I picked up an Eico ST70 because I wanted to use the iron to do a ground up power amp build. Anyway, being a noobie, I decided to use the original ciruit design instead of my own and see what happens.
Kept most everything to spec with the exception some of the power supply cap values and the use of two 12ax7's up front, instead of using 1/2 for each channel. Also, ignored the original grounding scheme which I found a tad strange.
First test without global feedback revealed a rather nasty hum that was eliminated by elevating the ac heaters (+17v). Then giving a listen, I was really surprised by the amount of gain. With the 7591's balanced and putting out a mild 35ma, I was pleased because the amp was actually working but thought something wasn't quite right about the quality of the sound.
So today I put a 1k sqare wave into it and had to laugh when I saw the snakelike output on the scope. Grounding the opt commons and trying some different value resistors in the feedback circuit left me with a pretty good wave that has just a small initial spike and slight hump. Another listening test and the change was dramatic, much less gain and sound quality greatly improved.
My question is, should the amount of feedback applied be determined by careful listening (with trained ears), by square wave analysis - or some other method?
Thanks again, Ed - and please, all comments / suggestions very welcome.
An externally hosted image should be here but it was not working when we last tested it.
Kept most everything to spec with the exception some of the power supply cap values and the use of two 12ax7's up front, instead of using 1/2 for each channel. Also, ignored the original grounding scheme which I found a tad strange.
First test without global feedback revealed a rather nasty hum that was eliminated by elevating the ac heaters (+17v). Then giving a listen, I was really surprised by the amount of gain. With the 7591's balanced and putting out a mild 35ma, I was pleased because the amp was actually working but thought something wasn't quite right about the quality of the sound.
So today I put a 1k sqare wave into it and had to laugh when I saw the snakelike output on the scope. Grounding the opt commons and trying some different value resistors in the feedback circuit left me with a pretty good wave that has just a small initial spike and slight hump. Another listening test and the change was dramatic, much less gain and sound quality greatly improved.
My question is, should the amount of feedback applied be determined by careful listening (with trained ears), by square wave analysis - or some other method?
Thanks again, Ed - and please, all comments / suggestions very welcome.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
Square wave analysis is the surest and swiftest. Get it stable into a resistive load, then try a reactive load. Also, it's a good idea to try no-load but with very low input levels- that's a pretty severe test.
When you want to really have some fun, you can play around with dominant pole compensation to get rid of the feedback cap. Also, like most transformers of that era from Chicago Transformer, there's some primary imbalance. You can experiment with connecting small caps (100-100pF, high voltage rating!) from the blue to red lead to see if you can get out any asymmetry in the overshoot.
And, finally, when you want to play with the topology a bit, I've done some mods which I think are worthwhile on that same input/phase-splitter circuit.
When you want to really have some fun, you can play around with dominant pole compensation to get rid of the feedback cap. Also, like most transformers of that era from Chicago Transformer, there's some primary imbalance. You can experiment with connecting small caps (100-100pF, high voltage rating!) from the blue to red lead to see if you can get out any asymmetry in the overshoot.
And, finally, when you want to play with the topology a bit, I've done some mods which I think are worthwhile on that same input/phase-splitter circuit.
Vespasian said:
Yes
.Feedback design can be problematic especially for noobies. All possible techniques can and should be used to evaluate the results.
Check carefully for frequency specific bumps or dips using a sinewave sweep. Using an oscilloscope watch carefully for oscillation bursts during certain parts of the test signal swing (usually a sine wave is best for this). And the big kicker, check for how the amp recovers from overdrive or clipping. Many feedback circuits make clipping much worse. The feedback circuit in effect tries to improve the waveshape during clipping, it can't, so earlier stages can be driven into overload when otherwise they wouldn't be. Watch out for this.
Sy, the feedback cap, you are refering to the C25 in the schematic? I didn't have anything new at hand that was close, so I've just got the 6k in there now. I think the addition of the cap will take out the small initial overshoot spike? Also have a little issue with the plate voltage at pin 2 on the 6sn7's - it is a lot higher than the specs (330), partly because I don't quite have the 28.75k called for (doing a series thing with 1w resistors). Using the original eico labelled tubes too...
Very interested to hear what you've done with similar topologies, please email or post when you have a chance.
Forgot to mention that I'm tapped into the 16ohm for feedback and my input is simply a .1, a 100k pot and a 470k to ground for the 12ax7 grids - plates are at 94v and .68 at the cathodes. Didn't have 25mfd's for the 1k bypass but plugged in some 16's for the time being. The different bypass cap value will also have an effect on feedback results right?
Going to do a lot more testing, will try the static and no-load suggestion - also, have yet to look at sweeps and see what happens there - (thanks for all the pointers Herman).
Is it even possible to get a good square wave output over the enitire frequency range? At what point do we say good-nuf?
Thanks again, Ed
Very interested to hear what you've done with similar topologies, please email or post when you have a chance.
