Push-Pull transformers and gaps.

[James D]

Hi Dave,

I've designed and built many solid state and tube amplifiers over the years both SE and PP. My preference is for PP amps and for lots of iron i.e input transformers, interstage transformers and output transformers. My designs are, therefore, global feedback free. As so often is the case, this is a legacy of ones education and in my case this is certainly true. I was taught electronics by a small group of enthusiasts starting at the age of 9 when they were in their sixties and seventies. This was in the mid sixties so they learnt their electronics in the twenties and thirties. They liked the amps they were brought up on and didn't think much of most fifties and sixties designs... not because they used feedback or pentodes but because they were not unconditionally stable and had very poor rfi design. Intersting isn't it? They designed audio amps for good rf performance, or rather for well defined rf performance. Use of transformers was regarded as mandatory because of the isolation of ground planes and loops... the golden rule was that no one stage should affect the performance of any other stage except via the well defined signal loop (voltage and/or current). Interesting eh???

Anyway, the purpose of this ramble is that they taught me to always spec. my PP transformers for at least 10mA unbalanced dc current as the gap this caused in the core would linearise the transformers performance under varying magnetisation levels i.e. varying ac current. I have always followed this rule as I found it always resulted in better sound than very high inductance, no gap transformers.

I have also found that best sound from a PP output stage usually is not at the dc current balance point but up to 15% away from that point but this I attributed ( as did my mentors) to the best dynamic balance point being slightly away from the best static balance point. However I have been wondering recently if this isn't due to increased 2nd harmonic in the 'out of balance' performance leading to increased third harmonic cancellation after intermodulation within the tube is taken into account...

But anyway, I thought this tied in rather nicely with the experience of members of this forum that larger than normal gaps lead to better sound... and that 'oldtimers' knew this but it dropped out of circulation during the fifties and sixties... like so much other tube lore.

So onto the big question. Do you make PP transformers? Input, interstage and output? Would you make then out of pinstripe cores i.e. 50%:50% nickel/steel. What sort of price would they be?

best regards

James

[dave slagle]

. the golden rule was that no one stage should affect the performance of any other stage except via the well defined signal loop (voltage and/or current). Interesting eh???

you betcha!!! its anecdotes like this that most people gloss over without seeing the true value and inspiration!

Anyway, the purpose of this ramble is that they taught me to always spec. my PP transformers for at least 10mA unbalanced dc current as the gap this caused in the core would linearise the transformers performance under varying magnetisation levels i.e. varying ac current. I have always followed this rule as I found it always resulted in better sound than very high inductance, no gap transformers.

can't add anything, just wanted to read it again

I have also found that best sound from a PP output stage usually is not at the dc current balance point but up to 15% away from that point but this I attributed ( as did my mentors) to the best dynamic balance point being slightly away from the best static balance point.

the important thing is the improvement! we can discuss all day why this might or might not be, but lets not forget the improvement is there! So often people will discuss the cons of the various explanations (low level zero crossing, DC further linearizing the core, added even order etc.) and forget the net result is what matters.

i'll talk all day about the possibilities as to why, but often we lose sight of the real important factore and that is it happens to sound better

However I have been wondering recently if this isn't due to increased 2nd harmonic in the 'out of balance' performance leading to increased third harmonic cancellation after intermodulation within the tube is taken into account...

Mark kelly has had some thoughts on this topic. there was a recent post in the asylum that was similar...

http://www.audioasylum.com/forums/tube ... 7198.html

But anyway, I thought this tied in rather nicely with the experience of members of this forum that larger than normal gaps lead to better sound... and that 'oldtimers' knew this but it dropped out of circulation during the fifties and sixties... like so much other tube lore.

i love it when two unrelated things seem to fit together so nicely! Thanks for adding this. If nothing else it gives us some support for our outdated ideals.

So onto the big question. Do you make PP transformers? Input, interstage and output? Would you make then out of pinstripe cores i.e. 50%:50% nickel/steel.

i am a student of the craft... i have read all of the stuff out there on how you are supposed ot make a PP transformer, some I agree with and some I don't . I have some very distinct ideas on a PP design and have sombody i'm talking with about a 8K PP trannie for class A PP 50's. Your best bet is to start up a discussion about what you need and see where it leads. The one thing i refuse to do is tell someone what is best, since that assumes i know.

I don't need to go into the need for the airagap for PP trannies, but i'll add i'd do it on a split bobbin wound as two discrete SE transformers with one section reverse wound. The secondaries would be placed in series with the CT used as the ground. I now have the nerve to mention this publicly since previous mentions have gotten me lectured on how the primaries need to couple to each other in order to assure proper cancellation of the even harmonics. I now know where this lecture originates from. As it turns out the origins of the close coupling comes from class B PP usage (and class AB) where the design heavily relies on the H2 cancellation. (the answer actually turned up in RDH4 as an explanation as to why we needed to couple the primaries).

