Ultimate bandlimited SE 801A transformer

Design and use of the various types.
Post Reply
shinebox
Posts: 88
Joined: Fri Jan 28, 2005 9:44 am

Ultimate bandlimited SE 801A transformer

Post by shinebox »

Dr Dave,

What are your thoughts for an all-out device for the (dreaded high impedance) 10Y/801A?

I'm thinking of the following:

Op point: 500V - 600V, 40mA max

1) midbass channel
100Hz to 800 Hz
4W at 100Hz (will prob x/o 1st order at 150 or 200Hz)
10k pri, inductance required of 50 - 80H?
4 and 16 ohm secondary connections (don't need the 8 ohm tap!)

2) upper midrange
As above but 800Hz-7kHz

What do you reckon, particularly as regards making something really, really special? Could the new namglas alloys be employed to good effect for (2) without breaking the bank? Also, is litz or silver wire worth considering.

Going to be going to ETF this year and will surely be itching to build some 10Y amps afterwards...

Best,
shines
JeffreyJ
Posts: 320
Joined: Fri Jan 28, 2005 2:53 am
Location: stuck in the USA for now...
Contact:

I love the idea...

Post by JeffreyJ »

I even got as far as talking to Voltsecond about advantages of bandwidth limited iron.. and then a quote on amoprphous parafeed versions with matching chokes from Tribute.. so I guess what I am saying is please build it! I have been talking about it for over a year.. Dave and I even talked about it last week when I stopped in.. he does believe there are advantages..

I was gonna do three way.. run the treble light.. maybe down at 20 mA or so.. just to be able to use those old globes.. then the mids the same, but with some reserve in the iron in case i wanted to run 35mA.. and the bass? Use 801A's and run them hard... I think we want some headroom down low.. mine was going to be a single supply 6x spud amp.... I still can envision an amp with a front row of six 10's.. ahhh...

Peace,
Me
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Re: Ultimate bandlimited SE 801A transformer

Post by dave slagle »

shinebox wrote:Dr Dave,

What are your thoughts for an all-out device for the (dreaded high impedance) 10Y/801A?
if you break it up, the high impedance is not an issue. 10 octaves of bandwitdh is cake, so i'd look to place the operating range in the center of the 10 octaves. (equal bandwidth above and below) maybe shift an octave in the direction of the X-over.

1) midbass channel
100Hz to 800 Hz
4W at 100Hz (will prob x/o 1st order at 150 or 200Hz)
10k pri, inductance required of 50 - 80H?
4 and 16 ohm secondary connections (don't need the 8 ohm tap!)
Big core, for lots of turns and low DCR. I won't even bitch about the taps since they only seem to screw things up at the top.

What do you reckon, particularly as regards making something really, really special? Could the new namglas alloys be employed to good effect for (2) without breaking the bank? Also, is litz or silver wire worth considering.
the namglas is really really really costly at this point. It is like 10X the cost of the amorphous in cut core form.

I'm really not sure of the "right" approach, but there are two different directions and i can argue either.

You could use amorphous for the bass, 49% nickel for the mids and 80% nickel for the highs and on paper it makes a lot of sense, but i always go back to something gordon says. He claims if you are going top multi-amp you need to use the same amp for everything otherwise the sound will suffer. I see the logic, but think you can split the difference between the two. I think conceptually the amps should be the same, but i do suspect a few slight changes would help the big picture out. I might stick with the same core material (amorphous) for everything to start and then just scale the outputs for the frequency range. For lack of an easier way to illustrate it, i'd just but the bass trannie in a "copier" and shrink it down 50% for the mids and another 50% for the highs so you have 3 nearly identical transformers that are just different sizes.

dave
shinebox
Posts: 88
Joined: Fri Jan 28, 2005 9:44 am

Post by shinebox »

Thanks for the reply Dave. I figured the taps wouldn't be an issue at those frequencies.

I heard the Gordon argument before, but I'm not sure I buy it fully (he also prefers M4 to Nickel which I find hard to fathom, but that's fair enough). I think if the amps are topologically similar that goes most of the way; the difference between speaker drivers is much bigger anyway if you ask me.

Gonna have a think. Despite the raw expense, I might consider the namglas for the treble transformer. A pal of mine heard namglas vs regular amorphous on a 2A3 PP and set the difference wasn't subtle.
What would the core cost for the 800-7KHz tx, roughly?

Anyway, insanely busy at work for the next couple months by the look of things but I really would like to get some intact iron at some point, hopefully will build a couple of amps up over the Christmas break.

BTW, seen the very honorable mention in 6moons? See the review of the FAB speakers...