Forgot to mention that I'm tapped into the 16ohm for feedback and my input is simply a .1, a 100k pot and a 470k to ground for the 12ax7 grids - plates are at 94v and .68 at the cathodes. Didn't have 25mfd's for the 1k bypass but plugged in some 16's for the time being. The different bypass cap value will also have an effect on feedback results right?
Going to do a lot more testing, will try the static and no-load suggestion - also, have yet to look at sweeps and see what happens there - (thanks for all the pointers Herman).
Is it even possible to get a good square wave output over the enitire frequency range? At what point do we say good-nuf?
Thanks again, Ed
Safety first...
It's somewhat off-topic, but could we please have a better arrangement for securing the mains cable? Tying a knot in it was marginally acceptable in the 50s, but not now when there are all manner of clamping cable glands available.
And yes, square waves are the fastest way of tweaking the negative feedback.
It's somewhat off-topic, but could we please have a better arrangement for securing the mains cable? Tying a knot in it was marginally acceptable in the 50s, but not now when there are all manner of clamping cable glands available.
And yes, square waves are the fastest way of tweaking the negative feedback.
My first stereo was a Sansui receiver. It most likely used a highly integrated amplifier design, maybe even an integrated output stage.
Though I never did it, I'm quite sure that any square wave passed through that amplifier would have come out quite pretty with no overshoot, ringing, tilt or excessive roll off causing slow rise time on the waveform sides.
The problem is, compared to my current (significantly more expensive) equipment it sounded quite awful. I'm sure it measured just fine, perhaps even exceptional. This amplifier was designed in those early transistor design days when very large amounts of negative feedback were quite popular and I believe all Japanese mass market electronics sounded equally awful.
The point I'm making is that while triming a design so that square waves look good isn't bad it also isn't even close to what you need to do to make a truly first class sound amplification system. It is a starting point, one tool, nothing more.
Even noobies are permitted to build amplifiers with exquiste sound, there is no license, guild or secret society. On the other hand neither is it trivial. To the best of my knowledge the difference between a good sounding design and a lesser one is is not fully understood, even experienced designers find a need to listen and modify the design based on listening results.
Though I never did it, I'm quite sure that any square wave passed through that amplifier would have come out quite pretty with no overshoot, ringing, tilt or excessive roll off causing slow rise time on the waveform sides.
The problem is, compared to my current (significantly more expensive) equipment it sounded quite awful. I'm sure it measured just fine, perhaps even exceptional. This amplifier was designed in those early transistor design days when very large amounts of negative feedback were quite popular and I believe all Japanese mass market electronics sounded equally awful.
The point I'm making is that while triming a design so that square waves look good isn't bad it also isn't even close to what you need to do to make a truly first class sound amplification system. It is a starting point, one tool, nothing more.
Even noobies are permitted to build amplifiers with exquiste sound, there is no license, guild or secret society. On the other hand neither is it trivial. To the best of my knowledge the difference between a good sounding design and a lesser one is is not fully understood, even experienced designers find a need to listen and modify the design based on listening results.
It's pretty well understood. John Atwood has a PowerPoint presentation on some fine correlative work he's done and the question of harmonic distribution has been known since the '60s. But there are still things which are MUCH easier to pick up by listening than by measurement, one important example being overload recovery. With real music in real rooms, amps clip. If they clip some peaks and move on without breaking a sweat, they'll sound much better than if they choke up for a few hundred milliseconds. This is all measurable, but correlation to ear-offense is difficult.
I'm still waiting for the amp that "measures good, sounds bad" unless one puts a Procrustean restriction on the measurements.
I'm still waiting for the amp that "measures good, sounds bad" unless one puts a Procrustean restriction on the measurements.
Why do some circuits just sound better than others?
Clearly improving the sum of all harmonics total distortion can't help but make a better amplifier. Still, I offer that op-amps on the whole do very well on this test, but do not automatically sound good or at least not as good as a discrete design executed by a competent designer.
If you ask the John Curl, Nelson Pass types if there is a single secret to making good sounding amps, my guess is they'll say the answer is just not that simple. Grounding and power supply impedance are important as are many other factors. To the best of my knowledge, all attempts to reduce amplifier design to a simple step by step cookbook have failed to automatically produce outstanding designs.
I have seen a number of solutions of the minute to this problem. Many were convined that SID or TIM distortion was the holy grail of understanding why some amps sound better than others. Although feedback slew limiting or timing delay was a part of the answer it certainly wasn't all of it. If your understanding was true any competent designer could make a giant killer amp for $400 that would challenge the likes of Levinson, Pass, Krell, Edge etc. They can't because it's just not the whole answer. After all a fairly simple (read low cost) voltage gain stage with excellent distortion measurements can be made for $20 worth of parts. Why use Vishays when a carbon film measures very well in total distortion?SY said:
Clearly improving the sum of all harmonics total distortion can't help but make a better amplifier. Still, I offer that op-amps on the whole do very well on this test, but do not automatically sound good or at least not as good as a discrete design executed by a competent designer.
