Hello all.
This is my first post.
I'm wanting to build my own tube amp from the ground up.
I'm an electrical engineering student, and I've had a couple of courses in transistor amplification, so I'm not all that scared of equivalent circuits or heavy formulas.
That said, here's my question:
I'm trying to design a good push pull output stage using a pair of 6L6s (or KT66s) in ultra linear mode, and I'm having a bit of trouble being able to analyze this and really to optimize it. Are there any good sources where I could read up on this? I'd even be willing to switch to pentode or triode operation just for the sake of being able to design it myself. I've leafed through the radiotron 4th ed. and Theory and Applications of Electron Tubes by Reich, and I've read Valve Amplifiers by Morgan Jones cover to cover, but I'm not finding them all that helpful.
Are there any amp designers in here that could help me out? I'm looking for a good graphical method, or mathematical method so that I can confirm what I'm doing. Or should I just be building this stuff in PSPICE and computer simulating it to find results? I'm just looking to be pointed in the right direction on how to design amps.
Thanks for all your help.
This is my first post.
I'm wanting to build my own tube amp from the ground up.
I'm an electrical engineering student, and I've had a couple of courses in transistor amplification, so I'm not all that scared of equivalent circuits or heavy formulas.
That said, here's my question:
I'm trying to design a good push pull output stage using a pair of 6L6s (or KT66s) in ultra linear mode, and I'm having a bit of trouble being able to analyze this and really to optimize it. Are there any good sources where I could read up on this? I'd even be willing to switch to pentode or triode operation just for the sake of being able to design it myself. I've leafed through the radiotron 4th ed. and Theory and Applications of Electron Tubes by Reich, and I've read Valve Amplifiers by Morgan Jones cover to cover, but I'm not finding them all that helpful.
Are there any amp designers in here that could help me out? I'm looking for a good graphical method, or mathematical method so that I can confirm what I'm doing. Or should I just be building this stuff in PSPICE and computer simulating it to find results? I'm just looking to be pointed in the right direction on how to design amps.
Thanks for all your help.
broadly you start at the output and work your way back to the input.
determine the power you want, the load you expect to drive.
from that you can determine the psu requirements, the transfer ratio of the transformer, etc.
from that you can determine bias and operating points, and the drive signal required.
the drive signal at the output grids determines if you need one or two stages of drive, the class of the output also determines drive characteristics, for example if you are going into class 2 land, you will need a low Z output driver, otherwise a normal hi Z driver will be ok.
class 2 output stages need grid current. best avoided if poss.
most amps 20W plus, comprise of a first stage, splitter then a driver, then the output stages.
Re read the bevois amp section in the jones book again, it is really excellent at going through the above.
what specifically is it in his book that you fail to understand?
If I can help, i am more that pleased to do so.
kind regards
b
determine the power you want, the load you expect to drive.
from that you can determine the psu requirements, the transfer ratio of the transformer, etc.
from that you can determine bias and operating points, and the drive signal required.
the drive signal at the output grids determines if you need one or two stages of drive, the class of the output also determines drive characteristics, for example if you are going into class 2 land, you will need a low Z output driver, otherwise a normal hi Z driver will be ok.
class 2 output stages need grid current. best avoided if poss.
most amps 20W plus, comprise of a first stage, splitter then a driver, then the output stages.
Re read the bevois amp section in the jones book again, it is really excellent at going through the above.
what specifically is it in his book that you fail to understand?
If I can help, i am more that pleased to do so.
kind regards
b
When you get to seriously thinking about your choices vis-a-vis output transformers I can help with the information that will allow you to go looking with a clear idea of what you want and why.
UL is not a mystery, it sounds slightly more sterile than a straight pentode, which sounds slightly less musical and colorful than a triode, all with the same tube too. Harmonic distortion is increasing at each step but so do all of the characteristics that play a part in believable reproduction of music. Just the strengths of each topology at play, not to say that one is inherently the best.
Has most to do with feedback, where it happens, how much and why you must have it, or can dispense with it. Sadly the numbers usually bandied about as markers of "good" performance have little to do with what you will hear. The folks here can certainly help you out with whatever level of performance you choose to start at.
Bud
UL is not a mystery, it sounds slightly more sterile than a straight pentode, which sounds slightly less musical and colorful than a triode, all with the same tube too. Harmonic distortion is increasing at each step but so do all of the characteristics that play a part in believable reproduction of music. Just the strengths of each topology at play, not to say that one is inherently the best.