I do have to admit i am not a fan of the whole pinstripe concept. i look at it as using a 45 in parallel with a 211 and saying the 45 will give you low level detail and the 211 power. If you want to try the whole pinstripe thing, i'll gladly encourage you in any way i can and gladly learn from your efforts.

dave

[Guest]

To add to what Dave just said; if you must mix core materials, I think arranging them in stripes is probably the worst way to go about it; the corners of an EI stack are prone to all sorts of edge effects etc, posibly local saturation. I'd put the nickel stuff in the middle of the core meself.

But then, if it wasn't feasible to make the whole core from nickel, I'd go amorphous C-core anyway...
Cheers
shines

[dave slagle]

I'd put the nickel stuff in the middle of the core meself.

i'd take it a step more. i'd do it in two discrete sections and isolate them from each other with a gap of some sort. the last thing you want is the flux that is supposed to be confined to the M6 sneaking into the nickel and locally saturating it. (think of what happens in the cross grain part of the E when there is some nice nonoriented nickel right next to it)

I'd suspect a piece of paper cut in the shape of an EE to match the lams would make a nice nickel to M6 barrier.

i'd still expect to see an odd "kink" to the loop when you hit the point where the lower Bsat material saturates.

dave

[James D]

Isn't it funny how people can get so caught up in discussing the mechanisims behind observations that they forget entirely the purpose for the observations in the frist place? I sometimes think we humans would far rather debate an issue than progress that issue... he said rambling on again.

Thank you for the Mark Kelly link. I have been following Lynn Olsen's re-discovery of the WE work on third harmonic cancellation and trying to generate a model for it. I think I understand the mechanisim now but can't produce any sensible equations for it. I hope Mark has more success - he is certainly more fluent in theory than I am. His A2 speculation is very interesting and your observation of how Lynn O. sets up his PP amps is fascinating. I need to think about that and re-read some references. BTW have you read John Broskie's explaination of using a similar mechanisim with SE stages to optimise their linearity? http://www.tubecad.com/2005/March/blog0039.htm

8K PP trannies sound interesting! That's very close to the optimum for an amp I've just started building. I was after 10K but 8K would work well too.

I think your suggested structure for a PP OPT sounds very good for Class A PP. The coupling we are after is into the secondary, equally from each primary - so as long as dc current field cancels rather than saturating the core, we should be swinging. I have to say I don't understand the RDH4 argument ... I thought in Class B that only one side of the PP was conducting at any one time so how can the two primaries generate fields than cancel? I must be missing something obvious... the most I can see it helping would be softening cross-over distortion.

I have no pre-disposition for pinstripping beyond thinking that it can look pretty I really like your analogy. I haven't particularly liked the amorphous core transformers that I have heard. They seem to have slight sterile to me. I have spent ages getting the tone of my amplifiers just the way I like it i.e. neutral - by my reckoning, and amorphous cores seem a backward step. I like the 50/50 cores that I have heard and wonder what the best way of arranging the mix is. Any suggestions or references that I might pursue? I guess I can experiment with various stacking schemes and see what works best?

What are the considerations that I am missing for an output trannie? I not up on the theory of how to build a transformer at all so some guidance or pointing at references would be appreciated. I guess RDH4 would be a start...

How are the considerations for a PP:PP interstage and for an SE:PP phase splitting input transformer (or phase splitting interstage) different from above?

Sorry to load you with questions, I know just enough to ask a transformer manufacturer for what I want him to build but nowhere near enough to discuss how to build it! This is exciting but scary, my ignorance is on show in public - so now I know that I am alive

shines

Thank you for the data point I will take that into account.

James

[Guest]

I'd put the nickel stuff in the middle of the core meself.

i'd take it a step more. i'd do it in two discrete sections and isolate them from each other with a gap of some sort. the last thing you want is the flux that is supposed to be confined to the M6 sneaking into the nickel and locally saturating it. (think of what happens in the cross grain part of the E when there is some nice nonoriented nickel right next to it)

I'd suspect a piece of paper cut in the shape of an EE to match the lams would make a nice nickel to M6 barrier.

i'd still expect to see an odd "kink" to the loop when you hit the point where the lower Bsat material saturates.

dave

Ah yes, good point - that cross grain would likely spoil things.
I'm not convinced there would be a kink in the loop; in my mind, at this is looking at the bulk characteristics, not local effects, if the Ni approaches Bsat, its perm drops. This would mean that the M6 would pick up flux, reducing that in the Ni. I'd expect to see a gradual, smooth handoff hence an unkinked loop.

But then again, this is for the bulk material. Throw in the grain mismatch, local saturation, etc etc, and who knows?

You're right Dave, it would certainly be interesting to find out what happens with different arrangements. Hope I haven't discouraged you too much James!

Besides, Dave - didn't you mention once that a mix of M6 and mumetal had a very linear characteristic...

