Testing unknown audio transformers

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

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

Michael A

Well-known member
Joined
Mar 9, 2006
Messages
67
Hey, I'm trying something different here, so I don't know if this post will work, but.

I picked up some old scientific equipment and it had three pretty little transformers in them that were feeding the grids of a 12AX7 from a low level input. Actually it was part of a lie detector. So I've not had the best luck in the past testing transformers and it continued today as my ancient bench equipment kept telling my it needed a little more attention.

So anyway I put the transformer on the analog outs of a MOTU 2408Mk2 and took the output of the transformer back into the analog inputs. I used RMAA5.5 to analyze. Is this going to give me valid data? So this is what it spit out. There are three taps on the primary and a CT secondary. Pic is of the frequency response of the three. Transformer secondary was loaded only by the input impedance of the 2408, except the highest step up that I had too add a 30k resistor in parallel to keep it from clipping the 2408 input. I couldn't handle it in a more elegant way, but that's another story.

image
Three%20Spectra.jpg

syntax changed by moderator -PRR

Let me know if this post is screwed up or if the picture doesn't show, as I have not tried posting one here before and all forums seem to work differently.

What do you think? Will this kind of test give quality data? Of the responses?

Thanks,

Michael

ps, what the heck is wrong with html. I don't think this image is going to post so if you could tell me what is wrong with the syntax above, I would appreciate it.

edit:partially fixed link
 
> what the heck is wrong with html

I found a couple things.

You seem to have tried both URL and IMG and they got tangled (this forum's edit-box will sometimes try to auto-complete-tags and get it wrong).

Inside the IMG tags, your syntax was correct BUT this forum's parser does not believe in image URLs that have a "?" inside them.

Your image's (awful long) URL ended in

Three%20Spectra.jpg?dc=4675565539199208412

That bit after "jpg" may tell MSN Groups something, but when I tried chopping the "?dc=467..." away, I still got the right image. When I put that ?-free URL inside IMG tags, it worked here.

A minor point: your image is 512 pixel wide and the poorly-enforced rule here is to stay below 400 pixels so the layout does not blow-up on smaller monitors. Ethan has a buggy script to re-size embedded images: today it works and is crunching your 512 pixels down to 400 pixels, making some of the labels ugly. I left the URL link with just the word "image" in it so folks can open the plot full-size in a new window; that would be prettier than 512->400, though not a big difference in this case. (When folks post embedded 1600-pixel porn, I just won't look at it.)
 
> feeding the grids of a 12AX7 from a low level input.

Mike-to-grid transformer.

> put the transformer on the analog outs of a MOTU 2408Mk2

The line-level output is probably way too hot for mike transformers; also a mike transformer should be fed from a known mike-like impedance. I'd use a 15K:150 pad to knock it down 40dB and show 149 ohms to the mike. Likewise a modern line input is much lower impedance than a couple grids, so you have loading effect that won't happen in the intended (or similar) use.

Try to set things up as near as possible to what you plan to use the transformer for.

> Actually it was part of a lie detector.

Hmmmm. Either they wanted exceptionally great audio of the answers, or this was a data channel. Then the center-tap push-pull might make sense. They could have used the CT to gain-control the tube, or they could have been chopping a low slow signal so it could be boosted in an AC amplifier. Dunno how much you want to learn about stone age polygraphs.
 
Thanks for fixing the link. I could probably think of several disadvantages of html for a moderator, but for a user it lets one use and external editor to generate that messy code stuff. I'll try a little different approach and see if that works to fix that 400 pixel thing. I agree those MSN names are very long and some sites' editor insists on wrapping and inserting linebreaks. I thought maybe that unfortunate sequence of letters generated near the front of the address might be giving me trouble too. What is the random chance of that?



Back on topic, these were out of a dc amplifier that has a dc-ac chopper in it. Not knowing much about such stuff I thought it might be a level to frequency converter.

The windings ratios seem to be around 1, 3.5,24:24. The DCR is, I think, very high. The secondary measures 10kohms end to end. I think I remember reading once that windings like these can somehow throw off the ohmmeter, but still I think it is high. I fear for the high frequency response. I have the perfect generator to test with, an ancient HP unit that has various output impedances from 50-600 ohms with some balanced and some unbalanced. Unfortunately, it decided to quit working properly when I fired it up yesterday. Would sure be nice if they would have useable response as they are lovely little round potted things about the size of the UTC HA series mic inputs.

