Cloning 2503 Api transformer

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

Graph

Well-known member
Joined
Jun 13, 2009
Messages
121
Location
Argentina
Hi, i am wandering about, i made some custom DOA preamps, and bought some 2503 output transformers. Preamplifiers run smoothly with some different opamps and transformer(from cinemag, cmmi2c and cmmi 75101apc). Really nice results, but i was wondering sending here to clone 2503 opamps, but because of import costs to argentina, if i need 10 or 20 units, it will be difficult to get them.

So, i was planning on cloning them, when an idea came to my mind. Is it possible to clone it with different laminations, like 80% nickel?(i know it will be a little expensive through) Will it help providing different harmonic distortion? Is it worth?

Thanks :)
 
If I remember correctly it use M6. you should be able to get that iron in argentina.
sadly we have just M19 here in Chile.

edcor is selling M6 laminations too
 
2503 transformer, not opamp?

i have an API rack with a Melcor on the left and a Melcor on the right,

a 29M6 steel xfmr on the left and a 80 Ni xfmr on the right,

they both sound the same, i can not tell the difference,

so save yourself some money and go for the stock core, API engineers were sharper than most, you can trust their work,

since the steel core version is good from 10hz to 1,000,000 hz, using a nickel core with fewer turns will give you 10 hz to 1,000,000hz.
 
CJ said:
2503 transformer, not opamp?

i have an API rack with a Melcor on the left and a Melcor on the right,

a 29M6 steel xfmr on the left and a 80 Ni xfmr on the right,

they both sound the same, i can not tell the difference,

so save yourself some money and go for the stock core, API engineers were sharper than most, you can trust their work,

since the steel core version is good from 10hz to 1,000,000 hz, using a nickel core with fewer turns will give you 10 hz to 1,000,000hz.
Ok! Well thanks boths, i will use M6 then.

If i dot get them here, i will just get some from edcor.

Thanks again!
 
you could also use a core from a defective or garage sale transformer if they used old school varnish, just cut the wires off and soak it in a strong solvent for a few days,

then wash it with something and then heat it up with a heat gun or put it on the Bar B Q next Sunday.  you should be able to pull it apart piece by piece with a pair on needle nose or small vice grips,

steel lams will take a lot more bending abuse than nickel, so just flatten them out if you bend them. this is especially true for "barn roof" lams, that is, hot rolled lams with no grain orientation.
 
I've made a couple of transformers with nice results, I get some lams from a factory in Avellaneda, Gran Buenos Aires, Zona Sur. Tendría que buscar los datos si te interesa..

Here is the freq response of my first one
freq.jpg


HF strange curve is because of the sound board but still 0.1dB from 20Hz to 20kHz, second one was some better but don't have the measuraments here now... I just put as much 0.30mm wire as I could in a 62 Lam and it worked nice, I used trifilar winding keeping 3 single wires together by the way.

JS
 
joaquins said:
I used trifilar winding keeping 3 single wires together by the way.
You must be aware that multifilar winding results in very high capacitive coupling, which gives these transformer astounding HF response. But You may notice a significant alteration of HF response when you unbalance the output.
See attached. Balanced in red, unbalanced in green. This is a simulation but it matches actual measurements. The xfmr model is the result of actual measurements.
 

Attachments

  • bifilar xfmr freq resp.jpg
    bifilar xfmr freq resp.jpg
    68.2 KB · Views: 81
This is the model of an actual xfmr, which, BTW, uses M6 lams. The inductance is actually quite low, because this xfmr is a quadfilar type designed to be embedded in an active circuit.
 

Attachments

  • bifilar sim.jpg
    bifilar sim.jpg
    25.8 KB · Views: 81
stock 2503 s/b good to 1 mHz...

Grossner says that matching the load to the xfmr Z is important,

if R not equal to Z, you get reflections on the high end, so you might want to play around with that on the simulator, this is transmission line technology \,
 
CJ said:
Grossner says that matching the load to the xfmr Z is important,
That is quite surprizing, since there is nothing such as the intrinsic impedance of a xfmr. The rated impedance of a xfmr is just a recommandation based on its intended use. A xfmr rated at 600:600 in the context of pro audio use (inferring ca. 20Hz LF cut-off) could be rated at 3k:3k if the intended application limits LF response at 100 Hz.
Response at supersonic frequencies is governed essentially by leakage inductance and winding capacitance. Construction (sandwich, chamber...) is only a means to achieve low leakage and low capacitance.
if R not equal to Z, you get reflections on the high end, so you might want to play around with that on the simulator, this is transmission line technology \,
Even in the MHz range, the effects of leakage, capacitance, DCr and source Z will be dominant over the SWR effects.
 