Cheers,
shinebox

[dave slagle]

I'm not convinced there would be a kink in the loop; in my mind, at this is looking at the bulk characteristics, not local effects, if the Ni approaches Bsat, its perm drops. This would mean that the M6 would pick up flux, reducing that in the Ni. I'd expect to see a gradual, smooth handoff hence an unkinked loop.

i see it this way... if we assume the gap fixed you will have two simplified loops one with the slope of the air-steel mix and the other of the air nickel mix. the slope of the nickel mix being steeper. if both cores are operating below the knee, they will simply average out. when you hit the knee of the lower Bsat material, it will slowly leave the mix leaving you with the lower total perm. To add insult to injury you now have 1/2 the core area which causes the remaining flux to increase at 2X the rate causing a momentary increase in perm until the second knee is hitl... i can see the kink from here. I will agree it may not be easily visible, but place a 45 in parallel with a 211 and see if you see the kink in the transfer function.

Besides, Dave - didn't you mention once that a mix of M6 and mumetal had a very linear characteristic...

actually it was 80% nickel and 50% cobalt... and the "alleged" linearity was presented to me by two different unconnected sources. I have no personal experience with the mixture. I do have 80% nickel, cobalt, and M6 in the EI75 size so if somebody want to play with various mixtures on a plate choke, i'm all for it.

Ferromagnetisim by Bozorth actually has an entire section on nickel cobalt alloys... but this is a different animal altogether than lam mixing. if this indeed were a valid approach, i would expect to see some pure nickel lams and some steel lams and a mix your own alloy setups. why bother with 49% nickel, 80% nickel etc, when pinstriping suggests 8 in 10 lams of nickel will get you 80% and 5 in 10 50%... then just throw in a slice or two of coblalt and you are cooking ...big

i'll accept the concept as valid when a proponant of the idea also supports dissimilar tubes in parallel since its the same concept in my book.


Cheers,
shinebox[/quote]

[dave slagle]

he said rambling on again.

I resemble that remark

I have to say I don't understand the RDH4 argument ... I thought in Class B that only one side of the PP was conducting at any one time so how can the two primaries generate fields than cancel? I must be missing something obvious... the most I can see it helping would be softening cross-over distortion.

hmm... i always thought of class B as class A with a ton of even order distortion you are right though... hmmm.

I like the 50/50 cores that I have heard and wonder what the best way of arranging the mix is. Any suggestions or references that I might pursue? I guess I can experiment with various stacking schemes and see what works best?

i'm all ears for your results, my only opposition to it is on a conceptual level, and we all know how far that gets us.

What are the considerations that I am missing for an output trannie? I not up on the theory of how to build a transformer at all so some guidance or pointing at references would be appreciated. I guess RDH4 would be a start...

i think for PP trannies the "mystery" is in the top end leakage and capacitttance are your enemy, and unmached leakages and capacitances are the devil incarnated (sorry for getting dramatic) my goal is to wind two of the best damn SE transformers as i can on a common core and call it a PP to PP and take it form there.

How are the considerations for a PP:PP interstage and for an SE:PP phase splitting input transformer (or phase splitting interstage) different from above?

with the PP to SE you have a phase reversal in one of the windings so capacitances cannot be matched. this is just another conceptual thing (like pinstriping) so i make no guarentees to its actual merit.

for restels 1:2's for his dac there was a phase reversal and early on i voiced concern about it (you cannot series bifilar windings without it) the saving grage was the small core and relatively small amount of turns. as a general rule, if you are going to do it keep the impedances involved as small as possible (can you say a 600R input transformer wired SE to balanced to split phase

i have had good luck with a 1:1 bifilar 600R input transformer feeding a CT choke for ground reference to split phase. the CT choke replaces the resistors normally used to derive a fake CT and can provide matched capacitances. I haven't tried this with higher impedances though. The 1:1 gives you bandwidth out the wazoo until you flip a ground (or move it to midpoint) so a high turn bifilar 1:1 with a CT secondary is a trainwreck, but at 600 ohms even the resistors keep things civil... the CT choke just puts those evil "non-conductors" where they belong. (the bin)

dave

[James D]

Chaps,

I'm sorry but I'm buried in work today and tonight so I won't be able to reply until tomorrow.

This discussion is great and certainly making me think - I can't wait to 'fix' a design to try

James

[James D]

I have just had a re-read of the thread and it's making me ask some fundamental questions about what I'm trying to achieve... Previously I have had good to great results by asking for a 10mA or thereabouts gap in the push-pull OPT that I have used. The best amps that I have made have been 'Amity' style for want of a better description. In that they had all the benefits of DHT SE amps but with better macro dynamics (not just power - 6B4G tubes are my favourites so about the same power in PP as a 300B in SE). However recently i.e. the last 12 months or so I have been concious that they are not quite a good as my SE amps at the smallest level of microdynamics i.e. they seem to miss the really low level nuances. Allen Wright says that putting a CCS on the cathode and making the amp differential restores this but I have not found that to be the case. I find it hurts the amp in terms of tone reproduction and macro dynamics whilst adding nothing to the microdynamics - so for me it's a dead end. I note that Mark J (IslandPink) has found the same and found that changing to an OPT from Bud Purvine (after using a Lundahl) restored the low level information. SO I am trying to find a way of improving my PP amplifers to get them to equal the SE performance with microdynamics whilst losing none of the other advantages. I hoped that using mixed cores would get some way towards this without losing out too much on other factors but maybe it introduces too many other compromises? It seems pretty clear that mixing lams will cause some steps in the transfer curves and that a full blended material such as amorphous would be more consistant in it's transfer charateristic. So I guess one needs a fully M6 or Mu-metal or amorphous core rather than mixed lams. Funny that MQ seem to get good results with pinstripping...

