Ah... I should have known that you would build amps this way!
Do you use a Choke?
How much capacitance before and after the V-reg?
(I will need about 330V B+ @ 200mA for the 2A3 stereo amp)
What do you think about a zener diode vs. a tube for the voltage reference?
I need about 277V at the plate of the pentode, so with a say 12V zener on the cathode I can have the pentode working at it's "Typical" level.
For a stereo amp, the same B+ for both channels?
Thanks
Depends a lot on the design goal. I do use chokes even with pentode connected pass elements when I need the maximum ripple rejection possible. I also use chokes in most cases where the pass tube is a triode.
I use input capacitors that are of acceptable values for the rectifier used and rarely more than 47 - 50uF, with a choke based pi-filter I may use up to 100uF, depends on the choke used. I generally use either a choke or RC based filter for the screen supply and here I may use up to 100uF (or more).
I like small gas references like the 5651 and the 85A2 and use them with a cascode connected error amplifier (12AT7/6DJ8, also the 12AX7A which is not designed for cascode connection but provides acceptable performance IMV for this application.)
There is a thread with one of my cascode error amp based regulators here:
http://www.diyaudio.com/forums/tubes-valves/90020-perfect-tube-regulator.html (Post #9) Note that although a 6BQ5 is shown a 6550 or two will drop in with minor adjustments at most.
Related: http://www.diyaudio.com/forums/tube...ce-tube-5651-85a1-etc-internal-impedance.html
A properly designed regulator will give you a very low noise output suitable for use with a dht based amplifier w/o gfbk and great stability. Ripple rejection of >60dB is possible with care, and stability with the gas tube is good. Regulation both load and line can be 1% or better depending on available loop gain and quality of the reference. Zeners interestingly drift more than the 5651 IME, but their low dynamic impedance promises slightly better ac performance. (Not enough to discourage me from almost exclusively using the 5651 as a reference where the voltages needed are conducive.)
I like separate regulators for each channel, that said you can parallel a couple of 6550 to get the current you need or even use a single 6336 as the pass tube. (Use a 5H choke based pi filter ahead of it for lower overall output ripple.)
I use input capacitors that are of acceptable values for the rectifier used and rarely more than 47 - 50uF, with a choke based pi-filter I may use up to 100uF, depends on the choke used. I generally use either a choke or RC based filter for the screen supply and here I may use up to 100uF (or more).
I like small gas references like the 5651 and the 85A2 and use them with a cascode connected error amplifier (12AT7/6DJ8, also the 12AX7A which is not designed for cascode connection but provides acceptable performance IMV for this application.)
There is a thread with one of my cascode error amp based regulators here:
http://www.diyaudio.com/forums/tubes-valves/90020-perfect-tube-regulator.html (Post #9) Note that although a 6BQ5 is shown a 6550 or two will drop in with minor adjustments at most.
Related: http://www.diyaudio.com/forums/tube...ce-tube-5651-85a1-etc-internal-impedance.html
A properly designed regulator will give you a very low noise output suitable for use with a dht based amplifier w/o gfbk and great stability. Ripple rejection of >60dB is possible with care, and stability with the gas tube is good. Regulation both load and line can be 1% or better depending on available loop gain and quality of the reference. Zeners interestingly drift more than the 5651 IME, but their low dynamic impedance promises slightly better ac performance. (Not enough to discourage me from almost exclusively using the 5651 as a reference where the voltages needed are conducive.)
I like separate regulators for each channel, that said you can parallel a couple of 6550 to get the current you need or even use a single 6336 as the pass tube. (Use a 5H choke based pi filter ahead of it for lower overall output ripple.)
Last edited:
A good point. Mains voltage variations (level) and quality (contamination) seem to change a lot throughout the day where I live. I think a regulator and decent inlet filtering help bring consistency. I see it as a foundation upon which to build a decent design.
Well I’m leaning toward using the Voltage Regulator in the next amp.
Without the regulator, the B+ of my current amp varies no more than 5V….. which I don’t think is an issue for a 312V B+ (sounds like I have some clean AC compared to other countries)
With the regulator, it’s always 312V, as expected.
I have attached the schematic of the regulator I’m testing on my 2A3 monoblock.
The monoblock psu is LC-RC. To test the regulator, I add a small cap (2.2uf) in front of the choke to increase my “raw” B+, and then replace the resistor of the RC stage with the regulator.
Without the regulator, the B+ of my current amp varies no more than 5V….. which I don’t think is an issue for a 312V B+ (sounds like I have some clean AC compared to other countries)
With the regulator, it’s always 312V, as expected.
I have attached the schematic of the regulator I’m testing on my 2A3 monoblock.
The monoblock psu is LC-RC. To test the regulator, I add a small cap (2.2uf) in front of the choke to increase my “raw” B+, and then replace the resistor of the RC stage with the regulator.
I think regulated supplies help a lot with these issues. My amps are immediately listenable, but I do have impression a lot of the time that sound does improve slightly with some warm up. IMO This however is not nearly as pronounced as I recollect with some of my earlier designs that did not feature regulation.
My main reason for using regulators is to achieve supplies with low source impedance, low noise and minimal variation in output voltage in order to have predictable operating points under almost all conditions. (Irrationally perhaps don't often use current sources as I like chokes, transformers and resistive loads generally better. I generally use fixed bias as well - regulated bias supplies of course.)
Your design is quite similar in some ways to what I am doing. I have found that bootstrapping a gas tube off of the same supply as the screen is a recipe for oscillation, but it works fine with zeners because of the much lower impedance. I would bootstrap the gas tube off of the output and create a new and independent voltage divider for the screen - note that it should be decoupled to ground.
