Some toughts.
You told your shift measurements are differential so your real zero phase could be in the high frequency range.
I don't know the specifications of your transformers,
however (from your first measurements) I would say you have a (parallel) resonance at (or just above) 20KHz.
A parallel resonance means you have the equivalent capacitance and inductance in parallel. It gives rise to a 180 degree shift (going from far below the resonance to far above it).
The phase is conventionally 0 degree at the resonance, so you go from +90 to -90, passing through the resonance.
Such parallel resonance is typical of any transformer: it is the first of a series.
The first resonance is usually due to the coupling of the stray capacitances (all in parallel among them) and leakage inductance of the transformer.
Note that the typical stray capacitances of audio transformers are in the range of nF (can be a fraction of nF or several nF's, depending on their inner coupling and the transformer design itself). So I am not suprised about the bad effect of the 33 nF cap!
If you say there is no problem with high frequencies when listening without the 33 nF, I think you have a resonance given by the parallel coupling of such 33 nF cap with the leakage inductance of the input transformer.
Lets assume the leakage inductance is 1 mH (typical), than you get a resonance at 28Khz!
Very probably it is not well damped beacuse your source has a low impedance.
Maybe if you tell the specification of your input transformer I could be more precise.
Cheers,
45
You told your shift measurements are differential so your real zero phase could be in the high frequency range.
I don't know the specifications of your transformers,
however (from your first measurements) I would say you have a (parallel) resonance at (or just above) 20KHz.
A parallel resonance means you have the equivalent capacitance and inductance in parallel. It gives rise to a 180 degree shift (going from far below the resonance to far above it).
The phase is conventionally 0 degree at the resonance, so you go from +90 to -90, passing through the resonance.
Such parallel resonance is typical of any transformer: it is the first of a series.
The first resonance is usually due to the coupling of the stray capacitances (all in parallel among them) and leakage inductance of the transformer.
Note that the typical stray capacitances of audio transformers are in the range of nF (can be a fraction of nF or several nF's, depending on their inner coupling and the transformer design itself). So I am not suprised about the bad effect of the 33 nF cap!
If you say there is no problem with high frequencies when listening without the 33 nF, I think you have a resonance given by the parallel coupling of such 33 nF cap with the leakage inductance of the input transformer.
Lets assume the leakage inductance is 1 mH (typical), than you get a resonance at 28Khz!
Very probably it is not well damped beacuse your source has a low impedance.
Maybe if you tell the specification of your input transformer I could be more precise.
Cheers,
45
45:
Thanks for the thoughts. You seem to be really zeroing on the input cap, but I think that capacitor is only contributing to a washed out tweeter, nothing else. In the first post of the thread I noted:
All the rising response and phase shift issues are independent of this cap. In fact, when I had experimented with possible values for this capacitor in choosing the HPF frequency, I had swept the system over the audible range looking for resonance. I did find some, with smaller values of capacitance. The higher I went in capacitance, the less prominent the peaking. The values chosen worked good with swept sine waves, without adversely loading the source.
Further, I noted there was zero phase shift and flat response 20 to 20K from source to 12B4 grids. While the capacitor may not SOUND good, it is not contributing any resonance that I can see.
I guess where I'm heading is asking the question if other people's non-GNFB IT coupled designs exhibit the phase shift (constant time delay) I am experiencing. It no longer sounds bad, with the input cap removed, but the shift is still there.
Thanks for the thoughts. You seem to be really zeroing on the input cap, but I think that capacitor is only contributing to a washed out tweeter, nothing else. In the first post of the thread I noted:
zigzagflux said:
All the rising response and phase shift issues are independent of this cap. In fact, when I had experimented with possible values for this capacitor in choosing the HPF frequency, I had swept the system over the audible range looking for resonance. I did find some, with smaller values of capacitance. The higher I went in capacitance, the less prominent the peaking. The values chosen worked good with swept sine waves, without adversely loading the source.
Further, I noted there was zero phase shift and flat response 20 to 20K from source to 12B4 grids. While the capacitor may not SOUND good, it is not contributing any resonance that I can see.
I guess where I'm heading is asking the question if other people's non-GNFB IT coupled designs exhibit the phase shift (constant time delay) I am experiencing. It no longer sounds bad, with the input cap removed, but the shift is still there.
Re: Re: New 300B PP amp completed, phase shift questions
In fact, in the last post, I was speaking in general about transformers as well.
The phase shift could be due to a transformer alone or in combination with the tubes (i.e. one or more differential pairs do have quite high and/or different capacitances within the pair).
The 33pF cap brings an additional and similar effect being in the same (or close) range of stray capacitances.
Cheers,
45
In fact, in the last post, I was speaking in general about transformers as well.
The phase shift could be due to a transformer alone or in combination with the tubes (i.e. one or more differential pairs do have quite high and/or different capacitances within the pair).
The 33pF cap brings an additional and similar effect being in the same (or close) range of stray capacitances.
Cheers,
45
If the tubes are distorting due to excessive capacitance in the loads, this defeats the whole purpose of the Amity/Karna family of amplifiers. The goal is to operate the most linear devices with most linear loading system - but - this is critically dependent on the right choice of tubes, transformers, and operating points.