Best,
cv
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

good point on the difference between drivers, personally i still like keeping things as similar as possible at least to start out with. Then at least you can hope to get a handle on future changes. I just don't like the idea of starting the game with lots of variables.

Think $600 for a set of namglas cores for treble outputs. (approx 5cm square cross section double C core "shell" configuration)

These cores will be size compatible with the metglas line so a comparison could be made to the more pedestrial AMCC core at 1/10 the price :-)

If you spring for the cores and let me play with them for a few months I'll wind the OT's for you gratis :-) I also have a quote for a slightly larger version that would allow for 80% nickel to be swapped on :-)

dave
jarek
Posts: 62
Joined: Wed Feb 02, 2005 10:02 am

Post by jarek »

Think $600 for a set of namglas cores for treble outputs. (approx 5cm square cross section double C core "shell" configuration)
sweet jesus...i guess you are talking about nano namglas 4? what´s the mass of such cores, i.e. price per kg? besides, with such a baby its a shame to go cut core (3-4x coertion increas). zen + manual toroid winding?

best
jarek
piotr_grzesik
Posts: 68
Joined: Mon Feb 07, 2005 9:59 pm
Location: UK

Post by piotr_grzesik »

So is this stuff suitable for full-range transformers? (apart from price) Or is it better suited to lower-level stuff? According to the info I found it's listed as a "better performing substitute for 80% nickel" - how 'bout an auto-former VC based on this?

Cheers/
Pete
Too much stuff and not enough time!
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

Lots of speculation and WAG's follow.


Other than shine's friend, I don't think anyone has heard a direct comparison. I think the target is as you say 80% nickel, which suggests low level uses. It seems the cobalt amorphous also targets 80% nickel so that poor material sure has a lot to defend against.

The word "nano" is getting thrown around a lot lately, it seems to become more of a marketing tool than an engineering choice. There are two different categories of the nano materials, the Finemet and the nanoperm. Actually there is a cobalt based one too Hitperm that offers the same Bsat as GOSS, but 10X the perm.

I am only slowly coming to grasps with the differences between the metglas cores I use and the 49% nickel. I have done a fair amount of listening to the two and there is a distinct flavor to each, but when you look at their BH loops they look pretty much the same.

Essentially the nano cores start their life as an amorphous ribbon that is put though a anneal with a superimposed field to "orient it". If you look at the data sheets for the SA1 amorphous material it shows three different anneals with three different loops. I was told years ago that the SA1 (powerlite) cores were the no field anneal since they were unable to produce the "field anneals" with the required results outside the lab. This makes me suspect that at least one of the nano cores (finemet since metglass and hitachi metals are one in the same) is something very similar to the SA-1 cores with longitudinal field anneal. The fact that they offer the same line of sizes (amcc series) in Finemet as they do in Metglas strikes me as interesting. My previous discussions with the metglass engineer also mentioned that the field anneal increased in difficulty as the core size went up and became impossible if the core was cut. (this was 6 or so years ago) I asked for two quotes on the finemet, one for amcc-16A size and one for AMCC-160. The 16A was ~$150 a core and the 160 was not an option. They do have the ribbon slit in that size, so why was the 160 out of the question. Did they figure the $800 core cost would be too much, or was it beyond their limit of the field anneal? I suspect the later.

In doing the Lriaa chokes for sB the amorphous surprised me with its nonlinearity WRT signal level. I tried to get it under control, but by the time I got reasonable results, my gap was 10X that that was required on a comparable nickel core. Given that the perm vs. flux level tends to become less linear with orientation, it makes me wonder what "oriented amorphous" actually does for us.

I think pieter at tribute does his autoformers on one of the nano toroid cores so they might be worth a listen.

If I were forced to make a guess, I would expect the cobalt amorphous to give 80% nickel the best run for its money since it is the one material that appears to be very linear WRT flux level. BTW, minimally gapped (1X1 stacked) 80% nickel is horribly nonlinear in that respect, but a small gap really seems to clean that problem up.

dave
piotr_grzesik
Posts: 68
Joined: Mon Feb 07, 2005 9:59 pm
Location: UK

Post by piotr_grzesik »

Hi Dave,

Great info, thanks! Having looked up the Finemet info, it does seem that the published curves are very similar to the longditudinal anneal of SA1.

Hmm...the link I was looking at for Namglass was here: http://www.magmet.com/newmaterials.html and they seem to use the term Namglass interchangably for amorphous SA1 (Namglass-1) and the nano-crystalline stuff (Namglass-4). In fact, in this PDF http://www.magmet.com/nam/images/NAM_ca ... nedCh7.pdf it seems to be indicated that the Namglass is just a trade name of Magnetic metals for Finemet. So are they one and the same?

