I'm on a quest: listen to and compare tube types and different bias currents.
I have some hardware (ultra-linear push-pull transformers and a 350V power supply). My goal is to build a push-pull output stage with octal sockets so I can listen to some 6L6, EL34, and KT88 tubes. The driver stage for the output stage will be all solid-state. The arguments for solid-state is a whole different matter, so I'd like to concentrate on the output stage.
Since the intention is to make a push-pull output stage that can handle a variety of tube types and (potenitally) mismatched tubes, the bias adjustment is critical. Push-pull amps can suffer when the bias currents are even a few mA mismatched.
In addition, the amp needs to be able to operate in lean class AB.
So I've committed most of the sins: using solid-state, push-pull output stage, and running in class AB. Well, consider this a learning opportunity.
Here's the question: what is the best way for the tubes to self-bias? Since I am willing to use solid-state, the tube can bias using op-amp servo loops, constant current stages, or whatever. Cathode bias or grid bias. The only requirement is it must be applicable to class AB operation.
I've posted some of my thoughts (including schematics) on my web site (under Tube Amplifier - Schematics).
http://www.stephenwmoore.com
Take a look. What's the best way to get a pair of mismatched tubes to get along, then yank 'em out, and plug in yet another set of mismatched tubes, and expect 'em to work right away?
I have some hardware (ultra-linear push-pull transformers and a 350V power supply). My goal is to build a push-pull output stage with octal sockets so I can listen to some 6L6, EL34, and KT88 tubes. The driver stage for the output stage will be all solid-state. The arguments for solid-state is a whole different matter, so I'd like to concentrate on the output stage.
Since the intention is to make a push-pull output stage that can handle a variety of tube types and (potenitally) mismatched tubes, the bias adjustment is critical. Push-pull amps can suffer when the bias currents are even a few mA mismatched.
In addition, the amp needs to be able to operate in lean class AB.
So I've committed most of the sins: using solid-state, push-pull output stage, and running in class AB. Well, consider this a learning opportunity.
Here's the question: what is the best way for the tubes to self-bias? Since I am willing to use solid-state, the tube can bias using op-amp servo loops, constant current stages, or whatever. Cathode bias or grid bias. The only requirement is it must be applicable to class AB operation.
I've posted some of my thoughts (including schematics) on my web site (under Tube Amplifier - Schematics).
http://www.stephenwmoore.com
Take a look. What's the best way to get a pair of mismatched tubes to get along, then yank 'em out, and plug in yet another set of mismatched tubes, and expect 'em to work right away?
317 and 337 are no-nos. They can make excellent DC constant current sinks, but their AC performance is poor. Grid bias is the way to go. Use a 1R sense resistor in each cathode circuit, an op-amp to give some gain and perhaps another to integrate, then use a high voltage transistor to control the (possibly -100V) bias supply. It's more complex than your solutions, but would allow any bias current you liked and would be self-balancing.
As requested, I won't comment on the SS drive to the valves.
As requested, I won't comment on the SS drive to the valves.
By "grid bias" I assume you mean "fixed bias"?
Why not just use a single, adjustable wirewound resistor? Ohmite makes a sliding one with a screw. They are 15 watts I believe. When you switch tubes, you just slide to the right value.
As a side note, I think you're going to be sadly disappointed at just how similar all the stages are going to sound.
Why not just use a single, adjustable wirewound resistor? Ohmite makes a sliding one with a screw. They are 15 watts I believe. When you switch tubes, you just slide to the right value.
As a side note, I think you're going to be sadly disappointed at just how similar all the stages are going to sound.
Unless you do something fancy, the sense-adjust type autobias circuits will not work for AB operation. You have to either build in a means of sensing when there's signal and having the circuit automatically hold the last value (a uP type circuit is probably the best implementation), or do a variation of Chater's circuit for solid state amps (Audio Amateur, 2/88).
The best solution is to bite the bullet and rebias for each tube change.
The best solution is to bite the bullet and rebias for each tube change.
... two quick thoughts...
1)
I did a comparision between cathode bias, diode bias (diode in the cathode circuit) and fixed (grid bias) and with DHT (2A3s) or IDHT (5B/255 aka 807s) and found that fixed bias sounds best - so I would go for fixed bias - not servo'd as the current through the cathode or anode sense resistor will vary with signal level (it's Class AB)...
And I agree with Joel that they will all sound much the same given the circuit you are suggesting using them in (I didn't mention ss
)
2) You should investigate your point on push-pull dc balance ... in practice best sound usually occurs with 10-20% dc imbalance (make sure your OPT can handle the dc current offset....then you can hear this...)
ciao
James
1)
I did a comparision between cathode bias, diode bias (diode in the cathode circuit) and fixed (grid bias) and with DHT (2A3s) or IDHT (5B/255 aka 807s) and found that fixed bias sounds best - so I would go for fixed bias - not servo'd as the current through the cathode or anode sense resistor will vary with signal level (it's Class AB)...