This is most awkward in the input stage, which has the least ability to drive the stray capacitances of the first interstage. (The 5687's and 7119's I am using now are barely adequate for the job.) Rather going to the additional complexity of a cathode-follower, you might revert to the previous differential input stage that is then simply RC-coupled to the grids of the driver stage. This eliminates the input and first interstage transformers, allows you to select the RC coupling value that gives you the highpass characteristic you are looking for, while retaining the Class A PP driver that is transformer-coupled to the output stage.
You still need a very low-distortion input stage, but the standing current and Zout no longer matter that much with RC coupling. A nice vintage 6SN7 will probably do just fine - and you can keep the current source you already have. Since the ideal coupling caps will be small at these impedances, you have the option of Teflon or other exotics, and don't have to chase after an exotic high-Z interstage transformer for the first interstage. High source impedances are the enemy of good transformer performance, and that's unfortunately what you tend to see in input stages.
To recap, the suggested topology is:
1st Stage: Input is direct-coupled to one side of a differential 6SN7 (or similar) dual triode with a 15~40 mA current-source on the cathodes. The plates are then conventionally RC-coupled to the following driver stage.
2nd Stage: The driver stage is the same as the previous recommendations - NO current source, merely a common resistor to ground, and the previously suggested bypass cap between B+ and common (virtual) cathodes. The grid resistors should be 100K or less.
3rd Stage: As before. No current sources here either.
This gives you an amplifier which is an Acrosound for the first stage, and an Amity/Karna for the driver and output stages.
Sidelight/distraction: If this works for you, an interesting alternative would be a Mullard first stage, with the PP/differential part of the Mullard using transformer coupling at the output (a Mullard with a twist). Brief walkthrough of a Mullard: single-ended first tube is plate-resistor loaded and directly-coupled to the grid of a second stage, which sits at an elevated voltage and uses a long-tail resistor to create a differential pair. The grid of the "other" half of the diff pair goes to two places; a 1 megohm resistor connected to the plate of the first tube, and a large-value bypass cap that goes to ground. If the coupling is RC to the following stage, it is a traditional Mullard, and the kinky variation is transformer coupling instead. But this nutty digression aside, I'd try the straight differential RC-coupled Acrosound version first.
By eliminating two sets of low-level transformers, you should see a lot less phase shift in the whole amplifier, while retaining their advantages in the stages that operate at higher power and at lower impedances.
This is most awkward in the input stage, which has the least ability to drive the stray capacitances of the first interstage. (The 5687's and 7119's I am using now are barely adequate for the job.) Rather going to the additional complexity of a cathode-follower, you might revert to the previous differential input stage that is then simply RC-coupled to the grids of the driver stage. This eliminates the input and first interstage transformers, allows you to select the RC coupling value that gives you the highpass characteristic you are looking for, while retaining the Class A PP driver that is transformer-coupled to the output stage.
You still need a very low-distortion input stage, but the standing current and Zout no longer matter that much with RC coupling. A nice vintage 6SN7 will probably do just fine - and you can keep the current source you already have. Since the ideal coupling caps will be small at these impedances, you have the option of Teflon or other exotics, and don't have to chase after an exotic high-Z interstage transformer for the first interstage. High source impedances are the enemy of good transformer performance, and that's unfortunately what you tend to see in input stages.
To recap, the suggested topology is:
1st Stage: Input is direct-coupled to one side of a differential 6SN7 (or similar) dual triode with a 15~40 mA current-source on the cathodes. The plates are then conventionally RC-coupled to the following driver stage.
2nd Stage: The driver stage is the same as the previous recommendations - NO current source, merely a common resistor to ground, and the previously suggested bypass cap between B+ and common (virtual) cathodes. The grid resistors should be 100K or less.
3rd Stage: As before. No current sources here either.
This gives you an amplifier which is an Acrosound for the first stage, and an Amity/Karna for the driver and output stages.
Sidelight/distraction: If this works for you, an interesting alternative would be a Mullard first stage, with the PP/differential part of the Mullard using transformer coupling at the output (a Mullard with a twist). Brief walkthrough of a Mullard: single-ended first tube is plate-resistor loaded and directly-coupled to the grid of a second stage, which sits at an elevated voltage and uses a long-tail resistor to create a differential pair. The grid of the "other" half of the diff pair goes to two places; a 1 megohm resistor connected to the plate of the first tube, and a large-value bypass cap that goes to ground. If the coupling is RC to the following stage, it is a traditional Mullard, and the kinky variation is transformer coupling instead. But this nutty digression aside, I'd try the straight differential RC-coupled Acrosound version first.
By eliminating two sets of low-level transformers, you should see a lot less phase shift in the whole amplifier, while retaining their advantages in the stages that operate at higher power and at lower impedances.
More thoughts - the data you're showing indicates that the tubes involved cannot handle the capacitances that they're seeing on their plates. Distortion in vacuum tubes isn't just matter of the real (resistive) term of the load, but the reactance as well - and elliptical load-lines have far more more distortion that straight load-lines.