Sounds like a marketing exercise to me...or, am I missing something?

Care to comment on the sonic signatures of amorphous vs Nickel? Does each have a consistent signature regardless of winding geometry and is one distinctly superior to the other? Or, is it a case of pick your poison and use it as a tuning tool?

Cheers/
Pete
Too much stuff and not enough time!
shinebox
Posts: 88
Joined: Fri Jan 28, 2005 9:44 am

Post by shinebox »

Allo dave,
thanks for all the info, *very* interesting.

What with the exchange rate as it is, I don't puke at the thought of a $600 pair of cores. My sowter 80% nickels (for 300B etc) cost me nearly $400 but they sound absolutely wonderful. They also sing along like nuts, so I'm thinking let's try something else if it has a fightin chance of sounding as good (or better).

I'm confused now as to what exactly we are talking about; you mentioned the amorph cobalt; this is, i think, the same stuff I used on my little toroidal input transformers. Is this the one you are talking about for my opts? If it is, I'll gladly go for it and let you play with the cores for as long as you like.

I know the stuff my pal tried out was cobalt based amorphous, but am waiting to see if he has any more details.


Cheers
jarek
Posts: 62
Joined: Wed Feb 02, 2005 10:02 am

Post by jarek »

The word "nano" is getting thrown around a lot lately, it seems to become more of a marketing tool than an engineering choice.
B_sat at 1.1-1.2T, coerction as small as with supermalloy, can be custom tailored to very low B_r, giving nice flat narrow loop. the notion of initial and max perm loose sense as the perm is constatnt till saturation. enough for me. the fact that nano technology is a catchy word is another story, but nano mags have little to do with real nano engineering (nanorobots, etc.)
in the mag realm "cobalt" seems to be the word.
Essentially the nano cores start their life as an amorphous ribbon that is put though a anneal with a superimposed field to "orient it".
i strongly doubt it. both materials can be annealed in transverse or parallel (with respect to the tape direction) field to get either flat or square loop. amo is made in the process of ultrarapid quenching. there is realy no crystal structure, hence the name "metal glass". nano, as far as i know, is different - it is intermediate between amo and usuall crytaline metal - there are nano-sized (hence the name!) crystals embedded in amo surrounding. i recon to get such different structures one requires different metallurgical processes, not just annealing. look at the chemical composition of nano. they add a lot of stuff (B, Sn, etc) to catalyze/stabilize the formation of nanocrystals.
If I were forced to make a guess, I would expect the cobalt amorphous to give 80% nickel the best run for its money since it is the one material that appears to be very linear WRT flux level
any data/listening tests on that or just a guess? i assume we are talking about low level linearity.

jarek
piotr_grzesik
Posts: 68
Joined: Mon Feb 07, 2005 9:59 pm
Location: UK

Post by piotr_grzesik »

i strongly doubt it. both materials can be annealed in transverse or parallel (with respect to the tape direction) field to get either flat or square loop. amo is made in the process of ultrarapid quenching. there is realy no crystal structure, hence the name "metal glass". nano, as far as i know, is different - it is intermediate between amo and usuall crytaline metal - there are nano-sized (hence the name!) crystals embedded in amo surrounding. i recon to get such different structures one requires different metallurgical processes, not just annealing. look at the chemical composition of nano. they add a lot of stuff (B, Sn, etc) to catalyze/stabilize the formation of nanocrystals.
Well, the info on the Metglas website states http://www.metglas.com/downloads/finemet_magamp.pdf that the Finemet (nano-crystalline material) is indeed produced by annealing an amorphous alloy from an amorphous state, though not necessarily from SA1 because the composition appears to differ slightly. So the base material is slightly different, but I believe the process of annealing is what creates the nano-crystals.

Some info on annealing, specifially WRT Hitperm: http://neon.mems.cmu.edu/mchenry/mchenr ... ns/198.pdf

nanoperm appears to be another iron-based nano material, some info at: http://www.magnetec.de

The cobalt-based stuff seems to be something different again.
Too much stuff and not enough time!
jarek
Posts: 62
Joined: Wed Feb 02, 2005 10:02 am

Post by jarek »

Czesc Piotr,

thanks for the enlightning info (the paper). It indeed starts as amo and then is partially crystalized - they grow nano-crystals.

