Peerless S-217-D Pultec EQP1a OPT Redesign

biggest bellyache around here is this Pultec OPT,
everybody wants to know what to do about this transformer situation, huge price tags on non-existing ebay auctions, lack of original laminations, feedback winding standing in the way of a suitable cross ref,
re-issues being expensive due to overseas shipping costs and exchange rate variance,

let's see of we can come up with a solution that will end this ongoing humiliation, 

an original 217 uses two coils on an L lam which is no longer available. L lams are hard to get even in the wrong size, plus, nickel lams are expensive, especially when they want you to buy 3 tons at a time,

nickel EI lams are semi easy to get, can we wind a dual coil job on an EI?  of course, just split the bobbin,

keep the same turns, the same wire, same winding structure, find the flux density of the original and size the EI core accordingly,

original turns and wire go like this:

Pri-1575T #36
FB-71T #36
Sec-352T #29

ratio is therefore 1575/352=4.47  F/B ratio is 1575/71=22.18:1

why the weird 4.47 number, where do they get that?  remember that these engineers want to get from impedance to impedance, not voltage to voltage like a power transformer,

what is the impedance ratio, the square of the turns?  yes. so 4.47^2=20

so this engineer, Ersel B Harrison,  wants to go from 500/600 to push pull plates of a 12AU7,

so 20 times 500=10K,      20 times 600= 12K,  what does the catalog say?

link>

http://www.bunkerofdoom.com/xfm/PEERLESS1953/Peerless_1953_08_0001.jpg

pic>

ok so they use 600 ohms and say 12.5K  Harrison is cheating already, puffed up 12K to 12.5K, naughty man,

level? 20 db, Freq Response? flat from dc to daylight,

remember these turns?

Pri-1575T #36
FB-71T #36
Sec-352T #29

that is for one coil, if we do flux calcs, we need the whole enchilada,  flux is the same for both pri and sec, but the business end of this thing is the pri since it has to work with the 12AU7a plate swing,

so we calculate pri flux,

we need lam dims, plate AC and turns and freq. everybody likes 20Hz so we have one constant,

turns = 2 x 1575 = 3150T, so we have two constants, now we have to hunt down the lam dims,

just so happens we took apart a Peerless with 12L lams so those dims are in a file somewhere,

yes, there it was, in the Peerless folder,

so we have 3150T, 20Hz, and a core cross section of 0.907 cm^2 but wait,

how big is the stack?

look at this, a top secret document, Winston O Boogie, partner in crime,  ;D

where is the stack ht?  check the former, 3/8 x 7/8 bingo,

so 7/8" = .875,  3/8"=.375, stack multiplier is .875/.375=2.33

so 2.33 times our sq stack area is our total cross sectional area for the 217,

so 0.907 x 2.33=2.116 cm^2,

cross check: 3/8 x 7/8=0.328 in^2  convert to metric, 2.54 x 2.54 is sq conversion factor= 6.4516

6.4516 x .328 = 2.116 check.

so we have turns and wire and structure, need to find EI core with a cross section of 2.116 cm^2

first let's see where they are running this core at,

Bmax=100,000,000E/4.44K1NAf
K1 stacking is 0.9 for a 14 mil lam laced 1 x1 

B=100,000,000E/4.44x.9x3150Tx2.116cm^2x20Hz

B=187E

rumor has it that Harrison, for reasons unknown, used an 80Ni core for an output transformer,

this has a saturation level of about 5 K Gauss. so how much voltage can this core handle on the pri wind?

5000=187E  so E=5000/187= 26.73 Vrms

lets get our sec volts so we ca use a dB calculator and see how honest that 20dB rating is,

