Transformer Secondary Hookup

[CJ]

fooling around with different hookups on this 100 EI xfmr secondary,

we have a coil wound Pri-Sec-Pri-Sec-Pri-Sec-Pri 

what happens to inductance when you only use the inner secondary versus all three secondaries in parallel, or only two secondaries in parallel?

here is a graph showing the variation in Henries due to different hookups,

the X denoted graphs are just the original graphs with the DCR subtracted from the inductance calculation,

since DCR is higher with just one secondary, the difference between the Henries calculation is greater,

we applied 1 volt to the secondary from 5 to 40 cycles per second, and then measured current, divided volts by current to get Z ohms, then corrected for DCR which gives us X-L ohms,

Inductance was calculated by dividing Z and X-L by 2 pi f

the red graph is just two sec's in parallel, we used sec 1 and 2, (inner and mddle)

what happens when we use inner and outer sec's only? see next post...

 

[CJ]

here are the graphs for 2 sec's in parallel, inner and middle vs inner and outer,

looks like inner and outer work better than inner and middle,

a Fisher OPT that i "borrowed" the winding topology from uses inner and outer sec's in series with a bi-fi sec in the middle,

so maybe we will try adding a bi-fi sec for the middle winding and see what happens,

so now you might ask  "why bother using parallel sec's if using only  the inner sec gives more inductance..." (see graph in first post)

two things, frequency response and phase shift at lower freqs,

both are improved when you use parallel sec's as we will show,

note that on the following graph, DCR starts to creep in as you go lower in frequency,  this is because 2 pi f L gets smaller as the frequency gets smaller,

so Z and Reactance are getting smaller, but DCR is constant, so X-L approaches DCR as you go lower, which is why engineers try to keep DCR low, distortion gets better with less DCR also, especially at the lower freqs,

note that DCR is subtrated quadratically from Z ohms,

so X-L = the square root of Z>(V-ac/I-ac)^2 - DCR^2

 

[CJ]

here is an good article on phase shift at the Magnequest site>

http://magnequest.com/tech2.htm

to get degrees phase shift, simply divide DCR by Reactance and take the arctan,

example:  a 2.25 Henry sec. winding is found to give 70 ohms of Inductive Reactance at a frequency of 5 Hz,

DCR from the 460 turns of #34 wire in the coil winding is 18 ohms,

Phase Shift = 18 Ohms DCR / 70 Ohms Reactance = 0.257, 

arctan of 0.257 = .252 radians or 14.4 degrees at 5 Hz.

this graph shows phase shift for 3 secondaries,

1 sec, 2 in parallel (inner and outer) and all 3 in parallel,

since DCR is lower with 3 sec's  in parallel, phase shift is lower,

a single sec gives a bit more inductance, but has 3 times the DCR = more shift,

Telefunken V72 and V76 amps use 3 sec's in parallel in the outputs,

bottom line is that DCR hurts the low end, what about the high end?

well, more DCR for the same amount of turns means smaller wire, which means less capacitance, which means better high end, so we balance the tradeoffs,


tomorrow we will check out series secondaries, f response and shift,

 

[CJ]

here is a whacked out graph showing the high end response from 20 K to 300 K of  three individual secondaries of equal turns,

Sec 1 is nearest the center of the core, Sec 2 in the middle and Sec 3 on the outside,

Structure is P-S-P-S-P-S-P, 1 2:1 ratio, 5600 T : 460 T,

variations are due the capacitance being different for each coil section,

next we will plot combinations of 2 secondaries, then 3,

 

[CJ]

bed of nails for easy hookup changes,

and a fresh roll of NOS Kester 44 because nothing has that smell anymore,

reminds me of wiring up a Heathkit, 

25 bucks off evilbay, free shipping of course, use to be 4 bucks a roll,  ???

unleaded is even worse,

 

[rafafredd]

Interesting. I didn't know inductance would fall so much with parallel connection of the same turns windings.

M6 lams?

 

[emrr]

Very interesting stuff CJ.

I'll add a tangent, more response observations, but L and C have to come into it.  Has to do with unused windings, in a way.  I've seen a few UTC A-10 type primaries that had an open on one side, between mid tap and center connection.  So, you can't get a 150-200 ohm humbucking connection that way you are supposed to (2-5), let's hook up to the other full winding (3-6).  Pretty close to the ratio you want, but response is different, with a big top end resonance peak.  I found you can resistively load the open winding (1-4), and get the resonance peak down.    So we have no DC connection in the winding , but the addition of R load tames the strays blowing up the response. 

Gates built an amp for awhile that sometimes had an ADC input designed for SE sec use, wired full winding.  Sometimes they subbed an ADC input meant for PP use, same ratio, but only used half the sec winding.  Left the other sec half unwired.  I tried wiring both windings in series, both in parallel, response was terrible, still best with the one side fully floating.  Still had a resonance peak the proper SE type didn't have, but better than the options you'd intuitively go for.  At that time I didn't think to try loading the floating winding, but I'm sure that would have improved things further. 

