Ampeg Gemini 1 OPT 7591A Push Pull

Dylan has successfully sent the Ampeg output transformer for us to enjoy and ponder,

a moment of silence for the transformers still suffering followed by the transformer prayer.......

oh Lord, grant me the power to dissect this transformer to the best of my ability,
the courage to un wind it as we see fit,
and the wisdom to know how to fix it,

ok then.

step 1  admit to ourselves that we need to post a few pics>

simple coil structure, pri-sec,

82 turns , 3 layers sec,

going to unwind sec tonight,

s/b close to 2355 turns according to Z ratio,

de-lam was fairly easy, not too heavy on the varnish, had to save the lams for rebuild,

heat gun and swiss army knife are the standard tools for this job,

first we take off the U frame,

then the grunt work,  :-X

photo doc the lead breakout to back up the hard copy,
first layer exposed, sec winding,

everybody is taped to the same side, staples into the insl type deal,

sec winds are taken down,

starting to smell like a bar b que somewhere, 

pri is cooked from the inside out, including the core tube, or former, whatever you want to call it,

all the heat is on the inside layers,

not shorted turns, but shorted layers in this case,

they used #36 AWG, which is good for 25 mils, should have used #35 which is good for 31.6 mils, 7591A push-pull needs 70 mils, after the juice enters the red lead, we can split this plate  current into two branches that need to carry 35 mils each, get me?
you might get a little leakage C added from the bigger wire, but at least the house will not burn down,
like a fine tuned sports car, more turns per layer from the smaller wire, gets that core up and running, but this OPT is gonna run hot, all part of the sound, core gets hot and looses perm, inductance goes down, tone changes,

you could run this on a 100 EI and use bigger wire, hi end changes from bigger wire and longer turn length, but you can play tricks with the winding structure to compensate,

looks like the inner pri shorted to the core which is grounded, notice the cooked wind has the 110 ohms DCR in the first print we did,  400 volts over 100 ohms = 4 amps current,  enough watts to fry half a hot dog, 

nylon bobbins resist flame-out better, so we can do that,

dumping some residual images on the LAB server,

pri turns were 1,115 outer, inner turns were baked, so they did not unwind well, so we double the 1,115 and get 2,230 turns of #36 which will go on the final print to be done A as i am cooked like the pri, 

zebra lams, ready to rewind on a 7/8 x 1 bobbin,

weird phase inverter, triode on the bottom and something else on the top,

Can't see on the photo, but lots of Ampegs had "paraphase" phase inverters. They sound great in my opinion, including when overdriven (altough it's problably not an unanimous opinion).

http://www.valvewizard.co.uk/floatingparaphase.html

Pretty sure the triode section of 7199 is the cathodyne phase inverter (tip off is 47K resistor on both plate and cathode).

Pentode section of 7199 is just a gain stage feeding into the triode PI.

Difficult to see because the schematic is weirdly drawn.

http://www.valvewizard.co.uk/cathodyne.html

That primary really got it 

Are zebra lams different than the usual GOSS?

don't know if the old lams are much different, will run inductance and freq tests on the rewind,

did a test on the #36 wire , 800 turn test winding went all the way up to 350 ma, it got warm but did break, so we will rewind with the original wire size,

going to use the old sec wire as it looks fine, this will keep the DCR  the same as wire was a bit different back then as far as copper content,

rewind complete!  here are some pics, was gonna space the turns for cooling but after winding a test coil of 700 turns and finding that the #36 wire did not break even with 370 ma running thru it, (got a lil warm) then we decided to lay it on down normal like,

re-stacked the lams 1 x 1 , they all fit as not too much varnish on them,

has a new fire proof bobbin and a Nomex cover, Nomex is good for 1 KV per mil, so this 10 mil cover is good for 10 KV from coil to core,

doing Pri.  inductance test, getting 381 Henries at 5 Hz so that means we did not screw up the lam grain structure during removal,

376 ohms DCR,  DCR/X-L @ 5 Hz =0.0314 Arctan = 1.80 degrees of phase shift,

have a Kingsoft spreadsheet program so all you do is input voltage and current and it does the rest, yes that is a VP opamp and a DIY T4b in the pic,

here are the pri Henries>

runs pretty flat, rez peak was at 300 K Hz,

when you have just a few turns on the sec, you do not need much coil structure to run flat, a 600 ohm output wind will need to be sectioned to run this flat, all about the leakage,

+1 on the "Cathodyne" PI. Giveaway is that Rp and Rk are equal. This was apparently (SWAG) an RCA app sheet at some point bc you see the 7199 used like this across multiple mfrs (Dyna ST-70, Ampeg, Eico too IIRC) with the pentode section going for voltage gain driving the PI.

Have to admit I have never seen these kinds of laminations before. More common IME are various annealing marks.

nice work on the rewind! Great write up.   The 7591 amps are some of the best sounding ever made IMO.  Did you make your own coil winder?

yes, made the coil winder,