I am hopeful that your suggested approach to a PP OPT will produce good results. If we treat the output stage as two SE amps with a shared cathode resistor and just have signal summation in the secondary then we might be onto something. Of course this assumes that the core just supports the two fields independently with it's magnetisation level merely the average of the two fields applied - so not explicitly summed but implicitly averaged... Is this accurate below saturation? Is the core just there to support a denser field from each primary across to the secondary than would otherwise be the case? And is it true that as long as it does this below saturation and above the intial knee then it doesn't impose it's own character onto the field? I would like to think so but I can't quite believe it is that simple. I think the field has imposed onto it the signature of the core material that is supporting it... but I'm not sure how... For instance in the case of microdynamics, these are usually present with higher level signals as well - so they are not intrinsically lost due to 'zero-point' crossing but they are due to the signal being asked to jump through a small (very small) BH loop in the presence of a larger( or many larger) BH loop. By my understanding that should be a very linear response since the larger signal overcomes hystersis leaving the small signal free to drive its own loop. However if we are losing microdynamics it appears that it might be that the small loop has difficulty establishing itself in the presence of the big loop - a BH 'momentum' effect that is not exactly hysterisis but has some of the character of hysterisis - of course it might also be that the loop is below the single magnetic domain strength limit and therefore just doesn't register... I don't know what the answer is or even if my speculation is valid... I just want my microdynamics to be as good with my PP amps as they are with my SET!

Meanwhile back at the plot!



Dave wrote

with the PP to SE you have a phase reversal in one of the windings so capacitances cannot be matched. this is just another conceptual thing (like pinstriping) so i make no guarentees to its actual merit.

Ahh Ok, so that's why many people have reported CT chokes as sounding better than SE:PP trannies. I had wondered why and couldn't really see why - in fact I thought they would be worse as they didn't give you the interstage ground isolation that a trannie does.

Dave wrote

i have had good luck with a 1:1 bifilar 600R input transformer feeding a CT choke for ground reference to split phase. the CT choke replaces the resistors normally used to derive a fake CT and can provide matched capacitances. I haven't tried this with higher impedances though. The 1:1 gives you bandwidth out the wazoo until you flip a ground (or move it to midpoint) so a high turn bifilar 1:1 with a CT secondary is a trainwreck, but at 600 ohms even the resistors keep things civil... the CT choke just puts those evil "non-conductors" where they belong. (the bin)

Ok so I build this as a 1:1 trannie followed by a CT choke. That makes sense to me I understand the 600R preferred impedance for signal level trannies too and actually since I started using 'Aikido' gain blocks I have nice low output impedances to drive the trannies with too Serendipity.

So should the amp topology be 1:1 input trannie to autoformer volume control to 'Aikido' gain/driver stage to 1:1 IT to CT choke to PP output stage to PP OPT? Assuming Class A1 throughout.

James

[dave slagle]

they seem to miss the really low level nuances.

This has been my experience too... Short term my PP amps are fun, but long term they never seem to satisfy. Somehow a good SE amp makes you forget the amp is there where even a good pp amp does something to let you know its there.

It seems pretty clear that mixing lams will cause some steps in the transfer curves and that a full blended material such as amorphous would be more consistant in it's transfer charateristic. So I guess one needs a fully M6 or Mu-metal or amorphous core rather than mixed lams. Funny that MQ seem to get good results with pinstripping...

Let me state again, I have no ears on experience with mixing lams and my only objections are theory based. There just seems to be something fundamentally wrong with the idea. Say you mix 80% nickel and M6, this puts 1/2 your core saturating at 7Kg and the other half at 18Kg. I'm not sure I could sleep at night knowing that all the power the M6 was buying me was coming at the cost of saturation of 50% of the core. If I were ot mix lams I would take the material with the lowest Bsat and use that as my maximum B for the design. I'll gladly empower someone to play with the idea and eat crow if the results are stunning. Looking at the other design choices from the folks who report the brilliance of the concept leads me to believe that we may just like different things. The expression "polar opposite" comes to mind.

I am hopeful that your suggested approach to a PP OPT will produce good results.

Me too

If we treat the output stage as two SE amps with a shared cathode resistor and just have signal summation in the secondary then we might be onto something.

In my perfect world, I would just want the coils to couple at DC this actually had me thinking for a bit that maybe M-19 might not be such a bad core for a PP transformer. Actually I really have to wonder about the use of any grain oriented material for audio, and given the choice, I might prefer M-19 to M6... Simply put, grain orientation gives you a square loop and non-oriented materials are round loop.

Of course this assumes that the core just supports the two fields independently with it's magnetisation level merely the average of the two fields applied - so not explicitly summed but implicitly averaged...

