BudP said:
If they don't look good off-band to test gear isn't a problem for me. They have just to sound "right".
The off-band peaking, in my experience, is basically only a technical problem if you want to employ some feedback.
Otherwise it has importance close to zero.....
Why?
Think about what comes out of 99.9% tweeters just around 15-20KHz (and very very often even around 10Khz) and the transformer peak will be nothing in comparison.
Here there are the waterfalls (CSD) of many common tw's....many of them are quite expensive, too!
http://www.zaphaudio.com/tweetermishmash/compare.html
45
I hate to rattle anyones cage, but I find that peaking all the way out to 70-80khz has an absolutely terrible effect on the overall sound quality. Very harsh and fatiguing. You notice the difference straight away as soon as the peaking has been tamed. I believe its down to intermodulation distortion bleeding down into the audio band.
I have heard people claim that interstage transformers with no resistive termination sound better than with, the claim is that the sound is more lively and airy. I agree that it initially sounds great but this is very fatiguing distortion and cannot be tollerated.
Peaking is totally unacceptable whenever it rears its ugly head.
shoog
I have heard people claim that interstage transformers with no resistive termination sound better than with, the claim is that the sound is more lively and airy. I agree that it initially sounds great but this is very fatiguing distortion and cannot be tollerated.
Peaking is totally unacceptable whenever it rears its ugly head.
shoog
Shoog said:
Technically speaking I completely agree with Shoog, and this is what I might do myself, however I think to some extent the acceptability of the audible consequences of the peaking depend on the speakers used and the individual's hearing. I suspect Shoog has both good hearing and speakers that allow him to hear the nasties. Full range systems in particular have very restricted response in the top octave and this might tend to make these effects less noticeable, I suspect an LPF at the front end of the amplifier limiting input bandwidth to something well below the resonance peak might also help. I have had situations where fixing that high q resonance was objectionable to the client, but his speaker system was much mellower sounding (rolled off
) than mine. In his system the underdamped behavior seemed to work better than my initial compensation - ultimately we compromised about half way..There are other ways to control the Q of the resonance such as controling the rp of the tube driving the IT with some local cathode feedback, effectiveness depends a lot on how good the transformer actually is.. Higher rp will lower the Q of the peak, (reduces the effect of leakage inductance, increases the effect of stray capacitances) and if there is insufficient primary inductance will also result in an early LF roll-off.. Zobels on the primary, secondary or simple resistive loading on the secondary also help.. Finding a well behaved IT transformer is better still, but expensive.
(My current system can sound very harsh when these sorts of issues occur, so I do understand what Shoog is talking about.. Vintage JBL pro horns on top of diy Onken bass cabs.)
My speakers are full range vintage isophons (10x8) but I also have the vintage isophon paper cone tweeters which I roll in at about 10K, this represents the mearest tickle of treble boost. I have very sensitive hearing as regards high frequency anomolies. I also use toroidal output transformers which will allow any anomolies to pass through unimpeded which cannot help if there is a problem.
Shoog
Shoog
Shoog said:
Shoog if your tweeter has a roll-off at 10KHz I can immagine what is going on. Especially in the time domain.
It is becoming directional with lots of time extended resonances up to 50KHz. I guess you already have some stuff to cope with well within the audio band.
Possibly the amplitude at those high frequencies is not dropping so quickly as one could imagine. Most likely it is like a roller coaster, especially off axis (5-10-30 degrees).
The time extended persistences of those resonances react (badly, IMO) to anything you feed them with.
Cheers,
45
You misunderstood what i said - I roll my tweeter in at about 10k with a single cap filter - so response will be down to 1khz or so.
It is difficult to know what is really going on. The fact is if you keep nasty resonances from arriving at the tweeters they don't tend to generate them themselves, all this with the amp feeding signal well past 20khz. What does that tell you.
Shoog
It is difficult to know what is really going on. The fact is if you keep nasty resonances from arriving at the tweeters they don't tend to generate them themselves, all this with the amp feeding signal well past 20khz. What does that tell you.