I don't remember now exactly but L was around 150mH or 200mH and capacitance between windings around 4nF, lekage inductance and capacitance I can't remember, DCR was around 5Ω or 6Ω, driven from a 300Ω source -3dB was around 8Hz, with the NF circuit it was for the freq response was the plot I posted.

I couldn't measure over audio freq with my computer and the software I was using, even at 192kHz of my sound board can't get over 95kHz... I don't have much more than that at home... all those measurements were made almost a year ago and I don't have now all the documentation so some data came out of my mind.

JS
 
HF response is fairly complex, about 30 pages worth of formulas, we use models to try and get close to reality but they are only approximations to the actual values, experiments save a lot of math work,


you need a good scope and leads to check from 100k to 1 meg,

measure current and phase shift if you want to understand these coils a bit more,

stuff:


 

Attachments

  • z1.jpg
    z1.jpg
    260.4 KB · Views: 50
joaquins said:
I don't remember now exactly but L was around 150mH or 200mH and capacitance between windings around 4nF, lekage inductance and capacitance I can't remember, DCR was around 5Ω or 6Ω,
These values seem extremely low. Typically, you're looking at 5-20H for inductance and 20-40r for DCR. 4nF for capacitance seems credible.
driven from a 300Ω source -3dB was around 8Hz,
Which suggests L=about 6H.
 
wide band, not broadband,  ;)

but yes, i was miffed a bit too as to the use of Mr.Grossner's "wide Band" definition,

however:

"wide-band transformer:
A transformer that can transfer electric energy from one circuit to another at any of a broad range of frequencies."

McGraw-Hill Dictionary of Scientific & Technical Terms, 6E, Copyright © 2003 by The McGraw-Hill Companies, Inc.

there is no RF XFMR section, but there is pulse transformer section, now if you want to really get into some heavy math, try pulse transformers, it is so bad that engineers are forced to use lab experiments instead of formulas,

remember that even if the 2503 is good to 1 meg, that you are still limited by the input transformer, i do not know the bandwidth of the opamp itself, but would imagine that it might have limitations also,
 

Attachments

  • wideband.jpg
    wideband.jpg
    148.8 KB · Views: 20
abbey road d enfer said:
joaquins said:
I don't remember now exactly but L was around 150mH or 200mH and capacitance between windings around 4nF, lekage inductance and capacitance I can't remember, DCR was around 5Ω or 6Ω,
These values seem extremely low. Typically, you're looking at 5-20H for inductance and 20-40r for DCR. 4nF for capacitance seems credible.
It was much worse before I get some grain oriented lams, of course, but I did a second one with some more turns it didn't get much higher inductance... Maybe I should try thiner wire but it will increase DCR and LF corner won't get lower...

driven from a 300Ω source -3dB was around 8Hz,
Which suggests L=about 6H.
I mis write here, from one measure to the other I change a 33Ω for a cap to block DC, some hundreds µF... So 33Ω, not 300, this would be around 600mH, wich would be a little higher than my measurement with my tester at, I guess, at 1kHz, lower voltage, but may be the diference in freq and level what make the diference.

Neve 1166 has around 300mH in each pri with about 300 turns each, and said they are for 50Ω with pri parraleled, so I'm higher than that (around 600mH we said) should be fine to be driven from a discrete opamp like 2520 or 990 I guess.

JS
 
try twisting the wire until you get about 2.5 turns per inch,

hi freq response is good with the 2503, but as you go up in frequency, current draw becomes pretty hiddy compared to a non helical wound xfmr,
 
CJ said:
try twisting the wire until you get about 2.5 turns per inch,

hi freq response is good with the 2503, but as you go up in frequency, current draw becomes pretty hiddy compared to a non helical wound xfmr,

I'll try that, would I go with trifilar or should I try with bifilar and get even more turns?

JS
 

Latest posts

Back
Top