26.73/4.47=5.9798 Vrms sec volts into a 600 ohm load,

puffed up another rating, says 20, actual is about 17.75

maybe their alloy can handle more than 5KG,

we can back track to see what there core would have to be like,

20dB = 7.75 Vrms,

flux is 187E, so gauss at 20db is 4.47 x 7.75 = 34.64 Vrms primary

34.64 x 187 =6.5KG, so maybe they have a 78 Ni or some custom blend that runs at a saturation flux of 6.5 K Gauss. or they pulled some advertising BS,

so we need an EI stack with a cross sectional area of 2.116 cm^2,

let's take a look in the catalog and see what's up,

well of course, you could use a 375 EI which has a 3/8 tongue like the 12L, and build a 7/8" stack,
this would keep you former aspect ratio the same as the original 217, or you could use 50EI and use a 1.3 aspect ratio, or a sq stack of 625EI,

you could also use a 75EI Ni lam and use less than a sq stack, this would give you a winding area closer to the L lams as far as the width of each layer, this 25EI will require more layers with fewer turns per layer, or we could use smaller wire which would lower capacitance, you do not really need #36 for a 12AU7,

what about a C core? an permalloy C core would cost a fortune, steel won't sound the same,

while searching ebay for Ni core i happened across this Tamura OPT with 38% Ni,

http://www.ebay.com/itm/Pair-Tamura-F-7001-output-transformer-permalloy-core-for-SE-tube-amp-Ex-RARE-/131584461418?hash=item1ea30b8a6a:g:a5QAAOSwgQ9V1bzA

so the 217 is not the only OPT with weird steel,

ok the biggest Ni lam i can find that is easy to obtain and won't break the bank is here>

http://www.ebay.com/itm/Nickel-Permalloy-EI-35-Core-and-Bobbin-for-Transformer/230646735924?_trksid=p2045573.c100033.m2042&_trkparms=aid%3D111001%26algo%3DREC.SEED%26ao%3D1%26asc%3D20131017132637%26meid%3D5294ab1ef76f4024b88c88c22f9a8307%26pid%3D100033%26rk%3D1%26rkt%3D4%26mehot%3Dpp%26sd%3D230646735924


it is a metric lam, so we convert it to English, tongue is 9.6 mm = .96 cm = 0.2438" or about a 25EI

we are going to use both sq stacks so this core is 27 bucks, but a small price to pay considering what the original goes for,

remember that we have an aspect ratio of 7/8  / 3/8 = 2.3 

this ratio can change the sound, so we will have a 2 to 1 ratio with a twice stack of 25 EI,

how does our area stack up to the orig?

well, a sq stack of 25EI is only .403 cm^2, so we will have a .806 cm^2 cross area compared to 2.16, but wait, this alloy that we bought is closer to 50Ni, so we do not need as much area to handle the flux, and 25EI weighs in at about .37 lbs per half stack so we will have a 3/4 lb core compared to the 1/2 lb weight of the original core,

and, looking at the inductance constants for both the EI25 and the 108L which is close to the 12L, they both come in at about .1^-8, so it looks like the L lam can with twice the stack height of the EI core will have the same inductance constant as the EI core, so Henries is a function of weight, not cross sectional area, interesting, no wonder they use such a big stack of 12L lams,

now we will have a lower perm with the 50Ni, but the extra 1/4 lb so make up for it,

free shipping from Korea, have to wait a bit but we can get bust winding the coil,

this core comes with bobbins, good deal as it is a metric core, we can cut the bobbins and make one big one out of the two small ones, so we will have a twice stack former,

big shipping window, but we get stuff from China in about a week which is amazing considering it is free shipping, so stay tuned...

here is what the coil will look like>

Good stuff sir.

My ear is on the station.

..tuned in...

Damn, 

Good job. I know Cinemag Already made a recreation of the 217 (they had to get some special tooling to make them) but, as you said, it's somewhat expensive.

I'll stay tuned

I'm so listening 

ok, just off the plane from S Korea, a twice stack of 375 EI and a pair of bobbins that we have cut on the band saw plus a bird's eye maple block for the mandrel,

so judging by the last post this took about two weeks for delivery, talking with the vendor we have found that he is willing to drop the price (27 bucks for this stack)for quantity 100 which is potentially good news,

so we will spool up some wire and test this thing out,

spinning #37 twice, had to chop turns from 1575 down to 1400 to get it to fit on this lam size as our winding window is smaller than the 12L lams,

but this 375 EI has a inductance coef of .156 compared to .105 for the 108 L which is close to the 12 L dimension wise, so we can get more inductance with less turns,

have to watch the flux, when you drop turns your dBu comes down,

but we show that with 1400 x 2 = 2800 turns, we can ram 60 volts down this thing as it is 49 Ni, good to 15 KG. our turns ratio is 2800/622=4.5:1 so this works out to 13.3 volts on the sec = about 24.5 dBu at 20 Hz, which beats the stock 20 dBu on the Peerless, we could drop the turns down even more to lower DCR and match up with the stock transformer better as far as saturation level, but then we lose Henries, i would rather have extra Henries for that low down freq that the Pultec is known for,