 

[CJ]

yes those hybrid transformers do some weird loading that i am still trying to understand,

 

[CJ]

here is a graph showing hi freq response of 2 out of 3 sec's being used in series,

different combinations were used, 1 and 2, 1 and 3, and 2 and 3.

looks like 1 and 3 win as far as flatness, stability,

the meter really jumps when you measure an unstable winding,

next will be 3 sec's in series using different combinations,

 

[CJ]

high frequency response for 3 secondaries connected in series,



note to Rafafredd, the early graphs did not start at zero on then Y axis, so differences in inductance are magnified a bit,


looks like using the outside sec first leads to the most instability,

next we will try parallel sec's in the hi freq band,

 

[CJ]

forgot some hookup combos for 2 sec's in series,

thought 1-2 was the same as 2-1 as with AC, order does not matter,

however, after further inspection, we see that 1-2 is not the same as 2-1,

if you look at the start-finish orientation, you see that we have

S1 F1--S2 F2  and S2 F2--S1 F1 

so we have S1>F2 and S2>F1,

so capacitance will be different from 1-2  and 2-1 if phase is kept the same,

so here are all 6 combos along with the schematic,

looks like using the outside wind first is the most unstable,

next up will be parallel connections and their hi freq response,

 

[CJ]

here is a graph for using 2 out of 3 secondaries in parallel,

1//2  1//3  and 2//3

next up is all 3 in parallel,

 

[CJ]

here is the graph for all 3 secondaries in series along with the best graphs for single and dual secondaries in parallel,

looks like there are flatter graphs up to 70 K, but after that, 3 sec's in parallel is much more stable,

there is a vintage transformer that is famous for blowing up tweeters, too much hi-freq energy that causes oscillation in the amplifier, so graphs like these can help predict problems.

we can also fool around with screens and reverse winding in the secondary as well as different winding topologies, too much fun these transformers,

 

[CJ]

now we are testing an 80 watt OPT secondary.

this is a quad-filar wound affair done on a EI 175 lam,

30 turns each coil,

DCR is per each coil as the winding progresses outward from the core,

we are leaving all the wires loose so we can check to see how things vary from wind to wind like inductance and frequency response,

inductance does not change by any great amount from coil to coil,

148 milli-Henries ea at 5 hz, and about 143 mH at 20 Hz,

voltage level was low as the signal generator was having a hard time with the low inductance and DCR, so test level was about 700 milli-volts AC,

next we will check freq response from coil to coil and see what happens,

here is a pic of the xfmr along with the schematic of the secondary,

why all the wires? to keep C down so we can get 10-100 K Hz +/- 1 db,

pretty tough to do with a coil of this size,

 

[CJ]

nice to have different colors to keep track of things.

red grn and natural is used on orig, no natural so just make up some aux spools of red and grn,

hack saw there in case of any problems,

about 8 types of insl of various thicknesses used to get the freq response right,

so a highly engineered coil wound P-S-P-S-P-S-P-S-P or 5-4

 

[CJ]

here is a graph showing how hi freq response varies with placement of sec windings,

same 80 watt OPT done on a EI 175 core, so it is a big mutha,

this graph is for the top section of a 64 ohm winding which uses 4 sections,

so response varies as each wind gets further from the center of the core,

not much difference was seen at lower freqs so the graph starts at 20 K Hz

 

[CJ]

here is a graph for the same xfmr showing a chunk of the 8 section windings,

sec structure can bee seen a few posts up,

this chunk is from leads 10 to G, which constitutes half of the 4 ohm winding,

response did not vary much from one quad wire to the next in each layer,

next we will combine all the winds in parallel as they are used in the Williamson amp and see if the curves cancel into a nice flat response out to 60 K and maybe beyond,

 

[ricardo]

http://www.r-type.org/articles/art-140.htm Ultra Linear Output Transformers - Leakey & Gilson

An article which might explain some of the design aspects of CJ's transformer.  Great Guru Baxandall deferred to Messrs. Leakey & Gilson when designing a transformer for his own designs.  Gilson were the transformer of choice for many famous valve designs eg the Mullard 5-20

It's about UL transformers but also discusses good practice for triode (eg Williamson) and pentode devices.

CJ, do you know if your transformer was designed for the Williamson?

 

[CJ]

thanks for the link Ricardo!

yes the 16431  was designed for the Williamson amp, this one for the Griefkit W-6M,

here is something from the Magnequest guy>

"Posted by mqracing (M) on January 22, 2008 at 05:41:31
This is a pair of outputs as built and used on the Heath W6M amplifiers.
Check out the specs for these. Quite good.

Peerless actually designed three different transformers for the W6M amplifier. The 16431 was the unit that Heath s elected for production.
But actually I like either of the two designs they didn't pick better
than the 16431.

Apparently--- Heath picked the 16431 because it had the 70 volt secondary tap which opened up some sales prospects for the amp. But the two other designs in the archives are IMO the better designs sans the 70 volt secondary tap. All three of the W6M tranneys share the design logic and coil geometry of the 20-20 Plus series that Peerless introduced in the early fifties. All three of the designs used all M6 EI based cores.

Interestingly the difference btwn the other two--- which had the same primary impedance, power rating, and response characteristics as the 16431--- was in how the secondaries were wound. One of the alternatives had bifilar secondary windings while the second alternate design had trifilar secondary windings. In the past several I've built several of the W6M versions which used the bifilar wound secondaries."

Western Electric had a transformer called the 197a which had some similarities to the Peerless,

cj

 

[CJ]

here is another graph of the 16431 Secondary,

it shows the hi freq response after combining four coils of the 64 ohm section,

there are 4 sets of four coils stacked vertically, but each of these sets look about the same so this graph shows a four 30 turn coils in parallel,

as expected, the curve flattened out quite a bit,

note that the graph starts at 20 K Hz,