Shoog
If HF peaking is due to any kind of feedback mechanism, that is extremely malign, since it indicates the onset of HF instability, long settling times, a substantial decrease in phase margin, and increased sensitivity to reactive loads at ultrasonic frequencies. These are all very undesirable, and have severe effects on HF quality. Any feedback amplifier must be free of overshoots in the time domain, and with a wide variety of loads.
But in a non-feedback amplifier, it is a different matter. None of the aforementioned effects happen, since there are no stability issues that can arise (any load is tolerable for an unlimited amount of time). Although overshoots in a non-feedback amplifier are not pretty, they have no effect on stability, since the amplifier is unconditionally stable, and for that matter rejects RFI induction that comes in from the speaker cable.
The RFI ends up on the power tubes plates and stops there - unlike a feedback amp, where it makes it back to the input section, and causes lots of trouble. So overshoots in a feedback amplifier (this includes local cathode feedback with a tertiary winding on the OPT) are very different than a non-feedback amplifier, which has much fewer stability and RFI-induction issues. Part of the sonic difference between the two classes of amplifiers is probably due to the stability alone - but of course with no feedback, there is no covering-up of the inherent sonics of the tubes or passive components. It's no accident many of the best tubes were developed during the era of no feedback.
But in a non-feedback amplifier, it is a different matter. None of the aforementioned effects happen, since there are no stability issues that can arise (any load is tolerable for an unlimited amount of time). Although overshoots in a non-feedback amplifier are not pretty, they have no effect on stability, since the amplifier is unconditionally stable, and for that matter rejects RFI induction that comes in from the speaker cable.
The RFI ends up on the power tubes plates and stops there - unlike a feedback amp, where it makes it back to the input section, and causes lots of trouble. So overshoots in a feedback amplifier (this includes local cathode feedback with a tertiary winding on the OPT) are very different than a non-feedback amplifier, which has much fewer stability and RFI-induction issues. Part of the sonic difference between the two classes of amplifiers is probably due to the stability alone - but of course with no feedback, there is no covering-up of the inherent sonics of the tubes or passive components. It's no accident many of the best tubes were developed during the era of no feedback.
That is an interesting observation as my current amplifier is a Tabor clone which has the local feedback mechanism.
Still my little 6080 had very audible artifacts from the interstage peaking out at about 50K. damped with a 10K resistor and everything smoothed down nicely. This was a none feedback amp. Maybe peaking just sounds nasty in itself.
Shoog
Still my little 6080 had very audible artifacts from the interstage peaking out at about 50K. damped with a 10K resistor and everything smoothed down nicely. This was a none feedback amp. Maybe peaking just sounds nasty in itself.
Shoog
Some quasi-random thoughts...
In actuality, I was considering using a cross between Lynn's Karna front end and Kevin Carter's PP300B front end (using 45's, not 6bx7 or 2a3).
Kevin doesn't have his PP300B schematic up anymore, but here is a similar front end...
http://www.kandkaudio.com/images/PPAmpInput.pdf
Izzy Wizzy went through a variety of topologies on his PP813 that I used for input too.
http://www.izzy-wizzy.com/audio/powamp813.html
I drew up a shunt regulated version of Kevin's front end with 6SN7/7N7 into 45 complete with operating points on graph paper. If I can figure out a schematic program, I'll draft it up for scrutiny.
When I asked Kevin why he chose to use a choke anode load for the 6SN7 and DC couple to the 45s (in his case 2a3/6bx7) instead of using another interstage he said because it takes one more interstage out of the equation. This may be an asset here, which is why I bring it up.
Lynn, you are saying the ails come from the 5687/7119 driving the first IX. Would Kevin's topology sidestep this issue, at least partially? I do like that this also allows for the venerable 6SN7 in lieu of the 5687 without giving up bandwidth. Slightly higher gain, and according the Morgan Jones' VA, lower distortion and even lower high harmonics then the 5687, at least at the OPs he tested.