That's the reason I've suggested trying a simple RC-coupled first stage to supply the voltage swing for the driver stage, so it no longer has to drive the stray capacitance of the first interstage transformer. It will have to drive the plate-load resistor, so there is merit in using a high value of B+ and a high value of plate resistor.
A mildly exotic variant of the RC-coupled input stage would use separate inductors between the B+ supply and the plate-load resistors, which would act to increase the effective value of the resistor in the midband (where it counts), while the existing plate-load resistor would buffer the stray capacitance of the inductor from direct exposure to the plate. This means that at DC and at extreme HF, the minimum load would be the plate-load resistor, while in the midband, it would be the much higher value (100K) of the grid resistor of the following stage.
The combination of the inductor, plate-load resistor, coupling cap, and grid-resistor would be a much easier load to drive than the existing transformer, and less reactive as well, since the plate and grid resistors would limit the phase angles and resulting elliptical load-lines to real values at the frequency extremes.
I agree the bias condition of the 46's does not look good. There should be lots of headroom at the 46 grids, and the input shouldn't be working that hard. You need to measure the frequency response of each stage, using direct injection on the primary side of the relevant transformer.
That's the reason I've suggested trying a simple RC-coupled first stage to supply the voltage swing for the driver stage, so it no longer has to drive the stray capacitance of the first interstage transformer. It will have to drive the plate-load resistor, so there is merit in using a high value of B+ and a high value of plate resistor.
A mildly exotic variant of the RC-coupled input stage would use separate inductors between the B+ supply and the plate-load resistors, which would act to increase the effective value of the resistor in the midband (where it counts), while the existing plate-load resistor would buffer the stray capacitance of the inductor from direct exposure to the plate. This means that at DC and at extreme HF, the minimum load would be the plate-load resistor, while in the midband, it would be the much higher value (100K) of the grid resistor of the following stage.
The combination of the inductor, plate-load resistor, coupling cap, and grid-resistor would be a much easier load to drive than the existing transformer, and less reactive as well, since the plate and grid resistors would limit the phase angles and resulting elliptical load-lines to real values at the frequency extremes.
I agree the bias condition of the 46's does not look good. There should be lots of headroom at the 46 grids, and the input shouldn't be working that hard. You need to measure the frequency response of each stage, using direct injection on the primary side of the relevant transformer.
Lynn Olson said:
Lynn, I have no experience with PP transformer coupling driver, however I do have experience in driving a 300B SE with an LC coupled 46 (triode connected).
This amp was a variant (with the cheaper 46 instead of a 71a) of this amplifier:
http://www.audiodesignguide.com/se/katelelo.html
My amp was able of 13W at the (extremely soft) clipping and the driver was still far from going into positive grid field. The power bandwidth was 30Hz-50KHz.
The operative conditions were:
1) driver : 250V / 22 mA (as from data sheet), L = 60H, C = 4,7uF and 10K for the 300B grid resitor.
2) 420V / 70 mA / 4 Kohm for the 300B
The 12B4 is certainly "stronger" than the PT8 (that is, more or less, like a 26).
It seems very strange to me here we have only 20W and the 46 is swinging into positive grid.
Cheers,
45.
Lynn:
Thanks for not giving up on me. I'll give some further opinions on what I hear (though I suspect nothing will surprise you) then I'll have a few questions in hopes of saving the current build.
Had two of my friends over to listen and compare this amp to my existing amp. Interestingly enough, your suggestions are very similar to the existing design; it is basically a Mullard 5-20 design with ST70 iron. 12AX7 gainstage, DC coupled to a 12AU7 LTP with CCS tail, cap coupled to triode connected EL34's for the output.
The basic summary was the 5-20 blows away the new amp. Not by design, of course, but my choices. Even with the input capacitor removed, BTW. Open and detailed soundstage versus a flat presentation, sounding compressed would be the best way to describe it.
1. One preliminary question for you, Lynn. Knowing you use similar iron, do you have the same type of phase shift on your amp? If you do, but your sound is still top notch, then I'll forgo stressing out about it any further. Part of me thinks the missing presence and imaging is in part due to this phase shift throwing off the recombination of sound, but of course I have no idea.
2. I will certainly consider your suggestions above, but if at all possible, I want to try and salvage what I have without replacing the entire front end. Since it works for you, it has to work for me. I placed the 12B4 at the front end because it has a very low Rp and just the right gain needed for my system. I am surprised to hear it may be a limiting factor in the amp. I DO have a population of 7119 tubes. What about paralleling two 7119 triodes for a really beefy driver for the first interstage? Would this not be more than adequate for the first interstage then?
3. If I need to bite the bullet and throw in 45's, I will. Again, if it works for you, it must work for me? However, would it not make sense that the 46 would do just fine at low output, before zero grid voltage is even approached? Because at present, the amp isn't right regardless of volume.
Not trying to ignore your advice, just trying to make right what I've done wrong. The 12B4 being a wrong choice really surprises me. Rp is 1K, how can that be a bad thing?
Thanks for your patience. And thanks, Bud, for some nice trannies, guess I just can't drive them properly.