Best,
jarek
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

Yo' shines
I'm confused now as to what exactly we are talking about; you mentioned the amorph cobalt; this is, i think, the same stuff I used on my little toroidal input transformers.
Yes, the little toroids are cobalt and look very similar to 80% nickel
Is this the one you are talking about for my opts?
No, I was talking about the nano (of some sort) I have yet to find the cobalt amorph in a cut core, but It is rumored that tamura uses this.
I know the stuff my pal tried out was cobalt based amorphous, but am waiting to see if he has any more details.
Oops... That's right, My bad :-( the $600 cores are finemet.


And onto Jarek...
B_sat at 1.1-1.2T, coerction as small as with supermalloy, can be custom tailored to very low B_r, giving nice flat narrow loop.
This sounds like the cobalt amorphous, it also has a flat narrow loop, but it has a Bsat of .6T and its loops look very similar to minimally gapped 80% nickel.
the notion of initial and max perm loose sense as the perm is constatnt till saturation.
Which I think is a very good thing and why I always advocate for a gap. Maybe its a poor mans way to get the nano performance :-)
nano, as far as i know, is different - it is intermediate between amo and usuall crytaline metal - there are nano-sized (hence the name!) crystals embedded in amo surrounding.
But from memory, several different references state that all nano starts out as amorphous.
i recon to get such different structures one requires different metallurgical processes, not just annealing. look at the chemical composition of nano. they add a lot of stuff (B, Sn, etc) to catalyze/stabilize the formation of nanocrystals.
Very possible, but I still believe the starting point is "Some" sort of amorphous ribbon. Then the nano structure is somehow formed.
any data/listening tests on that or just a guess? i assume we are talking about low level linearity.
Nope just a gut feel, and a bunch of unrelated empirical results. I have yet to see a BH loop for any of the nano stuff, and that tells the story. Then we need initial magnetizatiopn curves all on the same scale and frequency and we might get somewhere.

My only real data point is the published loops for the cobalt amorphous look like my loops for gapped nickel... That and $4 will get you a cup of coffee.

and Finally Pete,

Nickel=Globe 45, warm lush and inviting.
Amorphous= ST 45, fast accurate and detailed.

i think it comes down to a flavor thing.
jarek
Posts: 62
Joined: Wed Feb 02, 2005 10:02 am

Post by jarek »

But from memory, several different references state that all nano starts out as amorphous
me arrogant asshole, mea culpa... they do start from amo and then anneal it with high field to grow crystals in da tape. apart from piotrek´s ref, here is one more:

http://www.wtec.org/loyola/nano/US.Review/05_05.htm

as far as I understand for a wide range finemet type is preferrable.

another bs which i tried to promote is the role of B, Si, etc - their role is to help in the glass phase creation, crytals you dont have to help - they normally grow on their own in metals...to compnesate for that I link some static magnetization curves of nano (dave initial mag curves are always dc!)

http://www.vacuumschmelze.de/dynamic//e ... nloads.php

download their catalogue of emc cores, there are some curves as well as other info, also in other catalogues.

jk
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

hey jarek

thanks for the pointers... i seem to remember having this conversatiopn before :-)

look at the attached image... any guesses how they vary the perm?? Do you think someone might already have a patent on the idea??

dave
Attachments
sure looks like they are adding an airgap to me
sure looks like they are adding an airgap to me
Picture-2.gif (31.73 KiB) Viewed 20799 times
jarek
Posts: 62
Joined: Wed Feb 02, 2005 10:02 am

Post by jarek »

i seem to remember having this conversatiopn before
:oops: i also got a feeling that i started repeating myself, so probably i should just shut up and learn more
any guesses how they vary the perm?? Do you think someone might already have a patent on the idea??
the only bit of info i got from the first hand (engineer who develops and produces amo and nano) was that for flat loop (i guess that would be mu=20000) they anneal in the tranverse field, while for square (mu=80000) in longitudal...pfe..could never pronouce this word properly. mu=30000 could be no field anneal? i guess that´s so "easy" as with amo. note how wavy mu=20000 seems to be...just a graphical imperfection? the interesting question is to compare low level linearity of CoFe amo and flat loop Fe nano.

jk
agent.5
Posts: 12
Joined: Sat Apr 08, 2006 3:50 am

Post by agent.5 »

Just wondering if this set of output transformers was developed?

I am thinking something like this

for each channel

DHT
IT (plate loaded)
1st or second order passive filters

IT (I have a bunch of 1:1 Altec 15335 octal plug can transformers). So I will probably use them here

to three, separate, spud(or direct couped) second stage and a subwoofer

spud #1: >5000 hz
spud #2: 800hz - 5000hz
spud #3: 100hz - 800hz
subwoofer plate amp: below 100hz

I have not decided which DHT to use for the first stage buffer and these second stage spuds. 801A is probably as good as any. But I can also use 2A3. Maybe 2A3 for just spud #1 and #2 and then PSE 2A3 for spud#3. 801A or 10Y for the pre-filter stage.