#36 wire is stock, but took too much room, #37 is ok current wise and with the lower turns the DCR is not a factor,

all done winding, freq response looks good to 100 K, big 6 times rez peak at 300 K, then drops but still puts out at 600 K, this is leakage C, like the API 2503 OPT,

big test is F/B wind phase shift, the hardest thing to get right, if it is off, the amp will freak out,

inductance looks good, perm works out to 50K,

lets work out Sec Henries from Pri Henries, turns is 4.5. Z is then 4.5^2 = call it 20,

1200 H peak L / 20 = 60 Henries Sec Ind, which is a little above most 600 ohm winds, up there near the Langevin AM 16 C core stuff, or the K-241-D,

here is a 20 Hz level check, this curve will have the same shape as the BH curve, at the top at least,  increase in V-rms means more current which means more amp-turns which means more Force  which means more Oersteds (H) , more inductance means more Flux gathered by the core which means more B,

after the F/B wind check we will plop this thing inside a Pultec and see where the smoke comes out,

phase looks ok for the F/B wind, zero shift til 60 K where we have about 30 degrees, so we should have a stable line  amp, 

might not need the whole 680 ohm/.003 cap on the output, that is some pretty brutal damping there, 

here is pri input and F/B wind on the scope at 60 K Hz>

first proto works great but is labor intensive with the split bobbin, can we get away with one chamber?

so we rewind and use 2500 turns instead of 2800 since we had tons of inductance and plenty of headroom, with 2500 we can use bigger wire for the sec like the original, bigger wire means more capacitance, but turns are reduced from 352 twice to 279 twice, (558T) so C will drop a bit,

how much did inductance drop at 20Hz?

it went from 992 H with 2800 to 862 with 2500, still reasonable, and we get to use #32 instead of #33 for the sec, which drops DCR down from 40 to 28,

freq response suffers with the single chamber, down from 100K to 60 K, rez peak comes down from 300K to 120K, like a normal transformer, will check feedback wind phase today,

if we kill the response a bit, that might prevent osc. in the preamp as by the time the phase hits 180 out on the F/B wind, the voltage will be close to zero.

with less turns we find the sec saturate at 6.5 V-rms - 18.5 dBu,  finding out that KG-sat is 8,000 instead of the advertised 15KG, wtf, over?  still, a respectable level, 29 volts on the primary side,

so we decide on coil geometry after the phase test,

Actually , David buy a tooling and original type of lams is used on Cinemag recreation.One of the first come to me, and it become obvious that sound is a different from the solutions available on the market.More focus is the word that first come to mind.

looks like a CM-217-D wonder how much?

ok this single chamber bobbin version goes to 60K with no phase shift on the F/B wind, not flat out to 100K like the dual chamber, but who cares? would rather have a more stable amp with less response, right?

listening test next,

sasarist said:

Actually , David buy a tooling and original type of lams is used on Cinemag recreation.One of the first come to me, and it become obvious that sound is a different from the solutions available on the market.More focus is the word that first come to mind.

Click to expand...

That's the one I was talking about

Great work as usual CJ!
I wanted to ask, are you using any insulation between layers? Between windings?
It's 250V DC or whatever on that thing!

yes at least some poly tape to insulate layers, wire is so good now days, that arcing out is no longer a big concern, but the distance between layers can influence the response as the capacitance changes,

we stuck this thing in the Pultec and the vu meters on the mixer went up halfway but no sound came out, this is because the amp is oscillating above the hearing range, so the 180 phase flip at 120 K Hz was enough to cause problems,

so we have a few tricks up our sleeve to fix this, one is change the coil geometry, the other is to add cu foil screens, we will try the screens first, if that does not do it, then we will change the geometry and add screens, 

we have some pretty hideous ringing going on, the screens will damp this out also, no wonder the 217 has that big-azz zobel on the sec,

if we are lucky, we will have a better 217 than the original,

here is the ringing, in vs out>