In my particular case, I am planning to use 7N7s because I got a bunch of them (NOS Sylvania sleeves).
I wanted to use a, at least partially, modular top plate, so I can rework the front end and/or other parts of the design without having to completely botch the chassis. I admit that 26's in place of the *sn7/*n7's has its appeal (only in the anode loaded dc coupled version).
In actuality, I was considering using a cross between Lynn's Karna front end and Kevin Carter's PP300B front end (using 45's, not 6bx7 or 2a3).
Kevin doesn't have his PP300B schematic up anymore, but here is a similar front end...
http://www.kandkaudio.com/images/PPAmpInput.pdf
Izzy Wizzy went through a variety of topologies on his PP813 that I used for input too.
http://www.izzy-wizzy.com/audio/powamp813.html
I drew up a shunt regulated version of Kevin's front end with 6SN7/7N7 into 45 complete with operating points on graph paper. If I can figure out a schematic program, I'll draft it up for scrutiny.
When I asked Kevin why he chose to use a choke anode load for the 6SN7 and DC couple to the 45s (in his case 2a3/6bx7) instead of using another interstage he said because it takes one more interstage out of the equation. This may be an asset here, which is why I bring it up.
Lynn, you are saying the ails come from the 5687/7119 driving the first IX. Would Kevin's topology sidestep this issue, at least partially? I do like that this also allows for the venerable 6SN7 in lieu of the 5687 without giving up bandwidth. Slightly higher gain, and according the Morgan Jones' VA, lower distortion and even lower high harmonics then the 5687, at least at the OPs he tested.
In my particular case, I am planning to use 7N7s because I got a bunch of them (NOS Sylvania sleeves).
I wanted to use a, at least partially, modular top plate, so I can rework the front end and/or other parts of the design without having to completely botch the chassis. I admit that 26's in place of the *sn7/*n7's has its appeal (only in the anode loaded dc coupled version).
Notwithstanding all the excellent comments by all, and hi Lynn and Kevin, the original post compared a transformer coupled 300B PP amp with a ST-70.
The ST-70 is going to sound much much more "upfront" and "bright" than the triode/ZFB (is it ZFB?) design no matter what. The stock ST-70 employs "predistortion" in the 7199 and feedback to get the freq response that they advertise, plus it has driven screens (ultralinear). I have never ever heard an amp with tubes that have screens that sound the same as amps with triodes, especially in the output stage.
It is entirely possible that what is being heard is mostly this differential? Triodes, especially 300Bs tend to not only have their own characteristic sound (differing between mfrs too) but are so very much smoother on the highs (and the ears) that you might think that the highs are lacking!
Just an idea...
_-_-bear
The ST-70 is going to sound much much more "upfront" and "bright" than the triode/ZFB (is it ZFB?) design no matter what. The stock ST-70 employs "predistortion" in the 7199 and feedback to get the freq response that they advertise, plus it has driven screens (ultralinear). I have never ever heard an amp with tubes that have screens that sound the same as amps with triodes, especially in the output stage.
It is entirely possible that what is being heard is mostly this differential? Triodes, especially 300Bs tend to not only have their own characteristic sound (differing between mfrs too) but are so very much smoother on the highs (and the ears) that you might think that the highs are lacking!
Just an idea...
_-_-bear
Bear,
But a very good idea and a well made point. The DHT sound is deep and effortless and the high frequencies do not "stand out" as they do with "audiophile" amplification. In fact, neither do the lows. Everything is matched in phase and is coherent to the musical information driving the amp, rather than modifying and deconstructing that signal, to match a preconceived notion of what constitutes "good sound".
Having said that, I have a 40 watt AudioPrism amp, PP EL34, with UL taps and regulation on all voltages, that is really quite musical to listen to. Not at all like an ST 70 and others I have heard, but not like a DHT amp either. No less musical though! And this through Planet 10 Fostex 127 / Fonken speakers, with his pre coat work and EnABL too. Really an endlessly deep and clear musical presentation, but not at all "audiophile".