If you ask the John Curl, Nelson Pass types if there is a single secret to making good sounding amps, my guess is they'll say the answer is just not that simple. Grounding and power supply impedance are important as are many other factors. To the best of my knowledge, all attempts to reduce amplifier design to a simple step by step cookbook have failed to automatically produce outstanding designs.
Still waiting? I disagree strongly, there have been a number of amplifiers that measured reasonably good like my example of early Japanese transistor receivers with THD numbers around 0.0005% or better that nevertheless sounded quite awful. Nowhere near as good as some of todays amplifiers that measure more like 0.5%SY said:
Well, if you want to say "THD at one spot frequency with one load doesn't give a perfect correlation," I wouldn't disagree with that. But please note the last phrase...
edit: Do look up John's paper. He dissects the measured performance of one of those low THD wonders compared against an Eico (I think) tube amp.
edit: Do look up John's paper. He dissects the measured performance of one of those low THD wonders compared against an Eico (I think) tube amp.
Perhaps there is a semantics issue. The charts I've seen show THD with a swept sine wave source. They usually have a couple of typical overall shapes.
If you want to discuss interaction between multiple signals of different frequencies and levels that is a different horse, now we seem to be talking about intermodulation distortion, not the same as THD although most amps that measure good with one will measure well with the other.
Things I rarely see measured are differential gain or crossmodulation between one signal of stepped amplitude vs another frequency fixed small amplitude signal. The test isolates the small signal in the output but displays it synchronized to the step rate of the larger interfering signal, a measurement of amplifier masking and differential interaction. I used the test for video amplifier design with results that were often surprising.
I own a Pass Labs X250, it is a fine amplifier, my good friend owns a Bryston 4BSST also an excellent sounding amplifier. The design styles of these two products could hardly be more different, the Bryston is all complemetary conventional transistors and the circuit is remarkably complex. The Pass is a much simpler design built around FET transistors. The Pass is faster, the Bryston has a lusher midrange, I'm sure both amps would measure perfectly flat, but they still have a different presentation to the ear, sounding as if one is tilted up in response and the other is tilted down. Neither is intrinsically better, they are just different.
Even people with vast design experience do a lot of listening in the end. If this was a mere matter of equations or test measurement that shouldn't be very necessary.
A great number of very low power SET designs are sold on a regular basis, they all measure badly and often sound great, revealing detail that the massive transistor amps do not. I have never seen an audiophile commercial SE 5 watt transistor amp. Certainly the same single ended techniques can be, but never seem to be applied to a low power transistor design. I don't claim it's magic, but I think the reasons are less clear than you seem to believe.
Anyway, I don't have the answer, I merely object to he notion that if you follow the rules the design will automatically sound great. It will probably be acceptable, but great is harder.
I will read John's paper (not tonight, it's late), I have read similar things and I certainly learn new things even at my age approaching that of a fossil.
To borrow a term, cheers.
If you want to discuss interaction between multiple signals of different frequencies and levels that is a different horse, now we seem to be talking about intermodulation distortion, not the same as THD although most amps that measure good with one will measure well with the other.
Things I rarely see measured are differential gain or crossmodulation between one signal of stepped amplitude vs another frequency fixed small amplitude signal. The test isolates the small signal in the output but displays it synchronized to the step rate of the larger interfering signal, a measurement of amplifier masking and differential interaction. I used the test for video amplifier design with results that were often surprising.
I own a Pass Labs X250, it is a fine amplifier, my good friend owns a Bryston 4BSST also an excellent sounding amplifier. The design styles of these two products could hardly be more different, the Bryston is all complemetary conventional transistors and the circuit is remarkably complex. The Pass is a much simpler design built around FET transistors. The Pass is faster, the Bryston has a lusher midrange, I'm sure both amps would measure perfectly flat, but they still have a different presentation to the ear, sounding as if one is tilted up in response and the other is tilted down. Neither is intrinsically better, they are just different.
Even people with vast design experience do a lot of listening in the end. If this was a mere matter of equations or test measurement that shouldn't be very necessary.
A great number of very low power SET designs are sold on a regular basis, they all measure badly and often sound great, revealing detail that the massive transistor amps do not. I have never seen an audiophile commercial SE 5 watt transistor amp. Certainly the same single ended techniques can be, but never seem to be applied to a low power transistor design. I don't claim it's magic, but I think the reasons are less clear than you seem to believe.
Anyway, I don't have the answer, I merely object to he notion that if you follow the rules the design will automatically sound great. It will probably be acceptable, but great is harder.
I will read John's paper (not tonight, it's late), I have read similar things and I certainly learn new things even at my age approaching that of a fossil.
To borrow a term, cheers.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.