Has most to do with feedback, where it happens, how much and why you must have it, or can dispense with it. Sadly the numbers usually bandied about as markers of "good" performance have little to do with what you will hear. The folks here can certainly help you out with whatever level of performance you choose to start at.
Bud
Very good advice, but, hard to find and now quite ancient so usually somewhat shabby. An alternative is http://www.pmillett.com/ . Pete has a great collection of books for download, for free and the RDH is one of them, so go dig around in there, you must have a high speed connection to the internet though...
Bud
Bud
All good advice so far! A book by Morgan Jones called "Valve Amplifiers" is very popular (you can buy it from several on-line vendors including Amazon).
I understand your preference to design it yourself (maybe that's what diy really stands for) and for that purpose you may be best advised to start with triode-connected beam tetrode or pentode tubes for the output. You will need less NFB and have fewer stability problems. You can still redeploy that same tubes for UL later, if you want to, especially if you buy an OP transformer with UL taps - no need to use the UL Taps to start with.
Don't skimp on the quality of the OPT, because it really dominates the quality of the entire amp.
I understand your preference to design it yourself (maybe that's what diy really stands for) and for that purpose you may be best advised to start with triode-connected beam tetrode or pentode tubes for the output. You will need less NFB and have fewer stability problems. You can still redeploy that same tubes for UL later, if you want to, especially if you buy an OP transformer with UL taps - no need to use the UL Taps to start with.
Don't skimp on the quality of the OPT, because it really dominates the quality of the entire amp.
You should download and read the Norman Crowhurst
series of tutorials from 1955. These will outdo most of the
sources you've already found:
http://www.audioxpress.com/resource/audioclass/index.htm
The design discussions cover KT66 in several cases... The 807
is another tube to consider in your plans as they are inexpensive,
quite available and the predecessor to the KT66 and 6L6.
Simulation is rather moot since much has to do with transformer parameters that tend to be empirically determined or provided
to you buy the transformer builder who has taken the pains of
characterize their performance.
Simulation might illuminate issues regarding
stability when it comes to designing the feedback network though....
-- Jim
series of tutorials from 1955. These will outdo most of the
sources you've already found:
http://www.audioxpress.com/resource/audioclass/index.htm
The design discussions cover KT66 in several cases... The 807
is another tube to consider in your plans as they are inexpensive,
quite available and the predecessor to the KT66 and 6L6.
Simulation is rather moot since much has to do with transformer parameters that tend to be empirically determined or provided
to you buy the transformer builder who has taken the pains of
characterize their performance.
Simulation might illuminate issues regarding
stability when it comes to designing the feedback network though....
-- Jim
Hey-Hey!!!,
The most complicating thing about U-L are working with the curves. All I have seen from tube the manufacturers assume a typical maximum power operating point. In this scenario, the load applied by the output TX falls from half the plate-to-plate load to a quarter as the opposite tube cuts off( AB1 operation ). If you can stay Class A( with its lower power, lower voltage and higher idle current demands ), the load is a lot easier to deal with. Half the a-a load drawn single-ended-style across a single set of characteristics.
WIth U-L and the high power operating point the curves are drawn for, they're pretty much useless at any other starting point. You'd need Class A U-L curves, and nobody bothered to publish those. The startig point( as in plate and g2 voltage ) determine the shape of the curves. You can make some guesses, that are useful...and helpful and even fairly close to reality.
I'll have to disagree with Bud's generalizations on th sound of these amps. I build U-L amps and I have not found that result. They're an unusual design, with a new circuit and FB path. Even the high power ones are 2 stage, that's all that's needed. They're also all Class A.
On your first amp, I will suggest designing in some flexibility. Ability to draw much more current from the power supply, both te heater/filament and the B+ will be useful. Ability to adjust the B+ up or down will be useful in future modifications and/or refining processes. Also, leave some room, much more than you think will be needed, you don't need to be cramped and crammed..it is not at all useful( save for cramming a big valve amp into a small room ). My usual has been to start with a chassis twice as big as it needs to be, and that works out about 10% too small.
cheers,
Douglas
The most complicating thing about U-L are working with the curves. All I have seen from tube the manufacturers assume a typical maximum power operating point. In this scenario, the load applied by the output TX falls from half the plate-to-plate load to a quarter as the opposite tube cuts off( AB1 operation ). If you can stay Class A( with its lower power, lower voltage and higher idle current demands ), the load is a lot easier to deal with. Half the a-a load drawn single-ended-style across a single set of characteristics.
