Very simple variable HV psu with short circuit protection

[rotation]

Hi!

I just built myself variable hv psu for experiments with tube preamps, schematic is here:
http://www.geocities.com/tjacodesign/300vsup/300vsup.html
It works very well, specially part with short circuit protection Now i would like to understand this circuit a little better and add something that would make it work as a regulator.
I don't understand what is the purpose of fet here. Is it there only to make psu variable or it adds some regulation too?
Also, heatsink on fet is small piece of 2mm aluminium (5x8cm) which i got from old tv where it was used as a heatsink too. It gets quite hot, dissipation from fet is around 5W (300V/15mA). Should i change heatsink to something bigger?
Also, at the moment i only have one 150u cap, as on schematic. Do i need more capacitance to make ripple lower? And should i put it near C1 or at the end of circuit?

I managed to find another schematic where the same fet is also wired as a source follower:
http://www.welbornelabs.com/ps3.htm
Is it possible to add zeners to my existing psu in a simple way? Or can i do anything else to make it work as a regulator? I know, i could use something like G9's psu but i don't want to use low voltage ic-s for hv experiments. I would also like to keep it simple because i never use pcbs for tubes..

Yeah, i'm asking a lot. If i at least get responses to first part of this post, it would help me a lot.

Miha

 

[PRR]

> what is the purpose of fet here

Take it out. What happens?

The NO-load output is still variable, but ANY load makes it sag, and any useful load makes it sag to about zero.

The pot is an adjuster, but very low power.

The FET can pass power, and will pass what the pot tells it to pass, without loading the pot.

> make ripple lower?

Put about 0.5uFd from Gate to ground. Ripple vanishes. Explain why a small cap at gate does so much more than a big cap at input or output.

> heatsink ... 5x8cm ... gets quite hot

Hot enough to boil spit? Hot enough to melt solder? 100C is not intolerable for an experimenter supply. Silicon will survive solder-melt temperature for minutes.

> 5x8cm ... around 5W

In Missouri, we'd say 2 by 3 inches, 12 square inches both sides. Figure 100 degree C per watt per square inch. 12 square inches is 8 deg C per watt. 5 Watts is then 40 deg C temperature rise, say 65 deg C in a typical room. Or in Missouri, 150 deg F. That is not hot for Silicon, although it may be painful to the finger (it will actually burn flesh).

> possible to add zeners

Pot gives an adjustable fraction of input voltage. Zeners attempt to give a fixed voltage. Which do you want? Adjustable or fixed?

On an experimeter's bench, and certainly for any audo tube work, you have a voltmeter handy and 20% variation will not upset a vacuum tube. Set it to 250V, start experimenting, occasionally check that the supply is still about 250V.

In my shop, the 115V-108V wall voltage would make the "250V" wander 242V to 258V. The tube won't care (aside from +/-3% variation in maximum output, and who cares?). Since I would never regulate a vacuum tube audio amplifier, that is the way it will run in Real Life, and is entirely valid.

OTOH, my Boonton AC milliVoltmeter does use regulated supplies, because someone (not me) might complain about 1% errors in readings.

 

[sahib]

Also, at the moment i only have one 150u cap, as on schematic. Do i need more capacitance to make ripple lower? And should i put it near C1 or at the end of circuit?

Miha

Yes the higher capacitance in the input or output  will reduce ripple. But you can do that with a cost effective trick. PRR asked the question;

Put about 0.5uFd from Gate to ground. Ripple vanishes. Explain why a small cap at gate does so much more than a big cap at input or output.

Again he said the fet will pass power according to what the pot says. So the wiper of the pot must be supplying reference voltage to the gate of the fet. If you add a capacitor accross that reference, you will have a filtering effect (reduced ripple) on it. That effect will also be on the gate of the fet. Now there is the magic. The fet will multiply that effect due to the current gain and make the output appear as if there is a much larger capacitor accross it. This is called a capacitance multiplier.

 

[bcarso]

Strongly recommend you use a potentiometer with a plastic shaft for R3.

 

[rotation]

Hi, thanks a lot for explanations. I will read more about capacitance multiplier, it's a good idea..

I have some other questions; if i wanted to make a fixed supply and replace the pot with fixed resistor, should i measure resistance of the pot, add resistor instead and solder R1 from ground to gate? Or should i use two resistor instead of pot? I'm not sure about it.

Also, do we use regulation only to get fixed voltage, or are there any other reasons? Many people claim that regulation of anode supply for tube preamps sounds different from classic, well filtered RC supply. Could this be truth or is it another "internet invention"?

Miha

 

[rotation]

Strongly recommend you use a potentiometer with a plastic shaft for R3.

Ok, i will do it, i have some at hand. Thanks!

 

[sahib]

R1 stays there.

The pot is a voltage divider network so you can measure the high and low side resistances and replace it with fixed resistors. But why bother, you may struggle to find exact values. Therefore using a trimpot does not add to cost significantly.