I like the thought that it would be a simple addition rather than a complex interaction, but suspect the later is always the case.

Is the core just there to support a denser field from each primary across to the secondary than would otherwise be the case?

In a perfect (lossless) world, according to lenz law the power consumed by the load does not add any flux to the core. This suggests to me that given no losses a core could deliver infinite power

And is it true that as long as it does this below saturation and above the intial knee then it doesn't impose it's own character onto the field? I would like to think so but I can't quite believe it is that simple.

I like to look at the BH loop in the same way I look at a set of tube curves. With tube curves we want parallel equally spaced lines. With BH loops (remember they widen with frequency) we want parallel narrow loops that do not widen with frequency.

I think the field has imposed onto it the signature of the core material that is supporting it... but I'm not sure how...

Me either, getting back to lenz, the power transferred to the load does not cause flux in the core, but the the load isn't the only thing the source has to drive. The inductance and the losses can both be very nonlinear and make the source distort.

For instance in the case of microdynamics, these are usually present with higher level signals as well - so they are not intrinsically lost due to 'zero-point' crossing but they are due to the signal being asked to jump through a small (very small) BH loop in the presence of a larger( or many larger) BH loop.

This is where it gets tough to conceptualize. At any given instant in time there is just one single value of B and H, the only way multiple loops can exist is if you designate multiple timeframes.

By my understanding that should be a very linear response since the larger signal overcomes hystersis leaving the small signal free to drive its own loop.

This suggest there could be two different levels of B and H at a given time. I'll suggest that the small signal simply modulates the large signal. Think of it this way, Look at the loop derived from a 20hz signal that just reaches the Knee of the loop. It will appear as a diagonal "cats eye"

Now look at the loop from a small signal 1K signal on the same scale, it will appear as a short diagonal line (if you zoom in you will see it open up to the cats eye shape)

Now add the two signals together and look at the signal on the scope, you will have a 20hz sine made up of 1K "squiggles" the loops add in the same way. If you look over the period of 1/20th a second the 1K loops will be lost in the 20hz major loop, but if you zoom in on part of the loop you will see the 1K loops modulating the outline of the 20hz loop. Now change your timeframe to 1/1000 a second and look at the loops of the 1K signal. At a 1K timeframe, the 20hz signal just looks like a variable DC bias that follows a very nonlinear pattern so you would see a series of 1000 minor loops that would take the slope of the outline of the 20hz loop. This variable slope equates to a variable load for the minor loop over the period of the major loop. That’s IM distortion!

However if we are losing microdynamics it appears that it might be that the small loop has difficulty establishing itself in the presence of the big loop

How about the small loop is unstable in the presence of a larger loop??

of course it might also be that the loop is below the single magnetic domain strength limit and therefore just doesn't register

Oh how I want to know what that level is... Everyone seems so concerned where cores fail to resolve wrt saturation, but what about the other way. Since we avoid saturation by moving in the other direction, shouldn't we consider at some point we are going to hit a wall in the other direction?

My gut feel is the domains can reslove down to -200dB but if nickel goes to -200dB, maybe steel only goes to -150dB (remember steel goes higher wrt Bsat) also remember that 20-20K represents a 60Db range of flux.

... I don't know what the answer is or even if my speculation is valid... I just want my microdynamics to be as good with my PP amps as they are with my SET!

Me too... And discussion makes us all think.

Ahh Ok, so that's why many people have reported CT chokes as sounding better than SE:PP trannies. I had wondered why and couldn't really see why - in fact I thought they would be worse as they didn't give you the interstage ground isolation that a trannie does.

Are you talking a CT choke to split phase??? Time for a new thread and a new rant simply put its a 1:1 inverting interstage with crappy response. My guess is the saving grace is the noninverting half has perfect response so there are no sum and difference issues from having slightly different frequency responses (beat frequencies??)

Ok so I build this as a 1:1 trannie followed by a CT choke.

Remember this is really unproven... The logic is they use this all the time at 600R for SE to balanced conversion with resistors. The CT choke just dispenses with the resistors and mimics the "ideal" 1:1CT

So should the amp topology be 1:1 input trannie to autoformer volume control to 'Aikido' gain/driver stage to 1:1 IT to CT choke to PP output stage to PP OPT? Assuming Class A1 throughout.

Well, remember the autoformers are single ended, two can be assembled on a common core for balanced (and this works great from what I hear) and now that I think of it, a balanced autoformer might just do the trick of the CT choke perfectly... Thanks!!!

The 1:1 IT for the driver to the output can also be bifilar mirrored pairs on a common core (assuming you are at 300V or so DC offset) I have ideas for the 1KV you need for the 833's but lets save that for another day

Thanks for making me think a bit and remember some old ideas!

dave

[James D]

Hi Dave,

Sorry about the delay in responding I've been a little busy with work and things. Also I've been reading a few references on PP OPTs - that and your last post have really got me thinking!

