Loosening up magnetic "stiction"

[jkeny]

My limited knowledge in all this suggests to me that there might be something in the low level details getting through the transformer - I know John Swenson used a a Lundahl 1660 transformer on the output of his BDT preamp. Now I know there are too many variables in this but when replaced it with a Dave Slagle (*80% Nickel) transformer he reported significant increase in retrieval of detail AFAIR.

 

[smoking-amp]

"with a Dave Slagle (*80% Nickel) transformer he reported significant increase in retrieval of detail AFAIR."

Well, 80% nickel (permalloy) will certainly have a lot less hysteresis/magnetizing current than M6 etc, so will be less distorting typically. The permeability is high from zero flux up to below saturation level. (while M6 starts out with a lower Mu and then noticeably peaks half way up to satur.)

According to Bozorth, permalloy also produces its Barkhausen noise from the knee of the Hyst. curve upward. (Where-as M6 produces B noise in the steepest section) So low level signals that stay away from saturation in permalloy will avoid the noise. Wish I owned a Nickel mine.

I should mention one ameliorating affect for SE xfmrs: Due to the lowered flux swing available with DC current present, they typically use 3 or 4 times as much steel as a conventional P-P xfmr. This causes any B. noise to average out over the more material present.

My gut feeling is that B. noise isn't going to be a real noticeable factor in big output xfmrs. anyway. Little line and phono type matching xfmrs. on the other hand are made with selected materials, so the noise issue has probably been selected out there historically as well. But there are always possible exceptions.

Don

 

[jkeny]

Two things that strike me:
- in the diagram of the application of the JLTi LEM induced permeability the signal seems to be input at the front of the amplifier rather than at the OPT - surely this can't be optimal & what effect a HF (40 KHz -100KHz) signal has on the amplifier if indeed would it be filtered out before the reaching the OPT? See here: http://www.asonaudiophiles.com/ason/default.htm

- Would it be possible to use an SMPS as B+ feeding the center tap of the OPT to achieve the same result? Presumably one that operates in the 40 - 100KHz (or whatever is in the OPT range) & lets these HF signals through ie. badly designed?

Edit: - Finally, I have to say I like their concept of taking a reasonable tube amp import from Asia & applying these mods (LEM & super linear triode) to get a high performing unit!

 

[smoking-amp]

"Would it be possible to use an SMPS as B+ feeding the center tap of the OPT to achieve the same result? "

Unfortunately in this case, the center tap has great PSRR due to the two windings having opposite phase. So the HF won't get injected into the core unless the two windings are on different parts of the core. Even then, the HF magnetic path would be thru air mostly.

Keep in mind that a low output Z tube stage will overcome low permeablility problems. Just using some local plate feedback is likely to work fine for fixing low Mu and Barkhausen noise too.
Yes, I did say fix the Barkhausen noise too. With a very low impedance driven winding on the core, the domains cannot avalanche. Change in flux is solidly controlled by voltage with a zero impedance source. Maybe putting silver wire on the primary could help a little too, but the major impact by far will come from lowering the tube output Z.

Yes, the add-on concept is a nice one. You can probably build a LEM equivalent for $25 if you want one. Still a question is remaining in my mind about leaving the HF in the amplifier output though. Putting an LC filter in on the 8 Ohm side to remove that will mess the damping factor. No filter, and maybe good by tweeter. Not to mention howling dogs and cats. Might keep the mosquitos away though. Isn't that HF stuff what the military uses for subdueing unruly crowds?

Don

 

[Joe Rasmussen]

"Wehrmacht" was actually my nickname- it was a WWII-era setup using a single driver...

Thanks, you have set my mind at ease and ditto on the horns.

Joe R.

 

[Joe Rasmussen]

Two things that strike me:
- in the diagram of the application of the JLTi LEM induced permeability the signal seems to be input at the front of the amplifier rather than at the OPT - ...

- Would it be possible to use an SMPS as B+ feeding the center tap of the OPT...

Edit: - Finally, I have to say I like their concept of taking a reasonable tube amp import from Asia & applying these mods (LEM & super linear triode) to get a high performing unit!