And I agree with Joel that they will all sound much the same given the circuit you are suggesting using them in (I didn't mention ss
)2) You should investigate your point on push-pull dc balance ... in practice best sound usually occurs with 10-20% dc imbalance (make sure your OPT can handle the dc current offset....then you can hear this...)
ciao
James
OT, sorry
James, thanks, I try to change it every week or two. Mencken was from my hometown and remains, to this day, one of my personal heroes. Gene Kelly did a crappy job of portraying him in "Inherit the Wind." He was a nasty, cynical guy, who always took the dark view of humanity and spared no-one's feelings- my kind of fellow.
Back on topic, I'm not sure that I'd worry about AC effects, output impedances or whatever of the active circuit as long as what it's doing is controlling the grid- remember, you're firing voltage through a pretty big grid resistor. If a regulator's output impedance rises to something enormous (like 10 ohms), balance that against the notion that the grid resistor is 50-100K ohm. So that source Z is pretty well swamped.
James, thanks, I try to change it every week or two. Mencken was from my hometown and remains, to this day, one of my personal heroes. Gene Kelly did a crappy job of portraying him in "Inherit the Wind." He was a nasty, cynical guy, who always took the dark view of humanity and spared no-one's feelings- my kind of fellow.
Back on topic, I'm not sure that I'd worry about AC effects, output impedances or whatever of the active circuit as long as what it's doing is controlling the grid- remember, you're firing voltage through a pretty big grid resistor. If a regulator's output impedance rises to something enormous (like 10 ohms), balance that against the notion that the grid resistor is 50-100K ohm. So that source Z is pretty well swamped.
I am trying to combine the best attributes of technology for this output stage. My primary interest is creating a care-free output stage, and learning a lot while doing it. The whole point of DIY for me is to learn new things, and explore odd (or new) ways of thinking about problems and solutions.
I have presented the problem: mismatched tubes.
Since I can adjust the bias by hand, this tells me that I can design some electronics to do it for me.
Using a cathode current sense resistor is obvious. Will it work with class AB? I was assuming it would, if the sense resistor was bypassed with a capacitor. That way, all the AC would go around the servo circuit, leaving the servo to worry about maintaining DC bias.
Making the servo circuit a low-pass filter would also help, because it would only adjust the bias current in response to very low frequency changes. Since this output stage is for a tweeter in an active-crossover system, the only AC the output stage would ever see would be above 1kHz.
In my opinion, the push-pull output stage has been underdeveloped. All the technology and literature today is based on designs from 50 or more year ago. As new production tubes become more available these days (considering I can buy a new 6L6 from at least 4 different manufacturing sites around the world), I believe we have not seen the last of the push-pull, nor have we seen the developments and maturity that the push-pull output stage will eventually show.
Other comments:
- My push-pull transformers are quite cheap. I don't think they can handle much bias differential. If I am even slightly pleased with the sound, I will purchase nice new OPTs.
- I'm not on a quest to hear differences in output stage configurations, so I'm not setting myself up for disappointment. I am on a quest to have a robust amplifier that can put up with a lot of abuse.
- I'm not scared of lots of circuitry. Many circuit components can make design easier, because less is demanded from each piece. However, it does increase the chance that the negative attributes of each piece can accumulate and make lousy sound.
- I've already lost a very expensive tweeter to a solid-state amp. That's the reason I am going for tubes. The transformer-coupled output stage protects from deadly DC offset, and I can use capacitors in the output stage to shunt low-frequencies (such as anything below 500Hz). Ironically, I am running my solid-state amp with an 2:1 output transformer.
I have presented the problem: mismatched tubes.
Since I can adjust the bias by hand, this tells me that I can design some electronics to do it for me.
Using a cathode current sense resistor is obvious. Will it work with class AB? I was assuming it would, if the sense resistor was bypassed with a capacitor. That way, all the AC would go around the servo circuit, leaving the servo to worry about maintaining DC bias.
Making the servo circuit a low-pass filter would also help, because it would only adjust the bias current in response to very low frequency changes. Since this output stage is for a tweeter in an active-crossover system, the only AC the output stage would ever see would be above 1kHz.
In my opinion, the push-pull output stage has been underdeveloped. All the technology and literature today is based on designs from 50 or more year ago. As new production tubes become more available these days (considering I can buy a new 6L6 from at least 4 different manufacturing sites around the world), I believe we have not seen the last of the push-pull, nor have we seen the developments and maturity that the push-pull output stage will eventually show.
Other comments:
- My push-pull transformers are quite cheap. I don't think they can handle much bias differential. If I am even slightly pleased with the sound, I will purchase nice new OPTs.
- I'm not on a quest to hear differences in output stage configurations, so I'm not setting myself up for disappointment. I am on a quest to have a robust amplifier that can put up with a lot of abuse.