Dave said:

Somehow a good SE amp makes you forget the amp is there where even a good pp amp does something to let you know its there

Most of the time, I agree but I find SE amps get confused on complex orchestral pieces. One of my favorite pieces is Mahler's 8th symphony. This has a big big orchestra and a choir of up to 1,000 voices. I have yet to hear an SE amp that can keep good seperation on the different parts of this symphony. I find the same with complex pop and rock music. However the PP amps can cope with this complex mix much better. I want the best of both in an amp and I don't care if it's SE or PP just as long as it is the best at everything!

Let me state again, I have no ears on experience with mixing lams and my only objections are theory based. There just seems to be something fundamentally wrong with the idea. ... snip ...Looking at the other design choices from the folks who report the brilliance of the concept leads me to believe that we may just like different things. The expression "polar opposite" comes to mind.

I have to say that your reasoning feels right and doing a little research I think I can see what you mean about "polar opposites". I guess one day it will be worth trying to hear and to measure what difference it makes but I think I need more convincing now I've heard your reasoning and thought it through a little.

In my perfect world, I would just want the coils to couple at DC Smile this actually had me thinking for a bit that maybe M-19 might not be such a bad core for a PP transformer. Actually I really have to wonder about the use of any grain oriented material for audio, and given the choice, I might prefer M-19 to M6... Simply put, grain orientation gives you a square loop and non-oriented materials are round loop.

It took me a while to figure this one out but I got there ( I think), we want the primary fields to cancel at dc completely ( counter wound primaries) but at ac we want each primary to couple completely to it's secondary. We then run the secondaries together with the centre tap at ground.

If we build the PP as two SE OPTs then do we care about dc cancellation? I guess the question is does the dc cancelation get us anything other than incresed inductance? Well, yes, it gets us a much smaller and therefore cheaper transformer overall - so much so that an equivalent spec. PP OPT is cheaper than an SE OPT rather than twice the price...umm it seems that building the transformer with dc cancelation is a real world advantage...

In a perfect (lossless) world, according to lenz law the power consumed by the load does not add any flux to the core. This suggests to me that given no losses a core could deliver infinite power

Umm - interesting interpretation of Lenz Law. Let me think - Lenz Law is necessary for Conservation of Energy, a fundamental law of the universe. It is telling us that perfect coupling does not exist - there are always losses in the system... hence your smiley - phew I'm getting there.

But the question I'm still thinking on is 'Do we need a core?' I remember reading about an Italian chap who had built air-cored OPTs. They were big but the worked and the core couldn't impose its signature onto the signal because there wasn't a core! Except air... What do you think? To what degree would this work?

Ahh I found the website :http://www.autocostruire.it/old-site/trials/index.htm.

I understand the point about lossy mechanisims forming part of the load and therefore causing the source to distort by imposing an extra load - it also causes distortion by removing parts of the signal so that coupling is less efficent for these parts than others and possible by resonance efects causing enhanced coupling effects too. If we remove the core what are the remaining loss mechanisims?

This is where it gets tough to conceptualize. At any given instant in time there is just one single value of B and H, the only way multiple loops can exist is if you designate multiple timeframes.

I like this remark! Doing timeframe slices is exactly the way the ear/brain system seems to work! So analysing the system response in a sympathetic way to our hearing mechansism suggest that multiple magnitude BH loops might be exactly what we need to consider! We need the BHloop equivalent of an FFT to probe this and understnad what is happening Maybe not.

...That’s IM distortion!

Yes! and IM is very good at smearing and/or obscuring low level details... I think you may be onto something there. So how can one minimise this effect?

My gut feel is the domains can reslove down to -200dB but if nickel goes to -200dB, maybe steel only goes to -150dB (remember steel goes higher wrt Bsat) also remember that 20-20K represents a 60Db range of flux.

Downward dynamic range as Allen Wright says... It's really important that we come up with a way of measuring this. I need to think on this and read a whole lot more. Have you any ideas?

I've got to stop now...sorry much more to add. I'll complete it tomorrow.

James

[dave slagle]

I want the best of both in an amp and I don't care if it's SE or PP just as long as it is the best at everything!

Agreed. I just don't want to think how good or how bad an amp is. If I constantly am reminded how good an amp is by its sound, I find I fall out of love with it in time (the emperor's new shows always get dirty). It's the amps that disappear that I want.

I need more convincing now I've heard your reasoning and thought it through a little.

Well the ultimate convincing comes from ears on experience, but that is difficult to do so most people just accept and hear what they are told. If everyone says amorphous is the best, people just ask for amorphous and assume they have the best.

It took me a while to figure this one out but I got there ( I think), we want the primary fields to cancel at dc completely ( counter wound primaries) but at ac we want each primary to couple completely to it's secondary. We then run the secondaries together with the centre tap at ground.

Yes... That is the idea, I'm not sure if its what "we" really want... Its just an idea. A very interesting twist to this would be to ground and float the CT. With the CT grounded you could have imbalance currents between the halves, but with it floating we have a series circuit and both secondaries will have the same currents (difference between PP and differential?)

If we build the PP as two SE OPTs then do we care about dc cancellation? I guess the question is does the dc cancelation get us anything other than incresed inductance?