Hi John

Yes, common mode and also the low Z of the power supply would make it impossible to shift/modulate.

I have have found that the ideal place to inject the signal is on the input gain stage's cathode or if differential driver stage with individual cathode resistors, pick one of them. Keep in mind that the signal must be held constant, irrespective of audio signal level aiming at around 0.5V P/P across a dummy load on the secondary. To inject directly into OPT is simply not practical and even if possible would require a more costly solution.

Needless to say the result IS going to depend on the quality/behaviour of the OPT, but I don't think that the LEM is a cure-all. There are other things that also needs attention, such as getting the Anode Z as low (and stable) as possible. Did somebody say something about local feedback? Oh yes, Don. Use DC/resistor feedback around the Grid coupling cap as indeed Menno does (but uses U-L which I don't like). That is worth looking at even if that was not my chosen solution. I use VTO, Virtual Triode Operation, which I find complements the LEM. Re magnetisation effects, I will not put myself up as an authority, but with Push-Pull OPT any asymmetrical DC current either side of the centre tap is audible. I have heard this quite clearly and working on a solution, which, IF successful, will match the current of one output tube to another. The combination of these three incorporated in affordable Chinese amps leads to results that are highly satisfying (trying not to sound like an advert).

Joe R.

 

[jkeny]

Thanks Don,
I see (said the blind man). I have used plate-to-plate feedback very successfully in a Baby Huey amp using the very marginal OPTs from a Rogers Cadet III from which I built it. It gets some amazing bass & smoothness out of these small OPTs which I think is due with the partial feedback scheme and the consequent reduced Anode Z?

I was interested in changing out these OPTs for toroidal PS trafos & seeing how far this can be taken.

I'm also building a BDT preamp along with Zeus (using mosfet followers) & SS Tabor (using solid state pentodes) amps (all of which are fully differential) & wondered again about using toroids here.

So these ideas are of interest to me.

Don, I don't think my tweeters will cause the dogs any problems - they are not of the super tweeter variety one being a KEF B27 in a Rogers LS3/5a speaker and the other a Jordan JX92S full range in a DIY speaker (3D spiral horn)

One of the other compelling attractions of this idea is that I have a major problem with casing & chassis construction - it takes up the majority of the build process for me & is the least enjoyable part (I'm sure I'm not alone in this!)

 

[rdf]

DHT outputs with HF driven filaments; problem solved. =D

 

[jkeny]

To a man with a hammer every problem is a nail!!!!!!!!!!

But tell me more about your hammer

 

[cerrem]

"Would it be possible to use an SMPS as B+ feeding the center tap of the OPT to achieve the same result? "

Unfortunately in this case, the center tap has great PSRR due to the two windings having opposite phase. So the HF won't get injected into the core unless the two windings are on different parts of the core. Even then, the HF magnetic path would be thru air mostly.

Keep in mind that a low output Z tube stage will overcome low permeablility problems. Just using some local plate feedback is likely to work fine for fixing low Mu and Barkhausen noise too.
Yes, I did say fix the Barkhausen noise too. With a very low impedance driven winding on the core, the domains cannot avalanche. Change in flux is solidly controlled by voltage with a zero impedance source. Maybe putting silver wire on the primary could help a little too, but the major impact by far will come from lowering the tube output Z.

Yes, the add-on concept is a nice one. You can probably build a LEM equivalent for $25 if you want one. Still a question is remaining in my mind about leaving the HF in the amplifier output though. Putting an LC filter in on the 8 Ohm side to remove that will mess the damping factor. No filter, and maybe good by tweeter. Not to mention howling dogs and cats. Might keep the mosquitos away though. Isn't that HF stuff what the military uses for subdueing unruly crowds?

Don

Modulation of the DC BIAS grid supply is the way to go.....

Chris

 

[Michael Koster]

...
Keep in mind that a low output Z tube stage will overcome low permeablility problems. Just using some local plate feedback is likely to work fine for fixing low Mu and Barkhausen noise too.
Yes, I did say fix the Barkhausen noise too. With a very low impedance driven winding on the core, the domains cannot avalanche. Change in flux is solidly controlled by voltage with a zero impedance source. Maybe putting silver wire on the primary could help a little too, but the major impact by far will come from lowering the tube output Z.
...