- I'm not scared of lots of circuitry. Many circuit components can make design easier, because less is demanded from each piece. However, it does increase the chance that the negative attributes of each piece can accumulate and make lousy sound.
- I've already lost a very expensive tweeter to a solid-state amp. That's the reason I am going for tubes. The transformer-coupled output stage protects from deadly DC offset, and I can use capacitors in the output stage to shunt low-frequencies (such as anything below 500Hz). Ironically, I am running my solid-state amp with an 2:1 output transformer.
SWMoore said:
There are already tons of good designs out there that are stable.
As for learning and trying new things, great. Look at what the experienced engineers have done in the past, dissect it, and find the limitations in the designs and ways to remedy them. Also keep in mind that many of the 'classic' designs were done with a large eye on cost, because 50 years ago, the economics played as big a role as today, maybe more in the later days when they were trying to compete with cheap SS watts.
Why is this a problem. Buy matched quads/sextets/octets from a reputable supplier. they cost no more usually, or at least only a couple of dollars.
Secondly, get an OPT that can handle some offset.
Both of these will likely give you better long term satisfaction than futzing with a complicated biassing scheme.
As an engineer, I always ask will the added complication bring enough benefit from it's inclusion to justify the extra hassle, expense and possible reliability (you menttion this one below)
issues?
With a burned in set of matched valves, I'd usually check the bias every couple of thousand hours use, and seldom does it need adjusting. Playing devils advocate, are you looking for solutions to non-existent or minor problems?
There are a lot of designs out there that do this.
Well, why not simply make it class A and be done with it? Power will not be an issue, and if imbalance really worries you, use a CCS in the tail of the output pair. Voils, no imbalance, for the cost od a simple chaep and relaible -12V or so supply. Not hard at all to get 6 clean watts out of a class A pair of PP 6L6's. If you seriously need more than that, get more efficient tweets.
Agreed. But one of the advantages of the surge of interest in SETs, is it's gotten a lot of designers to look at some classic designs, as well as show much more interest in PSU design, which can easily be applied to PP amps. Also, look back further than the post WW2 Williamson designs to the WE and RCA theatre amps. These guys really knew what they were doing, and with modern transformers, caps etc can make outstanding amps.
It's also well worth reading articles 1 - 5 and 7 here
Apart from some very rare and cult types, I have no trouble obtaining a vast variety of NOS tubes of all types, at relatively good prices. Some of the new tets and pents available new are also excellent.
As this is to be for an optimised HF amp, get some OPTs that have excellent HF performance, even at the expense of the LF, so a gap will not be a problem, accomodating quite a bit of offset.
It's a tweeter amp. How much abuse is it realistically going to have to take?
Definately the latter is a concern. make everything as simple as possible.
How much power do you really need for this application? Not much I would think. A simple class A differential with 6L6 family tubes (and there are better types out there) will give a few watts, very clean, without even having to add NFB. The STC 6L6G and 807 datasheets give working and measured exmples.
No offense meant, but what you're trying to do sounds way too Rube Goldbergian for the required task at hand.
Good luck.
Hi Stephen,
I really don't think I can add to Brett reply except to say - try your ideas and see if they work for you...
There is a lot of info. out on the net and there is no substitute for searching it out and reading it yourself. (Indeed your site has helped me in my quest for my perfect speakers )
I agree with your feelings on push pull designs and I have knocked to gether several and the designs are on this site along with some comments by the assembled wisdom.
There are two places where the use of servo bias is justified (to my mind!) 1) is high power RF amplifiers and the other is high power guitar amps. I wouldn't use it else where for the reasons Brett has given plus I don't like the influence it has on the sound - even when the cathode resistor is bypassed by a capacitor the bias still moves around - we often forget that a music signal is asymmetrical so the average dc point moves around according to programme - this happens anyway in a 'natural way' or rather a programme sympathic way with triodes - adding a bias servo removes this link and sounds un-natural (all servos hunt and the ear can hear the tiny variations in the shape of the distortion spectrum.
Thats my two cents worth...
ciao
James
I really don't think I can add to Brett reply except to say - try your ideas and see if they work for you...
There is a lot of info. out on the net and there is no substitute for searching it out and reading it yourself. (Indeed your site has helped me in my quest for my perfect speakers
)I agree with your feelings on push pull designs and I have knocked to gether several and the designs are on this site along with some comments by the assembled wisdom.
There are two places where the use of servo bias is justified (to my mind!) 1) is high power RF amplifiers and the other is high power guitar amps. I wouldn't use it else where for the reasons Brett has given plus I don't like the influence it has on the sound - even when the cathode resistor is bypassed by a capacitor the bias still moves around - we often forget that a music signal is asymmetrical so the average dc point moves around according to programme - this happens anyway in a 'natural way' or rather a programme sympathic way with triodes - adding a bias servo removes this link and sounds un-natural (all servos hunt and the ear can hear the tiny variations in the shape of the distortion spectrum.
Thats my two cents worth...
ciao
James
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.