Good question, I can make a convincing argument either way

Well, yes, it gets us a much smaller and therefore cheaper transformer overall

I don't accept the "cheaper" argument. It makes to large of an assumption. I say design the best damn transformer you can and then attach a price, do start a design with a price in mind. Once you know what is best (yeah right) then you can look to "cut corners" to trim the price if required.

- so much so that an equivalent spec. PP OPT is cheaper than an SE OPT rather than twice the price..

That depends on where your money is going. I'll suggest a "decent" PP transformer should be more than a comparable SE. When it comes down to it, the bulk of the $$$ in a transformer goes to the time spent winding and not to the materials involved. Materials cost for a M-6 is like $3 a pound wire about $6 so say a 10 pound trannie has 7 pounds of core and 3 of wire we have $31 in materials for a $300 transformer. Going PP and dropping it to 6 pounds will net a $10 savings in materials, but suddenly you have a more complex winding. This is what takes the time, the physical machine winding time is trivial compared to the taping, insulating, lead attachment etc. and this is often more work for a PP trannie. Its actually double the work if you consider the dual SE approach.

.umm it seems that building the transformer with dc cancelation is a real world advantage...

And some real world disadvantages (possibly) I am not convinced that the 4 quadrant operation of PP is a problem vs. the single quadrant of DC biased SE, but I do wonder about the even order cancellation causing an unnatural harmonic spectrum.

But the question I'm still thinking on is 'Do we need a core?' I remember reading about an Italian chap who had built air-cored OPTs. They were big but the worked and the core couldn't impose its signature onto the signal because there wasn't a core! Except air... What do you think? To what degree would this work?

I guess it would work to the degree that the numbers (L, Leakage and C) allow it to. I think its incorrect to assume that since there is no core, the transformer will have no signature though since you still have the big three (L leakage and C) and for example the lack of L will have a distinct effect on the sound even if it is perfectly linear.

I understand the point about lossy mechanisims forming part of the load and therefore causing the source to distort by imposing an extra load -

But even with 0 core loss, a lack of inductance will cause distortions.

If we remove the core what are the remaining loss mechanisims?

I wonder if we have to break it down into two parts (the L and the reflected load) and see how the losses effect each.

The two key differences between the two are phase, and the idea that inductances cannot dissipate power. So we have the load drawing power and the inductance storing and releasing current. In a lossless world with 0 source impedance all is well, but when we consider losses and the Rp of the tube as a source impedance we start to get complicated. (the partridge articles cover this)

multiple magnitude BH loops might be exactly what we need to consider!

For any given period of time there will only be a single loop, multiple loops cannot exist under a single timeframe. All of the texts present a series of different timeframes on a single graph which leads some to believe they exist at the same time. For any period of time there will be a complete loop, but I see the multiple loop model as an artifact of measurement and not something that exists... The larger loops (in flux value) will dominate and the smaller ones will modulate the outline (did that make sense?)
Quote:

Yes! and IM is very good at smearing and/or obscuring low level details... I think you may be onto something there. So how can one minimise this effect?

Make the BH loop a BH line an airgap goes a long way to do this. I don;t see its necessary to go to the extreme with an aircore, but using low loss cores with large airgaps seems like a good compromise. Take a core with a perm of 1000 and gap it to get 100 and people will say you are throwing all the perm away. I disagree, you still have 100 times the perm of air with something that closely resembles the straight BH line of air, just with a greater slope (and saturation)

Downward dynamic range as Allen Wright says... It's really important that we come up with a way of measuring this. I need to think on this and read a whole lot more. Have you any ideas?

The only thing I can come up with is how much a large signal 20hz sine will screw up a small signal 2K sine. This should be measured in a real world situation. I'm not sure how to look at it though. It would be very interesting to look at the 2K sine as you remove the core and vary the size of the 20 hz modulation tone. This wouln't give any hard numbers, but sometimes a picture is worth more.

dave

[Paul Barker]

"the physical machine winding time is trivial compared to the taping, insulating, lead attachment etc. and this is often more work for a PP trannie."

You're not wrong. Drives you mad when you spend all that time and it didn't work anyway. I have bobbins full of wire all over the house which I have no idea what failed idea they came from!

[jarek]

Make the BH loop a BH line Smile an airgap goes a long way to do this. I don;t see its necessary to go to the extreme with an aircore, but using low loss cores with large airgaps seems like a good compromise. Take a core with a perm of 1000 and gap it to get 100 and people will say you are throwing all the perm away. I disagree, you still have 100 times the perm of air with something that closely resembles the straight BH line of air, just with a greater slope (and saturation)

Apparently Grossner (have I already thanked you for reffering me to that excellent book?) cites some Prtridge results, showing that introduction of the airgap alone DOES NOT change the "distortion". In his case "distortion" is the 3rd harmonic, as he considers unbiased core. As far as I understand, what they seem to consider as a source of the harmonic distortion is just the loop itself, i.e. they assume some shape of it (I think following Rayleigh two parabolas) and derive the 3rd harmonic content as a geometric ratio of areas of some portions of the loop. Since this ratio is unchanged (loop area is the same, airgap reduces inductance, etc) they conclude that the "distortion" is unchanged too.