Will see if I can get something to work for a proper test.

Thanks,
Michael
 
> think of several disadvantages of html for a moderator

On most boards, the moderator is just another member with some power to help users.

Raw HTML in messages is a very bad idea. Aside from cosmetic issues like messing-up the layout with dangling tags, many wonderful things are possible, such as scripts that steal log-ins. Some boards let you type naked HTML and try to filter out the bad stuff. Most now use non-HTML markup in the Compose box, so you can do the few essential tags but can't do anything unusual.

> for a user it lets one use and external editor to generate that messy code stuff.

I use UltraEdit to compose 99% of my postings. The first web-based board I saw used a dinky postage-stamp size Compose box, so small I could not hear myself think. Also it tended to lose new posts, especially long ones.
 
I finally got my sig gen working moderately well. Don't know yet if the output imedances are as stated. Anyway, I found these instructions for measuring the input and output impedances of an unknown transformer. To whit,
1. Apply low frequency signal to the primary and measure the voltage across the secondary.
2. Add a variable resisitor across the secondary winding and increase/decrease the resistance until the output voltage drops 3 dB. (Question: Is that to 0.707xoriginal voltage in this case?)
3. Leaving the resistor as set across the secondary, add a variable resistor in series with the signal on the primary. Set to value that results in an addt'l 3dB decrease in voltage measured across the secondary.
4. Sweep frequencies of sig gen to find lower and upper -3dB points.
5. Measure values of the variable resistors. These give the intrinsic impedances of the transformer and the sweep gives the small signal frequency response range.

Does this seem like a good procedure? Can anyone suggest a better one?

I have tried it and gotten some rather varied results on repeated application to the same transformers. Could be very sensitive to test lead dress on the high frequency sweep? Also seem to have noticed a bad effect when I add a 1:1 scope probe to the output.

Guess I will try to refine my technique in absence of other methods/suggestions.

Thanks,
Michael
 
Scope probe on 10X for hi freq sweep.

(probe capacitance close to leakage of xfmr)


PRR has a Crowehusrt article, measuring up a transformer.
Maybe he can look for it after he upgrades the Math Dept. server.
 
Actually, I scanned that article, but I don't remember at the moment where I uploaded it. It might have been on PRR's server.

OK, I did a search here on the correct spelling (Crowhurst) and found it:

PDF
 
Thanks for the articles. He gets pretty in depth in the second one. I wonder if he ever wrote the next one in the series, where he tells you how to use the measured parameters for optimum circuit design? I'm still struggling with simple frequency response measurements, but I see from the article that perhaps it is not so simple.

Thanks again,

Michael
 
Don't worry about frequency too much. Just about any audio xfmr will get close to 20-20 at some level.
I guess res peaks are interesting as far as looking at the relationship between leakage C and L.

If you want some fun, set up a B H experiment.
Of course were not going to buy a looper. We are DIY cheap, man!



bh_looper.jpg


http://vacuumbrain.com/The_Lab/TA/Transformer_Testing/bh_2.jpg
 
A simplified version of the Crowhurst setup. Should be fairly straightforward to put together. Part of the deal with the transformer I am most interested in testing is that I don't know that it will pass high audio frequencies since it wasn't from specifically audio equipment.

In a general way I would like to be able to measure transformer parameters to be able to calculate zobel networks and the like. I have built one once before for an UTC A11, but it was a trial and error thing as I recall.

CJ, I understand from other posts that you have a nifty winding setup. Did you diy? I have been thinking of putting something together along that line myself. I've collected a few parts including a small lathe and a variety of dc and stepper motors.

Thanks for the info,
Michael
 
I am just getting into the winding thing again.
I worked in a transformer shop for ten years.
Liked cutting core steel the best.
Set it up, turn it on, crank up the radio, and do what you want for 7 1/2 hours in the deep, dark, backroom confines of Aydin Energy.
We were in between Varian and HP over there in Palo Alto.

You can learn a lot by taking transformers apart, also.

You can classify xfmr's by

mic/line in
innerstage
output

like RDH4.

I think

1:1
1:5
1:10

is a better way.

Transformers of the same ratio tend to behave somewhat alike.
 

Latest posts

Back
Top