Thanks for not giving up on me. I'll give some further opinions on what I hear (though I suspect nothing will surprise you) then I'll have a few questions in hopes of saving the current build.
Had two of my friends over to listen and compare this amp to my existing amp. Interestingly enough, your suggestions are very similar to the existing design; it is basically a Mullard 5-20 design with ST70 iron. 12AX7 gainstage, DC coupled to a 12AU7 LTP with CCS tail, cap coupled to triode connected EL34's for the output.
The basic summary was the 5-20 blows away the new amp. Not by design, of course, but my choices. Even with the input capacitor removed, BTW. Open and detailed soundstage versus a flat presentation, sounding compressed would be the best way to describe it.
1. One preliminary question for you, Lynn. Knowing you use similar iron, do you have the same type of phase shift on your amp? If you do, but your sound is still top notch, then I'll forgo stressing out about it any further. Part of me thinks the missing presence and imaging is in part due to this phase shift throwing off the recombination of sound, but of course I have no idea.
2. I will certainly consider your suggestions above, but if at all possible, I want to try and salvage what I have without replacing the entire front end. Since it works for you, it has to work for me. I placed the 12B4 at the front end because it has a very low Rp and just the right gain needed for my system. I am surprised to hear it may be a limiting factor in the amp. I DO have a population of 7119 tubes. What about paralleling two 7119 triodes for a really beefy driver for the first interstage? Would this not be more than adequate for the first interstage then?
3. If I need to bite the bullet and throw in 45's, I will. Again, if it works for you, it must work for me? However, would it not make sense that the 46 would do just fine at low output, before zero grid voltage is even approached? Because at present, the amp isn't right regardless of volume.
Not trying to ignore your advice, just trying to make right what I've done wrong. The 12B4 being a wrong choice really surprises me. Rp is 1K, how can that be a bad thing?
Thanks for your patience. And thanks, Bud, for some nice trannies, guess I just can't drive them properly.
Great thread. lots of important information.
I built a little 5687 6080 amp with interstage phase splitting and it sounded good but had roll off issues. The pointers in this thread have clarified my thinking on a rebuild, keeping the two stages - but splitting at the front with a transformer and then running PP all the way to the 6080 finals. The 6080 works very well at 100V 100mA so the 5687 has plenty of gain and grunt to keep them happy. I was going to build it with differential stages - but your comments have given me pause. I have built a version of Gary Pimms Tabor amp which is differential front to back, so I was totally sold on the idea - but experience is a great educator.
Shoog
I built a little 5687 6080 amp with interstage phase splitting and it sounded good but had roll off issues. The pointers in this thread have clarified my thinking on a rebuild, keeping the two stages - but splitting at the front with a transformer and then running PP all the way to the 6080 finals. The 6080 works very well at 100V 100mA so the 5687 has plenty of gain and grunt to keep them happy. I was going to build it with differential stages - but your comments have given me pause. I have built a version of Gary Pimms Tabor amp which is differential front to back, so I was totally sold on the idea - but experience is a great educator.
Shoog
Experience IS a great educator; it's the lesson that can be most painful 
No problem being the guinea pig, as long as I find my way out of the cage.
I will confirm the prediction that removing the lower CCS and adding the B+ to cathode capacitor improved the sound. Funny thing is, I've found many designs of IT applications with anemic highish Rp drivers, often running in the sub-10mA range. If I'm hearing Lynn correctly, the idea is to run a ballsy driver at every stage of the amp, not just before the output stage. Distortion is important, obviously, but it seems to me if I can produce sweet sound with the 12AU7 in a LTP, which has a poor reputation for linearity, application is critical.
Not to bleaguer, but it is surprising to me a 12B4 doesn't have enough. It is no slouch in linearity or Rp, as is the 46. Most surprising.
If I migrate to the 7119, what about a 2:1 IT ??? Don't need additional gain... Hmmmm....
No problem being the guinea pig, as long as I find my way out of the cage.I will confirm the prediction that removing the lower CCS and adding the B+ to cathode capacitor improved the sound. Funny thing is, I've found many designs of IT applications with anemic highish Rp drivers, often running in the sub-10mA range. If I'm hearing Lynn correctly, the idea is to run a ballsy driver at every stage of the amp, not just before the output stage. Distortion is important, obviously, but it seems to me if I can produce sweet sound with the 12AU7 in a LTP, which has a poor reputation for linearity, application is critical.
Not to bleaguer, but it is surprising to me a 12B4 doesn't have enough. It is no slouch in linearity or Rp, as is the 46. Most surprising.
If I migrate to the 7119, what about a 2:1 IT ??? Don't need additional gain... Hmmmm....
zigzagflux said:
About the 46 (as for many 2A3's) you should need to have a good number and make a selection.....then you can get almost as good linearity as the 45.
With the 45 linearity is never an issue unless it has been heavily used.
Many 45's around look new at a first sight however then you find they have less emission tha expected....
45 is now hard to find in really new condition and is expensive, so ask for serious testing before buying.