In terms of regulated anode supply's sonic difference, I can not say very much. I never had the desire of making such comparisons but I heard and read the same comments from well established engineers. It is even claimed that increasing the reservoir capacitors on the power supply of a power amp will help tighten its bass response . It may well be true to a degree as increasing the capacitors in parallel will reduce the power supply impedance and may have an effect on the output signal swing. Yet, on the other hand Douglas Self says otherwise in his power amp design book. Though that refers to solid state amps. I would like to hear masters' view on this.

 

[Piedwagtail]

He doesn't appear well received here, though the criticsm seems to come from non-empirical scientists but you can fill your boots with
schema and knowledge here:

http://www.tubecad.com/2006/11/blog0087.htm
http://www.tubecad.com/2007/01/blog0095.htm
http://www.tubecad.com/2007/06/blog0109.htm
http://www.tubecad.com/2007/06/blog0110.htm
http://www.tubecad.com/2007/06/blog0112.htm
http://www.tubecad.com/2007/07/blog0113.htm
http://www.tubecad.com/2007/07/blog0115.htm

none are variable, there again Mr.Bench does a very nice expose on series regulators
http://greygum.net/sbench/sbench/reg1.html


Robert

 

[sahib]

I have had that web site bookmarked for years and I quite enjoy it but thanks again for the links. I had a quick read of them.

As I indicated Mr Self's comment related to solid state.  Although I am not taking his comments as the word of Bible he has a point. When a THD level of 0.0026% @ 20kHz on Class B seems well achieved with unregulated power supplies, it is quite reasonable to question what advantage can be gained with regulated types.

However, I am aware of the regulated power supply applications on valve amps, particularly the hi-fi. Unfortunately I don't have very much time to spend on experimentation. I have also not been so fortunate to work in an environment where high calibre engineers would be available to answer my questions (this forum excluded). So I heavily rely on books to find answers to my questions. But the difficulty with the books is that I have to work very hard to pinpoint and seperate the subjective views. And I can assure you, there are loads of dodgy comments from very well established engineers and writers.

 

[rotation]

In terms of regulated anode supply's sonic difference, I can not say very much. I never had the desire of making such comparisons but I heard and read the same comments from well established engineers.
I would like to hear masters' view on this.

Hi,
i'm not an expert but i just found two simple regs that are evry cheap to build and can be added to existing unreg psu. With this we can try reg vs unreg sound (if there is any difference at all) ourselfs:
http://www.jlmaudio.com/HVreg.gif
I'm thinking about second because i already have all material. It looks like 10k resistor is there to get some current through zeners (35mA?) to make them work.
I'm not sure what is 10u cap doing there, it might be there to take away zener noise or as capacitance multiplier (PRR,, i've done my homework, thanks). I would say it's capacitance multiplier because it's also on the first circuit which don't have zeners..
Can please someone tell me if i only have to recalculate 10k resistor if i want to make this circuit work with around 310VDC input?
Also, is first circuit really regulator? What is 330k resistor doing there?

Miha

 

[Curtis]

Can please someone tell me if i only have to recalculate 10k resistor if i want to make this circuit work with around 310VDC input?

The 10K resistor sets the current for the zener diodes so that they can start operating.  At 350V with 300V worth of zeners you're getting (350-300)/10000 = 5mA down the zener string.  With a 310V input you'll have (310-300)/10000 = 1mA, only a fifth of the zener current, probably a little light on.  Scaling the resistor by a fifth will get you back to 5mA.

310V input on a 300V out regulator is a little lean though, you will find that it will stop regulating properly given enough load.  Lowering the output voltage to 250V-ish will give better performance - gives a bigger safety margin.

Also, is first circuit really regulator? What is 330k resistor doing there?

Yep, but not as good as the second. 

The 330K resistor is just another way of setting a reference voltage for the regulator, it's just a resistive voltage divider (330K/(330K+22K))*350V=330V, although regulator output voltage will vary more with input voltage than the zener method.  The cap is there to help smooth out any blips that appear on the input voltage and provide a more stable reference voltage for the transistor.

 

[rotation]

Can please someone tell me if i only have to recalculate 10k resistor if i want to make this circuit work with around 310VDC input?

350V with 300V worth of zeners you're getting (350-300)/10000 = 5mA down the zener string.  With a 310V input you'll have (310-300)/10000 = 1mA, only a fifth of the zener current, probably a little light on.  Scaling the resistor by a fifth will get you back to 5mA.

The 10K resistor sets the current for the zener diodes so that they can start operating.  At
310V input on a 300V out regulator is a little lean though, you will find that it will stop regulating properly given enough load.  Lowering the output voltage to 250V-ish will give better performance - gives a bigger safety margin.

Also, is first circuit really regulator? What is 330k resistor doing there?

Yep, but not as good as the second. 

The 330K resistor is just another way of setting a reference voltage for the regulator, it's just a resistive voltage divider (330K/(330K+22K))*350V=330V, although regulator output voltage will vary more with input voltage than the zener method.  The cap is there to help smooth out any blips that appear on the input voltage and provide a more stable reference voltage for the transistor.

Hi,

aren't those zeners 5w type that need at least 20mA to start working? Can i use lower watage zeners here? Also, how do i calculate needed current if i have more zeners in series (when i can't find 2 150v when i need 300V)?

Miha