However, I think they are missing exactly the point you have raised - with the airgap loop becomes closer to the line. If the 3rd harmonic is related just to some area rations, surely higher order harmonics should be related to the shape of the loop borders too. In order words I expect that the higher the harmonic, the more detailed information of the loop shape it depends on. I suspect that all the irregularities of the loop get straightened by introduction of an air gap. I tried to plot an example but failed (stupid Maple or me).

-jarek

[IslandPink]

Hi Guys,

Just wanted to throw in a few comments which seem relevant to the discussion , mainly to respond to James's comment that I had solved the 'missing low-level information' problem , by going to the O-Netics output transformers on my amps .
As background, you need to know that I'm currently using ( exclusively ) a version of Lynn Olson's 'Gary's Aurora ' circuit, which was where those guys were about 2 years ago with the circuit :

http://www.nutshellhifi.com/GarysAurora.gif

It's very close to this design scheme right now, with a 76 at the front, but driving the very good ( and cheap ) Sowter 3575 input/phase-splitter interfacing to the ECC99 halves .

I've been running first the Lundahl 1620 outputs, then the O-Netics level 1 6.6 K outputs, now the O-Netics level III similar 6.6K outputs . I think they are well run-in now after about 2-3 months in place .

So to get to the nub of the issue, I had a visitor round on Sunday who brought a rough breadboarded SE amp using nearly the same tubes ( his were KR 300B XLS , mine VV32B ) and the normal Lundahl outputs ( 1623 ? ) and the Aurora is NOT THERE YET !
Now , I must go into as much detail as I can here. This current version of the Aurora is very strong on dynamics, detail and bass ( it always was ) . It's also much , much better with the O-Netics outputs than it was with the Lundahls . There is also , now , especially with the level IIIs , lots of tonal colour and tonal differentiation on all vocals and instruments, right through from bass to treble . The only aspect it is slightly behind on , is the 'emotion' and 'flow' aspect, as best I can describe it . This , technically, can be noticed in careful listening to be associated ( at the least ) with a slight loss of very small stuff , perhaps more so in the bass regions , what I would call the 'stuff between the notes' . The other listeners agreed on this , too . This is surprising too , in that the Auroras are MORE capable than this SE amp in reproducing soundstage , room size, note decay and reverb or echo-reflection aspects .
It's got to the stage where I feel I'm going to have to build an SE amp to a reasonable standard , alongside this project , to compare against . If you listened to the PP amp in isolation, you'd say it had everything , it's so much better than a typical PP !
Now, I haven't got to grips with the possible experiements that could be done regarding applying imbalance into parts of the circuit . I think it would be tricky with the Aurora in its current form ( answers on a postcard ? ) .
It may be that the remaining loss of whatever makes SE special comes down to the extra two transformers in the path , not any technical issues with the outputs . I know Bud Purvine had some in-depth ideas about how to make good PP outputs . The level IIIs use 0.009" M6 lams I think , 'special' seperation layers ( dielectrics ) and 140/40 litz wire for the secondaries . I don't think they are gapped in the conventional sense , I think there are three sections to the core stacking, though . It's interesting that the more basic Lundahl outputs on my friend's SE amp were able to do a better job on the low-level stuff, so it would seem . As I say, maybe the outputs are not the issue here ?

Anyway , comments appreciated .

Mark J

[IslandPink]

OK, more info ...
Built up a 2A3 SE amp using James outputs and EL84 drivers as per Thorsten's 'Legacy' with voltage-divider screen set-up . Power supply is a bit naff.
The areas the SE wins are in the upper midrange and treble, where the tonal 'integration' or realism is more convincing . There's not as much low-level detail on the SE as with the Auroras actually, Lynn is correct , I've changed my feelings on this , but I think the tonal or dynamic gradations are better in the upper frequencies on the SE amp . It's hard to describe of course.
I don't know yet whether this is actually transformer related . I'll swap 5687s for 6N6ps as the drivers soon, then maybe try EL84 as above but a pair ( obviously ), RC-coupled first, see if I can recreate the more finessed top-end at the expense of a little of the Aurora grip and clarity .
Lower mid and bass tone is better with the Aurora at the level of development of this simple SE amp anyway .

How much 2nd , 4th order are we seeing in this PP circuit , does anyone know ? I don't know if the differences I'm hearing are just due to this harmonic profile difference or not . The two ought to tend towards the same at low levels in a simple theory, but that's not quite what I'm hearing .

MJ

Ps. Last post for a week or so, I'm off on hol .

[Paul Barker]

Not fair to compasre base performance of an se 2a3 on Ariels with a push pull suped up 300b. Not a level playing field.

[IslandPink]

Hi Paul,
Well, that wasn't really my point ; agreed, you'd have to go a lot further with the SE amp to compare bass , timing etc etc in a valid way . What I think it shows is you don't need to try too hard to get good midrange and treble tone with the SE amp, it almost comes for free .
Mark

[Paul Barker]

Your right there. The base costs about another 20 grand if you go my route.