If you don't find what you are looking for could also consider something like Emission Lab 45 (solid plate) which can take up to 13W plate dissipation.
zigzagflux said:
IMO, you should go for those IT's designed for maximun overall bandwidth, minimal insertion loss and that can take a little unbalancing with minimal loss in performance.
Typically IT's (and OPT's) designed for minimal insertion loss and extended high frequency response give up something in low frequencies however, in my experience, I cannot hear a difference between a transformer that can go down to 1Hz and another that starts to roll off at 20Hz.
This is practically the philosophy of the best japanese transformer companies.
Cheers,
45
Well, you have two problems to solve - the wrong operating point on the 46's (or maybe they have low emission), and excess capacitance somewhere in the circuit. You have the know-how to get to the bottom of the problem with the 46's, so I'll leave that to you. A little fooling around with a DVM and test resistors (to determine standing currents) should resolve the problem. If the actual currents are within 5% of design, an amplifier should work as it should.
If I were you, I'd buy a pair of Lundahl LL1635's interstages as a cross-check. They are good for 80 kHz (that's what I used in the first version of the Amity), and sound pretty good as well. Replace either the first or second interstage and see what difference this makes to the measured peaking and HF rolloff beyond that. In the Karna I have now, there is NO peaking at 20 kHz, and I would not use any interstages that have peaking in the audible range, no matter how good they sound subjectively.
Since the output transformer has good bandwidth, you might try the RC-coupled differential front end, use the LL1635 for the interstage between the driver and output stages, and see how that works out. There should be no peaking at all with that setup, and plenty of bandwidth.
If I were you, I'd buy a pair of Lundahl LL1635's interstages as a cross-check. They are good for 80 kHz (that's what I used in the first version of the Amity), and sound pretty good as well. Replace either the first or second interstage and see what difference this makes to the measured peaking and HF rolloff beyond that. In the Karna I have now, there is NO peaking at 20 kHz, and I would not use any interstages that have peaking in the audible range, no matter how good they sound subjectively.
Since the output transformer has good bandwidth, you might try the RC-coupled differential front end, use the LL1635 for the interstage between the driver and output stages, and see how that works out. There should be no peaking at all with that setup, and plenty of bandwidth.
In my personal experiments with IT coupling I found that the real problem was in the initial driver stage. The amp in question was a 7119 in PP, driven with a phase splitter transformer, then driving 8 EL 34's in PP parallel. The IT in question was a 1.414 to 1 step down and there were 1 k ohm blocking resistors for each of the EL 34 grids. There were no other passive components in the circuits signal path.
The same flat, though highly detailed sound, was the result. After numerous revisions to the IT, without noticeable audible changes, and in desperation, I paralleled the 7119 driver tubes. All of the sonic lack disappeared.
I have had numerous folks, deeply versed in amplifier design, explain to me why this is an unneeded addition and complexity.
The only other change made to date has been to add ersatz ground plane in the form of Electron Pools, a hotly debated topic here.
http://www.diyaudio.com/forums/showthread.php?s=&threadid=102180&highlight=
This did not change the need for the extra tube, but did provide a significant addition to the dynamic colors and emphasis, applied to music by the performing artist.
The inter stage transformers used in my amp and in the Karna are wound as X coupled devices, with primaries cross coupled in adjacent wells, but as opposite physical side of secondary windings, so one of the primary drive sectors is wound first in one well and last in the other. They are paralleled in this scheme with the secondaries sandwiched and in series for the CT between them.
The core is constructed to provide about 60% self demagnetization on the back half of the B/H curve. Overall they are the lowest storage devices I can build, with no film plastic used in the coil construction. The design is so meta stable that a change in core material from M6 to M3 cuts the distributed capacitance by 30%. By metastable I point to their lack of storage mechanisms and lack of sustainable oscillation. The same design philosophy is found in the OPT and input splitter.
The IT is the most difficult to get a balance with and the use of better core material does not give enough boost to the core permeability to allow significant reduction in turns, thereby offsetting the drop in capacitive induced losses in high frequencies, due to the core material performance increase and the result is a steeper peak in highs at about 30K, with the attendant phase change.
Micro gaps in the core help here, but again are not sufficient to address what is a very difficult design criteria. And the attendant increase in size needed, rather than turns increase, which just serves to drop the distributed capacitance, gets out of hand very quickly, without solving the problem. Going in the other direction and utilizing the natural increase in inductance per turn in a smaller core / shorter length of wire per turn drives wire sizes into the realm where corona, even with careful vacuum impregnation of the coil, will kill the IT in a few thousand hours.
Suggestions for areas to poke my nose into would be welcome as these are not perfect devices by any means.
Bud
The same flat, though highly detailed sound, was the result. After numerous revisions to the IT, without noticeable audible changes, and in desperation, I paralleled the 7119 driver tubes. All of the sonic lack disappeared.
I have had numerous folks, deeply versed in amplifier design, explain to me why this is an unneeded addition and complexity.
The only other change made to date has been to add ersatz ground plane in the form of Electron Pools, a hotly debated topic here.
http://www.diyaudio.com/forums/showthread.php?s=&threadid=102180&highlight=
This did not change the need for the extra tube, but did provide a significant addition to the dynamic colors and emphasis, applied to music by the performing artist.