Bud
But a very good idea and a well made point. The DHT sound is deep and effortless and the high frequencies do not "stand out" as they do with "audiophile" amplification. In fact, neither do the lows. Everything is matched in phase and is coherent to the musical information driving the amp, rather than modifying and deconstructing that signal, to match a preconceived notion of what constitutes "good sound".
Having said that, I have a 40 watt AudioPrism amp, PP EL34, with UL taps and regulation on all voltages, that is really quite musical to listen to. Not at all like an ST 70 and others I have heard, but not like a DHT amp either. No less musical though! And this through Planet 10 Fostex 127 / Fonken speakers, with his pre coat work and EnABL too. Really an endlessly deep and clear musical presentation, but not at all "audiophile".
Bud
Otoh, quite often the iron being used really isn't up to the task... and I have heard IT coupled amps that were reticient on the top end...
I think Jensen makes transformers that tout a "Bessel" response (properly applied in circuit). That's tough to do. Even more difficult to achieve as the iron size gets physically larger (they make line and mic level stuff).
My experience with iron says that there is usually an "ultrasonic bump" in the response, and the higher it is - like >70khz. - the better the iron will work "in band." Of course a bump that is many dB is not desireable, and is then a "resonance". The bump is supposed to be low Q...
Because it is difficult to get good iron that works down into the bass and up up up into the highs, a lot of folks have taken to bi-amping or tri-amping, using iron that works best within the frequency range that is desired!
A lot of the permalloy core IT iron (especially the Japanese) seems to work well at higher freqs and is not so good at low - but there may be exceptions to that, I certainly have not tried all the iron out there, nor have the folks that I talk with experiences all combined add up to trying it all...
Part of the resulting response that a typical feedback toobe amp gives is a bit of overshoot on the leading edge of a square wave... without FB using triodes the leading edge (barring an IT/coupling issue) is often seen on a scope as a bit rounded, lacking that little peak at the front edge. Listening, that translates into a very very different sound to the ear. Especially with a reactive load (a real loudspeaker + xover) the sonic difference is often shifted around depending on the load and often increased between the two types of amps.
The phase shift could indicate that the iron involved isn't wide band enough... guess that is always a possibility.
Speaking of which, maybe post a pic or two of the square wave response into a resistive load at ~50Hz., 1kHz., and ~10khz.? That ought to tell a story...
_-_-bear
I think Jensen makes transformers that tout a "Bessel" response (properly applied in circuit). That's tough to do. Even more difficult to achieve as the iron size gets physically larger (they make line and mic level stuff).
My experience with iron says that there is usually an "ultrasonic bump" in the response, and the higher it is - like >70khz. - the better the iron will work "in band." Of course a bump that is many dB is not desireable, and is then a "resonance". The bump is supposed to be low Q...
Because it is difficult to get good iron that works down into the bass and up up up into the highs, a lot of folks have taken to bi-amping or tri-amping, using iron that works best within the frequency range that is desired!
A lot of the permalloy core IT iron (especially the Japanese) seems to work well at higher freqs and is not so good at low - but there may be exceptions to that, I certainly have not tried all the iron out there, nor have the folks that I talk with experiences all combined add up to trying it all...
Part of the resulting response that a typical feedback toobe amp gives is a bit of overshoot on the leading edge of a square wave... without FB using triodes the leading edge (barring an IT/coupling issue) is often seen on a scope as a bit rounded, lacking that little peak at the front edge. Listening, that translates into a very very different sound to the ear. Especially with a reactive load (a real loudspeaker + xover) the sonic difference is often shifted around depending on the load and often increased between the two types of amps.
The phase shift could indicate that the iron involved isn't wide band enough... guess that is always a possibility.
Speaking of which, maybe post a pic or two of the square wave response into a resistive load at ~50Hz., 1kHz., and ~10khz.? That ought to tell a story...
_-_-bear
The comment about the ST70 is a good one (actually, this is great thread, thanks all). It's not generally known the open-loop bandwidth of the ST70 is really pitiful - something like no more than 3~5 kHz before it starts rolling off. The amp requires feedback to get flat response, not to mention getting distortion down to reasonable figures.