WIth U-L and the high power operating point the curves are drawn for, they're pretty much useless at any other starting point. You'd need Class A U-L curves, and nobody bothered to publish those. The startig point( as in plate and g2 voltage ) determine the shape of the curves. You can make some guesses, that are useful...and helpful and even fairly close to reality.
I'll have to disagree with Bud's generalizations on th sound of these amps. I build U-L amps and I have not found that result. They're an unusual design, with a new circuit and FB path. Even the high power ones are 2 stage, that's all that's needed. They're also all Class A.
On your first amp, I will suggest designing in some flexibility. Ability to draw much more current from the power supply, both te heater/filament and the B+ will be useful. Ability to adjust the B+ up or down will be useful in future modifications and/or refining processes. Also, leave some room, much more than you think will be needed, you don't need to be cramped and crammed..it is not at all useful( save for cramming a big valve amp into a small room ). My usual has been to start with a chassis twice as big as it needs to be, and that works out about 10% too small.
cheers,
Douglas
If Reich and Jones aren't helpful to you, then (IMO) you're taking a totally wrong design approach. Your university education has taught you a reliance on simulators, which may be fine for many solid state circuits with parameters controlled by basic quantum mechanics, but fall down badly when asked about mechanical objects with real bits of glass and metal, which is what tubes really are. You'll need to adjust to the idea of working more empirically. That means tube data sheets, real (not virtual) oscilloscopes, load resistors, signal generators, and soldering irons.
I'm unaware of any tube amp designed by simulation from the ground up- the real role of sims in tube design is to knock out really stupid ideas and to get some ballpark idea of operating points. I admit to being an old fart, so I handle the latter much more quickly and easily with characteristic curves from the data sheets and a ruler.
I'm unaware of any tube amp designed by simulation from the ground up- the real role of sims in tube design is to knock out really stupid ideas and to get some ballpark idea of operating points. I admit to being an old fart, so I handle the latter much more quickly and easily with characteristic curves from the data sheets and a ruler.
d1983, you asked about a graphical method. If you read Jones and Reich and have RDH4, then I suspect you understand how to select an operating point and draw a loadline for a single tube (i.e. single ended.) To do the same thing with PP you need to make up a new set of plate curves for the pair of tubes at the operating point you have chosen, then draw a PP loadline - called a composite loadline. The method of creating the PP plate curves is described in RDH4.
It's a huge PITA. To be completely honest, in 30 years of studying and designing and building tube amps, I have never made a set of composite curves.
There is an easy cheat. If you limit yourself to class A then you can use the plate curves for a single tube. The load seen by each tube will be 1/2 that of the plate-to-plate impedance presented by the output transformer. You don't get the benefit of seeing graphically what the real PP curves look like, but you can get close enough to build a good amp.
UL is tricky to design, mainly because the curves just aren't available for most tubes. There are pentode curves available, but you still need to work with the screen voltages that are offered in the data sheets. If you want to try something else you're out of luck. Triode curves can be used for any operating point, so they might be a little easier.
Incidentally, it seems to me that it ought to be possible to use a good spice model generate a full set of UL or pentode curves for any set of conditions desired. Any Vg2, any screen winding, even full blown composite PP curves. I'm not talking about using Spice to model a circuit, just using the tube model to generate specific plate curves. There was a thread about some software that allowed you to use an existing set of curves to generate a model. I don't recall the name of the program or where to get it. At the time it ran most reliably on a Linux system, but I understand that people do run it under Windoze.
One other thing, if you use the 6L6, you should get a copy of the Kenrad engineering bulletin (available at http://www.tubezone.net/tubedata.html down near the bottom of the page.) There is a tremendous amount of information in there. It covers the original metal 6L6 (published in 1936!) The 6L6GC that you would probably use has extended ratings (higher max plate voltage and dissipation) so keep that in mind.
-- Dave
It's a huge PITA. To be completely honest, in 30 years of studying and designing and building tube amps, I have never made a set of composite curves.
There is an easy cheat. If you limit yourself to class A then you can use the plate curves for a single tube. The load seen by each tube will be 1/2 that of the plate-to-plate impedance presented by the output transformer. You don't get the benefit of seeing graphically what the real PP curves look like, but you can get close enough to build a good amp.
UL is tricky to design, mainly because the curves just aren't available for most tubes. There are pentode curves available, but you still need to work with the screen voltages that are offered in the data sheets. If you want to try something else you're out of luck. Triode curves can be used for any operating point, so they might be a little easier.