The inter stage transformers used in my amp and in the Karna are wound as X coupled devices, with primaries cross coupled in adjacent wells, but as opposite physical side of secondary windings, so one of the primary drive sectors is wound first in one well and last in the other. They are paralleled in this scheme with the secondaries sandwiched and in series for the CT between them.
The core is constructed to provide about 60% self demagnetization on the back half of the B/H curve. Overall they are the lowest storage devices I can build, with no film plastic used in the coil construction. The design is so meta stable that a change in core material from M6 to M3 cuts the distributed capacitance by 30%. By metastable I point to their lack of storage mechanisms and lack of sustainable oscillation. The same design philosophy is found in the OPT and input splitter.
The IT is the most difficult to get a balance with and the use of better core material does not give enough boost to the core permeability to allow significant reduction in turns, thereby offsetting the drop in capacitive induced losses in high frequencies, due to the core material performance increase and the result is a steeper peak in highs at about 30K, with the attendant phase change.
Micro gaps in the core help here, but again are not sufficient to address what is a very difficult design criteria. And the attendant increase in size needed, rather than turns increase, which just serves to drop the distributed capacitance, gets out of hand very quickly, without solving the problem. Going in the other direction and utilizing the natural increase in inductance per turn in a smaller core / shorter length of wire per turn drives wire sizes into the realm where corona, even with careful vacuum impregnation of the coil, will kill the IT in a few thousand hours.
Suggestions for areas to poke my nose into would be welcome as these are not perfect devices by any means.
Bud
BudP said:
Hi Bud,
do mean that in the single sector you start with a primary (secondary) and end up with a secondary (primary)?
Why do you parallel the primaries?
This will increase considerably your stray capacitances, for a given impedance. Am I wrong?
Isn't it possible to have all the primaries in series as well?
Cheers,
45
Series primary with interleaved series secondaries, or series / parallel secondaries, is best for SE OPT and mandatory for SE IT. Here I am trying to obtain full voltage coupling on both sides of each secondary, as coupling rules. But I need to do this with equal amounts of DCR as an unbalanced DCR in an IT seems to be a real problem. In series wound SE devices I step up wire dia as I interleave out to max diameter, to offset the drive voltage per sector problems, arising from widely different DCR using the same wire size throughout. In the PP IT, the two halves sandwiching the secondaries are two different wire sizes also, to help minimize this issue. I suppose I could look into a full series primary, in two wells, but I have not seen a benefit to a series primary in PP OPT's and so have not done so to date.
Even the M6 IT's have a signal peak at between 26 and 30 k, of about 2 dB. The same coils have a peak of 6 dB at 30 K when M3 is installed. In both cases 20k Hz frequency and phase are ruler flat, regardless of the drive level, right to hard saturation.
The theoretical inductance difference between M6 and M3 is on the order of about 40% when the anneal is nominal. The typical inductance at 1 volt and 120 Hz is not greatly different in real life though.
I think I am faced with not enough internal capacitance wound in this fashion, certainly the single well, stacked style, either series or parallel primary windings, same core size, same turns, very similar DCR were barely able to get to 12kHz flat and then both phase and FR took a nose dive from there.
I have had to shelve further investigations into IT's due to rather severe business die back in our OEM sector. Just cannot afford to build more prototypes at the moment.
Bud
Even the M6 IT's have a signal peak at between 26 and 30 k, of about 2 dB. The same coils have a peak of 6 dB at 30 K when M3 is installed. In both cases 20k Hz frequency and phase are ruler flat, regardless of the drive level, right to hard saturation.
The theoretical inductance difference between M6 and M3 is on the order of about 40% when the anneal is nominal. The typical inductance at 1 volt and 120 Hz is not greatly different in real life though.
I think I am faced with not enough internal capacitance wound in this fashion, certainly the single well, stacked style, either series or parallel primary windings, same core size, same turns, very similar DCR were barely able to get to 12kHz flat and then both phase and FR took a nose dive from there.
I have had to shelve further investigations into IT's due to rather severe business die back in our OEM sector. Just cannot afford to build more prototypes at the moment.
Bud
Interstage transformers are the hardest of any type to design - Bud Purvine, Per Lundahl, and Dave Slagle have all been valuable mentors. What does them in are the high impedances; input transformers have low impedances on the primary, output transformers have low impedances on the secondary, but alas, interstages have to make do with either Rp (for SE) or twice Rp for PP or differential. So everything is determined by Rp of the driving tube.
It's not all peaches and cream for the driving tube, either. The load as seen from the plate is very high in the midband, but drops and becomes more reactive towards the top and bottom edge of the band. Capacitances add up, and the stray capacitances in the transformer as not as pretty as the well-behaved Miller capacitance of the following tube, which has to be included in the net capacitive load.
The input transformer, if it uses a step-up ratio, can be a very difficult load for the preceding preamp. I found that out the hard way with the first version of the Amity, where I used a 1:4 step-up ratio. Big mistake. It multiplied the Miller C of the 5687/7119 an astounding sixteen times, added its own stray capacitance (winding to case, which is grounded), and then you have the 100~300 pF of the interconnect on top of that. This brought most preamps to their knees; solid-state sounded gritty and harsh, and tube preamps sounded "tubey" and dull.