IF you leave it as is you still need to decouple the screen to ground in order to get adequate ac performance out of the error amplifier, and because of the low values of the resistors used it will need to be large. (Probably on the order of 22uF - bypass with a good film.)
It is worthwhile to supply power to the error amplifier plate circuit from a well filtered supply running on the unregulated supply - this allows much larger plate resistors to be used at a given current and with your pentode error amplifier should result in more loop gain.
It's not intuitive, but watch out for the LF pole you have introduced in the feedback by shunting the feedback resistor network with that 0.68uF cap.
Also I abhor the use of pots, if you must restrict its range to roughly +/-5% or so of the desired output voltage. (resistor on top, resistor on bottom, pot in middle.) Normally if you have a precision reference and enough loop gain you won't need it. The reference/cathode currents summed are a bit too high overall for the 5651. Sadly a high power zener may perform better than the gas tube, but it does look cool. Note that most 10T pots are not specified for high voltages and I think that could be a bit of trouble waiting to happen. (But I could also be wrong)
The 0.47uF across the gas tube is pushing your luck - caps across these things result in relaxation oscillations in many instances, and subsonic frequencies modulating the output of the supply may trigger it.
I have covered pretty much everything I can think of here, and hope it helps. The basic design you have come up with is capable of excellent performance and unconditional stability with just a little more work!
Last edited:
First of all, Thanks for all the help so far.
Here is a new schematic based on Kevin’s recommendations:
1. Pentode Plate voltage from the Raw B+ via an additional RC filter
2. G2 voltage via a voltage divider to ground, with the lower resistor bypassed with 10uf
3. G1 voltage via voltage divider (pot removed)
4. .47uf cap removed from the 0A3
It will be interesting to see if the Z-out of this regulated PSU is any better than the others I tested.
Stay tuned.
Here is a new schematic based on Kevin’s recommendations:
1. Pentode Plate voltage from the Raw B+ via an additional RC filter
2. G2 voltage via a voltage divider to ground, with the lower resistor bypassed with 10uf
3. G1 voltage via voltage divider (pot removed)
4. .47uf cap removed from the 0A3
It will be interesting to see if the Z-out of this regulated PSU is any better than the others I tested.
Stay tuned.
I would also be very interested in the Zout and other characteristics of this regulator to have some comparison material for my own regulator.
jd
Jan, I couldn't find Zout vs frequency plots and/or step response for your regulator. Can you please point to where I could find them?
(As I have a moment...)
Zout here:
http://www.linearaudio.nl/Documents/T-reg fig 3 zout all.pdf
Linked from this page:
http://www.linearaudio.nl/t-reg-1.htm
Here is what I’ve learned after about 6 hours of experiments
1. Highly recommend a pot to adjust pentode G1 Voltage. From B+ to ground I used 1M-100Kpot-270K
2. Have to have a cap from the B+ to the pentode G1, ended up replacing the .47uf with .1uf
3. Don’t seem to need the cap across the 0A3…. Works fine without.
4. For the pentode G2 voltage, I had the same results with the G2 supply via a voltage divider that goes from B+ to the 0A3 as
I did with a separate voltage divider from B+ to ground
5. The biggest improvement was taking the pentode plate supply from the Raw B+
Now for the numbers:
Freq(hZ)-Zout
20-11.00
50-7.50
100-6.25
200-5.00
500-2.75
1000-1.32
2000-.65
5000-.22
With the regulator, I can dial in the exact B+ needed to get 295V at the plate of the 2A3, and it stays at 295V all the time.
I’m definitely using a regulator for the next amp.
1. Highly recommend a pot to adjust pentode G1 Voltage. From B+ to ground I used 1M-100Kpot-270K
2. Have to have a cap from the B+ to the pentode G1, ended up replacing the .47uf with .1uf
3. Don’t seem to need the cap across the 0A3…. Works fine without.
4. For the pentode G2 voltage, I had the same results with the G2 supply via a voltage divider that goes from B+ to the 0A3 as
I did with a separate voltage divider from B+ to ground
5. The biggest improvement was taking the pentode plate supply from the Raw B+
Now for the numbers:
Freq(hZ)-Zout
20-11.00
50-7.50
100-6.25
200-5.00
500-2.75
1000-1.32
2000-.65
5000-.22
With the regulator, I can dial in the exact B+ needed to get 295V at the plate of the 2A3, and it stays at 295V all the time.
I’m definitely using a regulator for the next amp.
My obsession with this continues and tonight I made a breakthrough.
Z-out at 20Hz is down to 3.5R, yesterday I reported 11R.
(my LC-RC psu is around 50)
Here are the changes:
Lower the pentode G2 Voltage so that we run the pentode on much less current, allowing for a much larger resistor on the plate of the pentode.
That’s the key.
To do this, I changed the pentode G2 supply to a voltage divider from B+ to ground (also learning that Kevin Knows Best)
And the 0A3 resistor is now 10K (was 20K) to keep the 0A3 running @ about 25mA (this includes the current from the pentode).
Schematic to follow
Z-out at 20Hz is down to 3.5R, yesterday I reported 11R.
(my LC-RC psu is around 50)
Here are the changes:
Lower the pentode G2 Voltage so that we run the pentode on much less current, allowing for a much larger resistor on the plate of the pentode.
That’s the key.
To do this, I changed the pentode G2 supply to a voltage divider from B+ to ground (also learning that Kevin Knows Best)
And the 0A3 resistor is now 10K (was 20K) to keep the 0A3 running @ about 25mA (this includes the current from the pentode).
Schematic to follow
LOL, yeah. A 0.01 ohm PSU for a 2A3 SET (just for example) just has to sound sooo much better than 1 ohms.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.