Incidentally, it seems to me that it ought to be possible to use a good spice model generate a full set of UL or pentode curves for any set of conditions desired. Any Vg2, any screen winding, even full blown composite PP curves. I'm not talking about using Spice to model a circuit, just using the tube model to generate specific plate curves. There was a thread about some software that allowed you to use an existing set of curves to generate a model. I don't recall the name of the program or where to get it. At the time it ran most reliably on a Linux system, but I understand that people do run it under Windoze.
One other thing, if you use the 6L6, you should get a copy of the Kenrad engineering bulletin (available at http://www.tubezone.net/tubedata.html down near the bottom of the page.) There is a tremendous amount of information in there. It covers the original metal 6L6 (published in 1936!) The 6L6GC that you would probably use has extended ratings (higher max plate voltage and dissipation) so keep that in mind.
-- Dave
First off:
Thanks everyone so much. I had no idea that I would get this sort of response, it's much appreciated!
To Sy: I hope that my education won't hinder my build too much, I've paid a lot of money for it so far. hahaha. But you were right on par with what I need to figure out with regards to data sheets, and I really want to get learning.
New, and More Specific Question:
I've found a method to start looking at Class AB1 from data sheets, but my problem is this:
What exactly is the 'plate current' on the data sheet? Is this the Q point current? Because I've tried to figure it out for myself and I've found that I"m pretty consistently halving this value. I.E the data sheet says 88mA plate current at zero signal, but I calculate it at about 44mA for class AB1 pentode. Is it the Q point for both tubes added together, or is it the point where the tube cuts to B operation?
I've also found that I'm calculating the max plate current wrong as well. Am I wrong in assuming that in class B the Ra (or Rp), is 1/4Ra-a?
Also, I figured out how to use PSPICE today to create approximate plate curves for just about any situation. (UL, Pentode at different screen voltages, triode) and I have some common tube models kicking around so if anyone wants a curve set maybe I could help them out.
Thanks for all of the help again.
Dave
Thanks everyone so much. I had no idea that I would get this sort of response, it's much appreciated!
To Sy: I hope that my education won't hinder my build too much, I've paid a lot of money for it so far. hahaha. But you were right on par with what I need to figure out with regards to data sheets, and I really want to get learning.
New, and More Specific Question:
I've found a method to start looking at Class AB1 from data sheets, but my problem is this:
What exactly is the 'plate current' on the data sheet? Is this the Q point current? Because I've tried to figure it out for myself and I've found that I"m pretty consistently halving this value. I.E the data sheet says 88mA plate current at zero signal, but I calculate it at about 44mA for class AB1 pentode. Is it the Q point for both tubes added together, or is it the point where the tube cuts to B operation?
I've also found that I'm calculating the max plate current wrong as well. Am I wrong in assuming that in class B the Ra (or Rp), is 1/4Ra-a?
Also, I figured out how to use PSPICE today to create approximate plate curves for just about any situation. (UL, Pentode at different screen voltages, triode) and I have some common tube models kicking around so if anyone wants a curve set maybe I could help them out.
Thanks for all of the help again.
Dave
Also, here's a page that helped me quite a bit if anyone should be reading this with the same problems as me.
Data sheet analysis for Class AB1 Pentodes
Data sheet analysis for Class AB1 Pentodes
I'm delighted that you've dug into cellulose and graphite simulation.
Generally, the data sheet will say something like "values are for two tubes" or something like that. There will be single ended stuff, then the push-pull characteristics. I'd suggest you look at the Mullard datasheets for EL34, EL84, and the like, just to see how a datasheet should be done.
Frank Philipse's site is a treasure trove, use it well! And with your new-found discoveries, go back to Jones's book- you'll get more out of it now.
Generally, the data sheet will say something like "values are for two tubes" or something like that. There will be single ended stuff, then the push-pull characteristics. I'd suggest you look at the Mullard datasheets for EL34, EL84, and the like, just to see how a datasheet should be done.
Frank Philipse's site is a treasure trove, use it well! And with your new-found discoveries, go back to Jones's book- you'll get more out of it now.
Dave,
If you haven't encountered LT Spice look here, it is free and there are a lot of people using it. http://www.linear.com/designtools/spice_models.jsp
One of the hot beds of LT sims is at Intact Audio and Dave Slagle is a good person to get to know. as are the rest of us. Check the Forum.
http://www.intactaudio.com/
Bud
If you haven't encountered LT Spice look here, it is free and there are a lot of people using it. http://www.linear.com/designtools/spice_models.jsp
One of the hot beds of LT sims is at Intact Audio and Dave Slagle is a good person to get to know. as are the rest of us. Check the Forum.
http://www.intactaudio.com/
Bud
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.