I think the problem is a combination of poor performance from the 7199 pentode section and the choice of driving the EL34 grids directly from the split-load inverter. People who have built ST70's with a discrete pentode front end and triode split-load inverters (EF86 and 6SN7) report much lower distortion and greater bandwidth - and I'm sure they sound different, too. It has to be remembered the 7199 was originally chosen because it was a cheap, readily available television part, and lent itself to the simple ST70 circuit. Both the pentode and triode sections were considerably inferior to commonly available discrete equivalents of the day - the decision was based on cost and size, not electrical performance.
With the 7199 having ridiculous NOS prices today (thanks to all the ST70 chassis out there), it makes no sense at all to continue using the 7199. The Dynaco design error that runs the cathode of the triode section at a voltage well above the published H-K limits of the 7199 assures the tube of a short life in a standard circuit, an additional reason to alter both the tubes and the circuit from the original.
Although the Dynaco decision to drive the EL34's from the split-load inverter exacts a penalty in driver linearity, the circuit is much more stable than a Williamson, which were always on the verge of instability. The big secret of the original Williamson was the extraordinary performance of the original Partridge output transformers - replace those, and the stability goes way down. The "cleanness" of the ST70 in comparison to an equivalent Williamson is mostly the result of much better square-wave performance and greater insensitivity to load reactance.
As for cathode degeneration/stabilization resistors, 100 ohms is too much. Remember the value of the resistor is multiplied by mu by the time it appear on the plate - so a 100 ohm resistor adds another 2K to Rp, which about doubles the Rp of the 5687/7119 family. That's a substantial degradation. I'd pick a more modest value, something like 10~20 ohms - that should be plenty. Be sure to use a top-quality wirewound resistor - Mills or Ohmite, not those awful sand-cast things. Do NOT use a wirewound for the grid resistor - that has to be a good-quality Allen-Bradley carbon-comp or a modern SMD chip resistor (these are really good, by the way).
I think the problem is a combination of poor performance from the 7199 pentode section and the choice of driving the EL34 grids directly from the split-load inverter. People who have built ST70's with a discrete pentode front end and triode split-load inverters (EF86 and 6SN7) report much lower distortion and greater bandwidth - and I'm sure they sound different, too. It has to be remembered the 7199 was originally chosen because it was a cheap, readily available television part, and lent itself to the simple ST70 circuit. Both the pentode and triode sections were considerably inferior to commonly available discrete equivalents of the day - the decision was based on cost and size, not electrical performance.
With the 7199 having ridiculous NOS prices today (thanks to all the ST70 chassis out there), it makes no sense at all to continue using the 7199. The Dynaco design error that runs the cathode of the triode section at a voltage well above the published H-K limits of the 7199 assures the tube of a short life in a standard circuit, an additional reason to alter both the tubes and the circuit from the original.
Although the Dynaco decision to drive the EL34's from the split-load inverter exacts a penalty in driver linearity, the circuit is much more stable than a Williamson, which were always on the verge of instability. The big secret of the original Williamson was the extraordinary performance of the original Partridge output transformers - replace those, and the stability goes way down. The "cleanness" of the ST70 in comparison to an equivalent Williamson is mostly the result of much better square-wave performance and greater insensitivity to load reactance.
As for cathode degeneration/stabilization resistors, 100 ohms is too much. Remember the value of the resistor is multiplied by mu by the time it appear on the plate - so a 100 ohm resistor adds another 2K to Rp, which about doubles the Rp of the 5687/7119 family. That's a substantial degradation. I'd pick a more modest value, something like 10~20 ohms - that should be plenty. Be sure to use a top-quality wirewound resistor - Mills or Ohmite, not those awful sand-cast things. Do NOT use a wirewound for the grid resistor - that has to be a good-quality Allen-Bradley carbon-comp or a modern SMD chip resistor (these are really good, by the way).
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.