I known Gordon said that one need to use the same amp for everything. But I am not sure if he considers a regular 2A3 SET and a 2A3 PSE the "same."

I also have not decided whether parafeed for the output stage or not. One advantage is that the parafeed cap will allow a place for a high pass filter (maybe even second order by adding a choke after the output transformer in parallel with the speaker). But then I will need 6 additional plate chokes.

I probably need to draw out some simulations and see how this thing is going. But I am wondering if you have already thought of the problem about using different sized cores and related inductances and resistances of these transformers and can provide some data for simulation.
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

agent.5 wrote:Just wondering if this set of output transformers was developed?
on paper I have done a number of designs, but I haven't gotten beyond that... mainly since I am not multi-amping at the moment and shines is on the 20 year plan.

DHT
IT (plate loaded)
1st or second order passive filters

IT (I have a bunch of 1:1 Altec 15335 octal plug can transformers). So I will probably use them here

to three, separate, spud(or direct couped) second stage and a subwoofer

spud #1: >5000 hz
spud #2: 800hz - 5000hz
spud #3: 100hz - 800hz
subwoofer plate amp: below 100hz
I have been playing with a number of line level filters lately and your choice here will greatly influence the rest of the designs. I would be very cautious attempting to use a parafeed filter since the CL resonance can cause a number of headaches.

I like the spud amp idea, it keeps things simple for the output, but the driver now needs to deliver a lot more.

I probably need to draw out some simulations and see how this thing is going. But I am wondering if you have already thought of the problem about using different sized cores and related inductances and resistances of these transformers and can provide some data for simulation.
as I mentioned before I have done a number of different paper designs and will gladly entertain any conversations you might want to have.

dave
agent.5
Posts: 12
Joined: Sat Apr 08, 2006 3:50 am

Post by agent.5 »

on paper I have done a number of designs, but I haven't gotten beyond that... mainly since I am not multi-amping at the moment and shines is on the 20 year plan.
This is a long term project for me too. So, it may take a while for me to finalize anything.

I have been playing with a number of line level filters lately and your choice here will greatly influence the rest of the designs. I would be very cautious attempting to use a parafeed filter since the CL resonance can cause a number of headaches.
Agreed. Let's forget parafeed. Just use normal SE output transformers. Starting from the speakers, I say we use 2A3, and require a bunch of 2.5K output transformers, bandwidth limited. Maybe 2.5K:8+16 for the high spud and mid spud, and 2.5K: 4+8 for the lower mid/mid bass spud.

I like the spud amp idea, it keeps things simple for the output, but the driver now needs to deliver a lot more.
I am aware of that.

A 2A3 spud will need about 40V rms to the grid. A spud using one of those two tubes in one bottle duplex thingy may need about 10 - 20V rms to the grid. Whereas a couple volts are all is needed to an output tube with a driver stage.

So, this spud thing may or may not work. What advantages or disadvantages in building passive line filters (LR, or LCR) requiring higher voltage rather than lower? I suppose the chokes will be bigger and cost a bit more.

I see that 3 gain stages are required, although one of the gain stage can be a step-up transformer. We can have

[1] triode --> IT --> filter --> IT -->triode --> DC --> triode, or
[2] triode --> IT --> filter --> SUT --> triode, or
[3] triode --> DC --> triode --> IT --> filter --> IT --> triode

(IT = interstage transformer, SUT = step-up tranformer, DC = direct coupled)

I am not sure which way to go either. The filter is sandwiched between two transformers, because I am thinking that this will isolate all these spud amps, and that plate amp that does not use tubes, from each other. Is my idea of isolationism a viable or necessary idea?
Last edited by agent.5 on Mon Jul 28, 2008 1:05 am, edited 1 time in total.
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

I'm not sure i see the need for isolation and don't really see too much of an issue doing the filters at higher voltages. I do have a concern that the 2A3 grid may not be happy being driven through a filter.

I am also not a fan of step up transformers at high impedances (the only place i like them is after a MC cartridge)

I keep coming up with ideas, but keep hitting hurdles that are a bit too high to clear.

I'm sure there is a elegant solution for the general topology, we just need to find it.

dave
agent.5
Posts: 12
Joined: Sat Apr 08, 2006 3:50 am

Post by agent.5 »


I'm sure there is a elegant solution for the general topology, we just need to find it.
Just give it some time. I am sure somthing will come up. The two amps that I found utilizing this topology used way too many coupling capacitors.
Post Reply