Nop its more of tolerancing. If you happy to adjust the LTP CCS if you change KT88's that's OK. Having discovered the distortion cancellation I would be looking for a better way to introduce it while still retaining full power. Its not common to let the driver stage saturate before the output stage. However - at least we understand better why you are getting such good results.
What exactly do you mean by ‘tolerancing’ and why do you insist on ‘making the most power’ is the proper way of going about a design? Isn’t the latter a design choice? If I had wanted maximum output power I would’ve made different choices to begin with. As mentioned I’m not after extracting every last Watt from this design, but rest assured I’ll try very hard to make sure I reduce distortion in every way I can.
Yes I am just wondering if the KT77 may need less pre-distortion. Tolerance means making the design repeatable but for a one or two off that not really an issue. Also whether you go for low distortion or full drive does not require component changes so why worry.
Last edited:
Might be worth exploring whether (or not) the cold running LTP is improving distortion at full output at the expense of distortion at normal working levels. As proposed previously, the cold LTP bias and resulting higher distortion in that stage could be compensating for output stage operation beyond class A by pre-distortion.
Also, forward biased LEDs are very low impedance, so not easily bypassed by capacitor. Not needed.
All good fortune,
Chris
Also, forward biased LEDs are very low impedance, so not easily bypassed by capacitor. Not needed.
All good fortune,
Chris
I’m pretty sure the design is reproducible and isn’t one-off, how do I know? Simply because I have been swapping out tubes and output transformers since I’ve been tinkering with the prototype for well over a year ago. I’ve used Hammond, Ogonowski and VanderVeen output transformers, and despite their differences I could get all of them to offer similar performance.
I’ve tried EF86 tubes of different brands, same for the 6922. The only constant have been the EH KT88s of which I have 4pcs that I use for prototyping. I have swapped them out and mixed them up a few times already and all that’s required in that case is setting the bias again. So from where I am sitting this is not some one-off project?
I’ve tried EF86 tubes of different brands, same for the 6922. The only constant have been the EH KT88s of which I have 4pcs that I use for prototyping. I have swapped them out and mixed them up a few times already and all that’s required in that case is setting the bias again. So from where I am sitting this is not some one-off project?
I think you think there's criticism where there is non. As I said I would always design the output stage to saturate first rather than the driver. Designing an amp is a balance of things.
All my amps have been 1-offs not intending to do a small production run of them.
All my amps have been 1-offs not intending to do a small production run of them.
Fair enough, I felt somewhat backed into a corner, thank you.
Frankly I’m not sure which saturates first, I honestly haven’t pushed it to clipping, the setting I use in my measurement software is to stop the measurement when the THD hits 5%, and I franky don’t plan to push it past that in practice. I rarely use more than 8-10W during listening to music at realistic levels.
Frankly I’m not sure which saturates first, I honestly haven’t pushed it to clipping, the setting I use in my measurement software is to stop the measurement when the THD hits 5%, and I franky don’t plan to push it past that in practice. I rarely use more than 8-10W during listening to music at realistic levels.
Ok I will explain the issue a bit more carefully. If the LTP is running out of current and saturating and that's an if.
If the KT88 grid can be driven to say -1V on the peaks before LTP saturates and you get 50W say. Suppose you change the KT88 and have to adjust the grid bias two more volts negative to get the same current. This will move the -1V to -3V before saturation. This will drop the output power to say 30W. The the HT changes by 2V on the plates of the LTP same thing. That is my concern but not proven.
If the KT88 grid can be driven to say -1V on the peaks before LTP saturates and you get 50W say. Suppose you change the KT88 and have to adjust the grid bias two more volts negative to get the same current. This will move the -1V to -3V before saturation. This will drop the output power to say 30W. The the HT changes by 2V on the plates of the LTP same thing. That is my concern but not proven.
My first two rules about clipping . . .
1. During amplifier testing:
Find out at what amplitude it clips.
Turn it up and down, above and below clipping, and see how well behaved it is, or does it go "Bonkers", or does it clip cleanly?
2. During listening sessions:
If it is clipping, turn the volume down a little.
1. During amplifier testing:
Find out at what amplitude it clips.
Turn it up and down, above and below clipping, and see how well behaved it is, or does it go "Bonkers", or does it clip cleanly?
2. During listening sessions:
If it is clipping, turn the volume down a little.
As for clipping, I don’t think the LTP runs out of steam, simply because I have raised the input resistors on the output stage from 100K to 220K to see whether that would reduce distortion further. It didn’t, nor did it raise the output power. Also, in a previous iteration of the prototype I used a negative supply for the LTP, adding an additional 12V of headroom, again the output power remainded the same. Take note though that I stop the measurement at 5% THD, so it could well be that the output power at 10% THD was higher in these two cases, I’ve just never explored that.