About the only thing that would drive it were ultrafast video buffers with 1000V/uSec slew rates. Those things are designed to drive 50~100 MHz signals down 75-ohm lines, so the Amity was no problem. When I backed off the step-up ratio (to 1:2), more preamps became acceptable. Never overestimate the ability of your preamp to deliver current - most can't, unless they are headphone-capable.
With the exception of microphone and moving-coil transformers , most transformers sound better if there's a lot of linear current to push them. You don't connect a transformer to the plate of a 12AX7, in other words, but something more like a power tube, and the more power, the better. Interstages are perfectly well suited for connecting the driver and output section, since they offer many advantages of Class A1/A2 operation, instantaneous recovery from overload, and if built right, excellent phase match that is level-invariant. Lots better than the split-load inverter of a Dynaco, in other words.
If you do parallel both sections of a 5687 or 7119, don't forget to use grid and cathode-stopper resistors, otherwise the thing may oscillate at RF frequencies.
The output and driver section of the Karna are close to flawless, and distortion is extraordinarily low. I am not as happy with the input, which is contributing most of the upper-harmonic distortion (5th and beyond). Admittedly, this is at very low levels - in the -80 to -100 dB range - but you don't see anything at all when the 45's are driven. This is confirmed by the amplifier changing its sonic character when different input tubes are tried. I am serious when I suggest an alternate input circuit, since both the input tube and first interstage are too parts-sensitive for my taste.
This isn't quite as bad as it sounds; the rest of the amplifier is so transparent that even tiny changes in the rest of the system are very audible. With a more conventional amplifier, I don't think this kind of small change would be audible at all.
As suggested earlier, straighten out the 46 bias issues by feeding signals direct to the 46 grids, or through the interstage, as you like. You will then have a functioning (if low-gain) amplifier. If there are any questions at all about the current delivery in the first stage, consider paralleled 7119's, a monster tube like the 6H30, or a quick conversion to RC-coupling.
When the capacitance issue goes away, you will hear it in the first few seconds of listening. It isn't subtle. Remember, by doubling the current and halving the Rp, the trouble frequencies are now twice as high. That can make all the difference.
It's not all peaches and cream for the driving tube, either. The load as seen from the plate is very high in the midband, but drops and becomes more reactive towards the top and bottom edge of the band. Capacitances add up, and the stray capacitances in the transformer as not as pretty as the well-behaved Miller capacitance of the following tube, which has to be included in the net capacitive load.
The input transformer, if it uses a step-up ratio, can be a very difficult load for the preceding preamp. I found that out the hard way with the first version of the Amity, where I used a 1:4 step-up ratio. Big mistake. It multiplied the Miller C of the 5687/7119 an astounding sixteen times, added its own stray capacitance (winding to case, which is grounded), and then you have the 100~300 pF of the interconnect on top of that. This brought most preamps to their knees; solid-state sounded gritty and harsh, and tube preamps sounded "tubey" and dull.
About the only thing that would drive it were ultrafast video buffers with 1000V/uSec slew rates. Those things are designed to drive 50~100 MHz signals down 75-ohm lines, so the Amity was no problem. When I backed off the step-up ratio (to 1:2), more preamps became acceptable. Never overestimate the ability of your preamp to deliver current - most can't, unless they are headphone-capable.
With the exception of microphone and moving-coil transformers , most transformers sound better if there's a lot of linear current to push them. You don't connect a transformer to the plate of a 12AX7, in other words, but something more like a power tube, and the more power, the better. Interstages are perfectly well suited for connecting the driver and output section, since they offer many advantages of Class A1/A2 operation, instantaneous recovery from overload, and if built right, excellent phase match that is level-invariant. Lots better than the split-load inverter of a Dynaco, in other words.
If you do parallel both sections of a 5687 or 7119, don't forget to use grid and cathode-stopper resistors, otherwise the thing may oscillate at RF frequencies.
The output and driver section of the Karna are close to flawless, and distortion is extraordinarily low. I am not as happy with the input, which is contributing most of the upper-harmonic distortion (5th and beyond). Admittedly, this is at very low levels - in the -80 to -100 dB range - but you don't see anything at all when the 45's are driven. This is confirmed by the amplifier changing its sonic character when different input tubes are tried. I am serious when I suggest an alternate input circuit, since both the input tube and first interstage are too parts-sensitive for my taste.
This isn't quite as bad as it sounds; the rest of the amplifier is so transparent that even tiny changes in the rest of the system are very audible. With a more conventional amplifier, I don't think this kind of small change would be audible at all.
As suggested earlier, straighten out the 46 bias issues by feeding signals direct to the 46 grids, or through the interstage, as you like. You will then have a functioning (if low-gain) amplifier. If there are any questions at all about the current delivery in the first stage, consider paralleled 7119's, a monster tube like the 6H30, or a quick conversion to RC-coupling.