Don

Menno says as much in the paper. If he uses a 300B triode instead
of the pentode output stage, even with 105db loudspeakers, there
is not a problem.

Michael

 

[Elias]

Hello,

... Actually the A/B listening test would probably be more effective with normal music. If the effect is worthwhile it should show up there. There's all kinds of differences the ear can hear that instruments don't really know how to measure (yet). And if part of the effect that creates the SE signature is caused by small signal hysteresis, Menno suggested, that "should" show up in listening tests.

I think better than listening test in this case would be comparing samples in numerical domain, i.e. record the amp output with and without the added high freq compensation signal. Can use a sample of music of specific level and record it with e.g. 24bit 192kHz sampling, then syncronize recordings, then subtract them to see the error. Usually the eye is more revealing than the ear.

- Elias

P.S. I have not done this, just an idea.

 

[exeric]

Ummm, I don't think I can agree with most of that Bel Canto quote. Even for a SE DC biased xfmr, flux domains have to flip in order for the flux to change. The DC bias acts just like the internal material coercive forces to pin domains down in one direction. Lowering the permeability as a result. No energy is supplied by the DC field, it would violate energy conservation since no energy is put into the DC field except resistive losses in the winding.

Certainly putting a DC bias on a core will quiet down the Barkhausen noise, but its most likely in direct proportion to how much the permeability has been reduced as well. So the noise reduction is just a good measure of how useless the core has become. High quality materials like permalloy have low internal coercive forces so the domains can move easily. Putting a DC field into the core is equivalent to reducing the material quality.

Don,
I'm no expert on magnetics either but I know some physics. I think you may be confusing the cause for the effect of the Barkhausen effect (love that word!). It appears that the magnetic domains really do have some inertia at low levels of magnetization that results in the so-called stiction. The reason it doesn't appear in SE amps as much is because inertia, of whatever form it takes, is a result of resistance to acceleration.

An SE opt will saturate with the DC running through it unless it has a gap. This means it is running higher up on the magnetization curve of the core at all times. That constant electrical current running through the coil can be thought of like a river that has mass and energy. Even the smallest change in the velocity of that river (the current) automatically implies an acceleration that is sizeable and can overcome that stiction.

A PP opt, at crossover especially, will have current that amounts to a tiny trickle - (don't let your imagination go to far with this) And the change of the current, or acceleration, in direction at the exact moment of crossover approaches zero. Since the current at that moment is so small its electrical inertia is also small, and requires very little energy to change it. And that lack of inertial mass of that small amount of coil current is what creates the stiction.

EDIT: This also makes sense with "generalized" ideas of PP being very dynamic and SE having more low level detail but being TOO smooth. Of course those are generalizations -YMMV

 

[exeric]

Hello,



I think better than listening test in this case would be comparing samples in numerical domain, i.e. record the amp output with and without the added high freq compensation signal. Can use a sample of music of specific level and record it with e.g. 24bit 192kHz sampling, then syncronize recordings, then subtract them to see the error. Usually the eye is more revealing than the ear.

- Elias

P.S. I have not done this, just an idea.

Elias,
Well, maybe... I seem to recall something about CDs and their players having perfect reproduction when they first came out in the 80s. My ears said something different. (Like, holy **** that sounds bad!)

 

[jlsem]

Even for a SE DC biased xfmr, flux domains have to flip in order for the flux to change.

But they flip as relatively small minor AC hysteresis loops and within the positive quadrant of the B-H graph.

John

 

[Joe Rasmussen]

Video: Menno's Lecture

OK, here goes.

Menno's Lecture - Video:

The following links are approx 65MB each for the first three zip files and the last is about half that.

First 20 minutes avi file: www.yousendit.com/download/bVlCM25QcGtRR2VGa1E9PQ

Second 20 minutes avi file: www.yousendit.com/download/bVlEZGVjTkxPSHdLSkE9PQ

Third 20 minutes avi file: www.yousendit.com/download/bVlEZGVlcTJxRTBLSkE9PQ

Last 10 minutes: www.yousendit.com/download/bVlEZGVhV3J6RTkzZUE9PQ

Total 1 hour and 10 minutes. There is a small overlap between the files. Suggest you download the first 20 minutes to see if it is of interest, if so, download the rest.

