs in steady state. Triple is nicer.
Its all a matter of how big a heat sink you can use. The current used by your audio circuit is not heating the PSU shunt Mosfet. It heats at its most (Vout*CcsIma) when unloaded. The upper Mosfet can hold up to 200V Vin-Vreg. 50V difference is OK to run at its smoothest though, nobody wants to discard more just heating up the P-Mosfet. It heats for Vdif*IccsmA. It can stabilize with just 10V Vdif non the less, but it will be compromised somewhat for transients. R1 sets the CCS current, Iccs=(Vleds-VgsQ1)/R1.
The reg voltage reference is made up by 2 BC560 CCS feeding a bypassed resistor voltage drop. You can use a fixed R5 or the +trimmer arrangement for a variable shunt. Be careful there not to go beyond 200V Vin-Vout, if you like your nice P-Mosfet alive.
Great performance in practice, tried to keep em practical with easy to source components. 2N6520 is a -350V PNP and KSP94 is an alternative -400V PNP for the higher voltage shunt, as 2SA1625 is. That is if you wanna go to that 360V mark or go to a higher range. The IRF840 works up to 500V.
Regards
Salas
Its all a matter of how big a heat sink you can use. The current used by your audio circuit is not heating the PSU shunt Mosfet. It heats at its most (Vout*CcsIma) when unloaded. The upper Mosfet can hold up to 200V Vin-Vreg. 50V difference is OK to run at its smoothest though, nobody wants to discard more just heating up the P-Mosfet. It heats for Vdif*IccsmA. It can stabilize with just 10V Vdif non the less, but it will be compromised somewhat for transients. R1 sets the CCS current, Iccs=(Vleds-VgsQ1)/R1.
The reg voltage reference is made up by 2 BC560 CCS feeding a bypassed resistor voltage drop. You can use a fixed R5 or the +trimmer arrangement for a variable shunt. Be careful there not to go beyond 200V Vin-Vout, if you like your nice P-Mosfet alive.
Great performance in practice, tried to keep em practical with easy to source components. 2N6520 is a -350V PNP and KSP94 is an alternative -400V PNP for the higher voltage shunt, as 2SA1625 is. That is if you wanna go to that 360V mark or go to a higher range. The IRF840 works up to 500V.
Regards
Salas
Attachments
Sinks
In this (regrettably washed out) picture, you can get an idea of the heat sinking needs. At right hand side from about middle line and down to the toroid, you see the Mosfet shunt board. In this case a 230V tension is applied at the input, 60mA constant current, and 20mA idle load at 190Vreg by the Steve Bench cascode RIAA for 2 channels. First little CCS sink next to the LEDS receives 2.4W heat. Second aluminum color big sink receives 7.6W from the shunt N-Mosfet under load. Could stand 12W if unloaded. 120 Ohm R5+trim, fixed those conditions with an MPSA92 shunt driver BJT. Because different BJT types can give more output voltage, it is good practice to set the R trimmer 100 Ohm, half way for a 50 Ohm value and then power up for the first time. By measuring the VregOut you will be in a position to move up or down more knowledgeably and conservatively. Use a multi turn trimmer. Remember that the less the R5+R trimmer value, the more the V out. Also a 20W 10-15k dummy load resistor will be helpful for testing and setting the regulator without oscillation dangers and then power off, let it bleed all charges
, disconnect the dummy, and everything is ready so to attach the real load circuit, power on, and enjoy.
In this (regrettably washed out) picture, you can get an idea of the heat sinking needs. At right hand side from about middle line and down to the toroid, you see the Mosfet shunt board. In this case a 230V tension is applied at the input, 60mA constant current, and 20mA idle load at 190Vreg by the Steve Bench cascode RIAA for 2 channels. First little CCS sink next to the LEDS receives 2.4W heat. Second aluminum color big sink receives 7.6W from the shunt N-Mosfet under load. Could stand 12W if unloaded. 120 Ohm R5+trim, fixed those conditions with an MPSA92 shunt driver BJT. Because different BJT types can give more output voltage, it is good practice to set the R trimmer 100 Ohm, half way for a 50 Ohm value and then power up for the first time. By measuring the VregOut you will be in a position to move up or down more knowledgeably and conservatively. Use a multi turn trimmer. Remember that the less the R5+R trimmer value, the more the V out. Also a 20W 10-15k dummy load resistor will be helpful for testing and setting the regulator without oscillation dangers and then power off, let it bleed all charges
, disconnect the dummy, and everything is ready so to attach the real load circuit, power on, and enjoy.
JC951t said:
Yes no problem, and a CLC filter or RC before it if you like.
SY said:
It can be done. The particular P-Mosfet though proved stable and good for noise. Its a simple concept, it can easily be modified & implemented to suit different part inventories and precautions. That is the beauty of it. Build a tailor one and listen to it SY. Highly recommended.
Not after. It is totally noiseless as it is and you will ruin its transient advantages that way. It is already correctly terminated.
Before it is the right thing, so to pre filter noise passively up high.
Actually you can see a 10H frame choke in the left side of the PSU box pictured above. Part of an input CLC.
If you have read my previous directions, you will see that the constant current is defined by the total Led string voltage drop minus the Q1 Vgs divided by R1. For the IRFP9240, Vgs at such voltage conditions is about 3.1-3.2V. Measure the string Led in series with a 10k resistor using a 9V battery. Note their total voltage drop. Then I=(LedsVdrop-3.1)/R1. In practice it will deviate around that, but you are in the ballpark.
Before it is the right thing, so to pre filter noise passively up high.
Actually you can see a 10H frame choke in the left side of the PSU box pictured above. Part of an input CLC.
If you have read my previous directions, you will see that the constant current is defined by the total Led string voltage drop minus the Q1 Vgs divided by R1. For the IRFP9240, Vgs at such voltage conditions is about 3.1-3.2V. Measure the string Led in series with a 10k resistor using a 9V battery. Note their total voltage drop. Then I=(LedsVdrop-3.1)/R1. In practice it will deviate around that, but you are in the ballpark.
http://www.diyaudio.com/forums/attachment.php?s=&postid=1684277&stamp=1229193896
Hi Salas,
Very nice work. Exactly what I was looking for my tube preamp (on the works)
I have a question about the input,
How does one get 400V rectified and filtered? can you give an example?
I mean what transformer did you use?
Thanks
Hi Salas,
Very nice work. Exactly what I was looking for my tube preamp (on the works)
I have a question about the input,
How does one get 400V rectified and filtered? can you give an example?
I mean what transformer did you use?
Thanks
chatziva said:
P.S. Is this the preamp you are talking about? If yes, the PSU examples that you see here are overkill for current, dissipation - heat sinking. So, I instruct you a tailor made power scaled schematic if I may. You can see it attached. A 230V 100mA transformer and 4X UF4007 arranged as a bridge rectifier feeding a 220uF/350V capacitor is all you need for Vin. Also 2X rudimentary 3W heat sinks like those they mount LM317T regulators on.
Attachments
JC951t said:
Stixx is correct.
Also if there isn't any special Wattage need for R6, you can use 1/4W. C3 is a dead end for DC, so virtually no energy consumption on R6.
Together they form a Zobel terminal block that helps transient overshoot performance. See how it overshoots at IRF840's gate for current without the Zobel.