Everybody . . .
Who wants to listen when the amplifier distortion is 10%?
What is the distortion of your speakers at 50 Watts?
What is the distortion of your speakers at 70 Watts?
Power is one thing.
Hi Fi is another.
Often those two do not meet.
Who wants to listen when the amplifier distortion is 10%?
What is the distortion of your speakers at 50 Watts?
What is the distortion of your speakers at 70 Watts?
Power is one thing.
Hi Fi is another.
Often those two do not meet.
If I could offer a more general comment on design and design goals: Many commenters, including me, come from generations before Spice modelling and so have instinctual aversion to the blind spots that Spice dependence (move to Arrakis) allows to creep into design. For example, anyone who's built a few valve amplifiers with long loop feedback knows that they require a dominant LP pole (and usually a stop) and well designed HP pole to work reliably and optimally. Also grid stops, and often output Zobels, but these are not necessarily included in the simplified models like Spice, because Spice ignores parasitic reactances and judges stability on a pass/fail basis, and with make-believe loading.
Secondarily, but I believe more insidious in contemporary thought, is the very often incorrect assumption of monotonicity of distortion with level. We have become conditioned to assume that all errors decrease without exception as signal level decreases. This is really only true in special circumstances and even then only over certain ranges of signal level. Extreme examples of failure of monotonicity would be semi-con or even extremely cold-biased valve amplifiers (McIntosh, etc.) with obvious crossover issues at normal low signal levels. Measures fine, if you're measuring at full power output, but performs poorly at normal signal levels.
These are not really criticisms of Spice per se, but instead are warnings about the limitations of modelling in a single (and inherently limited) dimension. Spice is a 2D model in a 3D world, and that's fine, but its limitations need to be kept in mind. We must always examine our goals even more critically than our results.
And finally, a plea to other geezers: to encourage upcoming generations to walk the walk, to draw actual load lines, maybe even on actual paper (shocking!), to get a good working feel for what's good design and what's marginal or strained or sloppy or just not optimum. In older years, that's somehow our purpose. (Until we have to go into the failed reactor core, but that's next week).
All good fortune,
Chris
Secondarily, but I believe more insidious in contemporary thought, is the very often incorrect assumption of monotonicity of distortion with level. We have become conditioned to assume that all errors decrease without exception as signal level decreases. This is really only true in special circumstances and even then only over certain ranges of signal level. Extreme examples of failure of monotonicity would be semi-con or even extremely cold-biased valve amplifiers (McIntosh, etc.) with obvious crossover issues at normal low signal levels. Measures fine, if you're measuring at full power output, but performs poorly at normal signal levels.
These are not really criticisms of Spice per se, but instead are warnings about the limitations of modelling in a single (and inherently limited) dimension. Spice is a 2D model in a 3D world, and that's fine, but its limitations need to be kept in mind. We must always examine our goals even more critically than our results.
And finally, a plea to other geezers: to encourage upcoming generations to walk the walk, to draw actual load lines, maybe even on actual paper (shocking!), to get a good working feel for what's good design and what's marginal or strained or sloppy or just not optimum. In older years, that's somehow our purpose. (Until we have to go into the failed reactor core, but that's next week).
All good fortune,
Chris
@Chris Hornbeck,
My generation of electronic engineers have been educated to use EDA tools as much a possible, so yes, my eyes have turned blue from the prolific intake of Spice for decades and I won't deny it is usually the first tool in my toolbox. That being said tubes have never been covered during classes, so I'm just learning as I go, and by reading some of the books available. However feel free to mentor engineers like me, I welcome any insights you can offer, especially if they're relevant to the project. I've always found that the best way to learn and gain understanding is from hands on experience, which is why I always build prototypes to correlate the Spice results with those of the real world. Going forward I would appreciate your take on the mechanism behind the low distortion I'm seeing in my measurements, as I'm trying to understand, and subsequently further exploit this. Thank you.
My generation of electronic engineers have been educated to use EDA tools as much a possible, so yes, my eyes have turned blue from the prolific intake of Spice for decades and I won't deny it is usually the first tool in my toolbox. That being said tubes have never been covered during classes, so I'm just learning as I go, and by reading some of the books available. However feel free to mentor engineers like me, I welcome any insights you can offer, especially if they're relevant to the project. I've always found that the best way to learn and gain understanding is from hands on experience, which is why I always build prototypes to correlate the Spice results with those of the real world. Going forward I would appreciate your take on the mechanism behind the low distortion I'm seeing in my measurements, as I'm trying to understand, and subsequently further exploit this. Thank you.