When the capacitance issue goes away, you will hear it in the first few seconds of listening. It isn't subtle. Remember, by doubling the current and halving the Rp, the trouble frequencies are now twice as high. That can make all the difference.
45
p1A p2b
s1 s2
p2a p1b
windings are separated by a molded center flange, 0.-45" thick and contained by outer flange walls of the same thickness. Material is glass filled nylon.
I did try a more elaborate primary secondary interleave, seeking to increase coupling surface and reduce winding depth, but ended up with too much dielectric material and a pretty wild phase event, a sharp peak at 30 khz and then a near straight fall. Sound quality was horrible.
Also tried setting the primary windings adjacent to one another in opposing wells and that also presented problems with very irregular response above about 9 k and a bad phase roll.
As Lynn has pointed out this is a difficult device to get to work and sound completely transparent. I have heard these IT's and some I built for Gary Pimm, and know that when the various stages can drive the load as seen through the IT the sound can be as good as the most expensive capacitors, without the negative sound stage vs amplitude pumping that capacitors all provide. The IT's in the Karna are very transparent and very music competent devices, able to pass an infinite shading of emphasis and tonal color, but at the moment mine don't look good to test gear, out of the audio bandwidth by a 1/2 octave or so.
Do please try putting some muscle in that driver stage, the IT there will thrive on the resulting load.
Bud
p1A p2b
s1 s2
p2a p1b
windings are separated by a molded center flange, 0.-45" thick and contained by outer flange walls of the same thickness. Material is glass filled nylon.
I did try a more elaborate primary secondary interleave, seeking to increase coupling surface and reduce winding depth, but ended up with too much dielectric material and a pretty wild phase event, a sharp peak at 30 khz and then a near straight fall. Sound quality was horrible.
Also tried setting the primary windings adjacent to one another in opposing wells and that also presented problems with very irregular response above about 9 k and a bad phase roll.
As Lynn has pointed out this is a difficult device to get to work and sound completely transparent. I have heard these IT's and some I built for Gary Pimm, and know that when the various stages can drive the load as seen through the IT the sound can be as good as the most expensive capacitors, without the negative sound stage vs amplitude pumping that capacitors all provide. The IT's in the Karna are very transparent and very music competent devices, able to pass an infinite shading of emphasis and tonal color, but at the moment mine don't look good to test gear, out of the audio bandwidth by a 1/2 octave or so.
Do please try putting some muscle in that driver stage, the IT there will thrive on the resulting load.
Bud
Many thanks for chiming in, Bud. Much appreciated info from the master of magnetics!
Your tip about paralleling 5687/7119's is something I think I'll be trying - twiddle with the current sources to raise the total standing current 1.4~2 times, put in grid-stoppers on each grid (at the pins of course), and reserve a separate 5687/7119 for each phase (thus side-stepping electron leakage from one half to the other, which can have weird effects on distortion).
The comment on the "dry", analytic sound is well noted - that's what an unhappy, not-optimized Amity or Karna amplifier sounds like. With no feedback and an extremely simple straight-through circuit, there's nowhere for coloration to hide.
P.S. I'm now working out at the local gym with a personal trainer to get from 90% recovery to the full 100%. Three weeks in, and I'm starting to see good results - strength is up 20% thus far, with more to come. Thanks again for kick-starting the "Beyond the Ariel" thread all those months ago!
Your tip about paralleling 5687/7119's is something I think I'll be trying - twiddle with the current sources to raise the total standing current 1.4~2 times, put in grid-stoppers on each grid (at the pins of course), and reserve a separate 5687/7119 for each phase (thus side-stepping electron leakage from one half to the other, which can have weird effects on distortion).
The comment on the "dry", analytic sound is well noted - that's what an unhappy, not-optimized Amity or Karna amplifier sounds like. With no feedback and an extremely simple straight-through circuit, there's nowhere for coloration to hide.
P.S. I'm now working out at the local gym with a personal trainer to get from 90% recovery to the full 100%. Three weeks in, and I'm starting to see good results - strength is up 20% thus far, with more to come. Thanks again for kick-starting the "Beyond the Ariel" thread all those months ago!
BudP said:
Thanks for your response Bud.
OK, the windings arrangement is as I hoped....i.e. you start with a primary and end up with the same type.
With more than 2 primaries, for example 3 primaries and 2 secondaries, the number of turns of the outer primaries are 1/2 of the inner, I guess.
About your elaborate interleaving it seems you ended up with the same resonant frequency but a different Q.
The dominant stray capacitance comes from the "interface" between primaries and secondaries.
While depends much less on the turn-turn coupling and layer-layer coupling of the same type.
The capacitance is proportional to the surface and the (average) lenght of the turn.
Also it depends on the insulating material and the thickness.
Classical Latheroid paper gives 4 pF/square cm for 0.25 mm thickness and 5pF/square cm for 0.15 mm thickness.
If you increase too much the winding surface without reducing accordingly the depth and/or use too much insulator you end up with more capacitance.
On the other end, increasing the number of sections and/or reducing the depth of the winding you reduce the leakage inductance.
One has to find the best compromise which changes every time, from one transformer to another!!
I agree with you it is expensive and above all it requires a lot of time.
45
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.