Files will be available for the next SIX days and a MAX of 100 downloads. First come, first serve.

Downloads guaranteed virus free - at least they when I uploaded them.

Cheers.

Joe R.

 

[jkeny]

Thanks Joe,
That's great - I cheated & downloaded the conclusions as it was the smallest, to see if I wanted the whole recording - I'm now downloading the rest. He favours SE OPTs for better immunity to both the Barkhausen & memory effects of OPTs. How does this square with others experience - I have no experience of SE OPTs (& very little experience of PP OPTs) so trying to gather valid data.

BTW, You can host these files on www.4shared.com indefinitely for free with a file size limit of 100MB & storage limit of 5GB.

 

[smoking-amp]

exeric:
"Since the current at that moment is so small its electrical inertia is also small, and requires very little energy to change it. And that lack of inertial mass of that small amount of coil current is what creates the stiction."

Well, the usual picture given is that the domains have a statistical Gaussian profile of energy barriers to overcome in order to flip. The DC current in SE causes the low barrier ones to permanently flip and pin in one direction. (the air gap prevents all of them from doing so) The higher barrier (or stiction) domains then account for the AC variation remaining. Since the total number of functioning domains has been reduced, the Mu goes down. The air gap has the effect of lowering the flux variation around the core (and hence voltage induction) from any single domain flip or group avalanching. This lowers the Barkhausen noise picked up by the windings. Since a lot more core material has to be used to compensate for the lowered Mu, more noise averaging occurs also. Since the AC flux is utilizing the higher energy barrier (or stiction) domains, these are also less likely ones to avalanche in groups, keeping the flip noise smaller in magnitude but now more numerous (hence averaging out well).

Electrically, one could model this as a noisy (Barkhausen) steel core in series with a quiet air core inductor. The air core inductor filters out the Barkhausen noise, and also reduces the large scale hysteresis by diluting out the hysteresis in the steel core. The model is a little misleading though, since the inductance of the steel core part (and final result too) is greatly reduced (not summed higher).

jlsem:
"But they flip as relatively small minor AC hysteresis loops and within the positive quadrant of the B-H graph."

Yes, but I think the original context was with respect to the much smaller micro variation of flux with domain flips, causing Barkhausen noise. I was just saying that for any given amount of flux change, similar numbers of domains still have to flip in either case (for the case of similar size cores; in reality, larger SE cores with air gaps require more domain flips to get the same flux change around the core).

The macro effect of hysteresis is mainly reduced with DC bias due to the reduced effectiveness of the magnetic material (ie. lowered MU, so less induction per current) and the diluting effect of the air gap causing most of the inductive field energy to be stored in the air gap rather than in the steel.

Don

addendum:
In my earlier posts I did state that the DC biased SE core (with air gap) does quiet down the Barkhausen noise. No doubt about that. My point was that the benefit is in direct proportion to how much the useful (inductance) properties of the core were crippled. To take this benefit/loss ratio to the extreme, we would be using totally air core OTs.

As a practical solution though, the higher Mu of a P-P non-gapped xfmr allows one to use a low Z (relative to the primary winding reactance) driving source very effectively to eliminate Barkhausen noise by strict control of the flux variation.

 

[jkeny]

As I said I've not much experience with OPTs, so I'm relying on other more experienced inmates opinions but is it agreed that the sound differences are SE OPTs pick up more low level detail but have less dynamics & PP is the opposite of this?

Don, comes down firmly on the side of PP OPTs as being less compromised & it's problems are overcome by low Z on the finals anyway (through either Triode use or NFB?) Is it possible, without feedback to achieve low Z using pentode finals?

Is this a fair statement of position?

 

[SY]

I have never found SE to have better low level detail; quite the opposite, my subjective (uncontrolled) perception has always been that they act like make-up, covering the wrinkles and pores.