rackmonkey

A bunch of transformer freq response test curves
« on: June 04, 2017, 12:35:14 PM »
I've taken an awful lot from this forum over the years without giving much back, so hopefully this will find some use with someone.

EDIT: Bottom line: if you are tempted to/must use oddball transformers for projects (PA, broadcast, scarce or little known manufacturer, etc), then you might find these useful. Skip the dissertation and go straight to link below.

I don't know how many audio transformers I've accumulated over a 15+ year period but it's a lot. In addition to the usual suspects, my congenitally cheap nature drives me to pull from old PA equipment (think Bogen, TOA, later Altec, etc) that I get for less than the price of a single UTC ouncer, and quite a number of these have proven surprisingly good in projects.

To aid in buying decisions for this sort of stuff I wanted to have a quick and dirty reference to remind me whether a particular transformer or unit is worth dropping the change or not. The result is a Google Drive folder with dozens of freq response curves for various input and output transformers under varying primary/secondary source/load impedances. Also helps when I'm trying to decide whether a particular transformer might be workable in a given project.

The folder is here, accessible to anyone with the link: https://drive.google.com/drive/folders/0B__Vr0ik2GcgLVlidXRjd0c2b1U?usp=sharing

The big caveat is that I'm also congenitally lazy about tests like this. So the measurements were done with - *gasp* - white noise rather than freq sweeps. This saved me the time of having to trim the image of the curves at the start/end of the sweep. I'd call that pretty lazy. But I found that I got similar results whether I used a traditional sweep or WN. So I just stuck with the WN approach.

I also just limited the curves to 20 - 20k. But that has served my purpose well enough of just helping me make a buy/no buy decision or a try/don't bother decision for a given project. When selecting for a project, more extensive tests would then be performed.

I guess it brings up the question of the value of WN for this sort of thing vs a sweep. Haven't seen this discussed here before, but I might have just missed it. Again, whatever flaws there may be to this method, it's served my purposes well enough.

Some might be interested in the test setup, so here it is. I used REW for the RTA Fourier analysis. A cheap chinese noise generator was used for the test signal. Cheap, but as you can see from the direct (no xformer) freq response plot, it's dead flat from 20 - 20k. The unit has a 50ohm output impedance. Depending on the transformer/application I wanted to test, I put a series resistor in to get the source impedance to where I wanted it (e.g., 50ohm output Z + 150ohm series R = 200). This unit only outputs 5V RMS - fine for most mic/line input transformers, but inadequate for anything else. So when required, I ran this signal through a cheap DC amplifier (100v p-p, 1A, DC - 50k +/- 0.5dB). This unit also has a 50ohm output impedance, so I used the same approach to source impedance as with the generator.  Always use this with output transformers and it seems to excite the cores adequately.

I ran the transformer secondary into a GenRad decade resistance box to set the load impedance and from there the signal went into the PC sound card. I didn't mess with Zobel networks or add an output capacitor to the setup. Obvious caveat on the latter, but remember the purpose was quick and dirty, "is it worth even buying/messing with in project X".

My soundcard has an obvious slight response dip from about 16k - 19k, then comes back up to flat at 20k. That's consistent across all these tests, so draw an imaginary line across that dip for a more accurate response picture. I'm currently building a Pete Millett SC interface that will hopefully improve things in this regard.

Oh, and don't bother trying to glean anything from the levels in REW, other than relative response levels. I constantly use that program and am always fiddling with the SC input sensitivity which affects the levels REW RTA displays. So those are all over the map. No, I didn't test mic input transformers at +30dBm or anything like that. Signals into mic input transformers were generally around 50 - 70mV, which would be around the high end of what a modern condenser mic puts out (based on my tests w/a microvoltmeter, anyway). Sometimes I tested w/lower levels depending on the transformer and mic type it (appeared) to be designed for. Line input transformers varied by experimentation in the 1.25V ballpark and up. Output transformers were subjected to varying signal levels depending on either specs (for known transformers) or experimentation (unknown) at up to about +26 - +30dBu (tops - not all were tested to these levels. Just depended on what the RTA showed). Sorry these are not documented in the tests. Again, L.A.Z.Y. But the levels were  at least "appropriate" enough to give a decent enough picture. Project results bear that out. Remember the purpose.

Anyway, hope someone can use this stuff. I've found it very handy and it's saved me quite a bit of time and (I assume) money in buying/picking xformers for my projects.

I've got a backlog on transformers I still need to test. But I've largely skipped over my rather sizable collection of off-the-shelf UTC transformers since the old catalogs have plenty of info on those. There are some tests on those, but I probably have 20 or so A-x, O-x and LS-x's I'll ultimately test that I haven't yet. I did test all the custom UTCs I've acquired though. Some nice surprises there.

(BTW I didn't bother to detail how I come up with impedances for unknown transformers. In short, DCR tests on the various pin/wiring combinations gives the roadmap. Then (if needed) test for winding ratios. From there, terminate with common resistor values via GenRad decade box and test opposite windings with a little Peak Atlas LCR45 set at a 1kHz signal. Read magnitude and phase of impedance from there. Works surprisingly well and matches up with nominal specs for known transformers incredibly close. Never had major problems with using this approach in projects.)

Anyway, enjoy (or scoff, if you prefer). Thanks for all that this forum gives to us trying not to electrocute ourselves on a daily basis.

BT

« Last Edit: June 05, 2017, 12:01:32 PM by rackmonkey »
Whether you think you can or you think you can't, you're probably right.


rackmonkey

Re: A bunch of transformer freq response test curves
« Reply #1 on: June 07, 2017, 02:30:57 PM »
Added:

Carnhill VTB2280 (600:600 and 600:150)
ADC 114C Input/Splitter 200:600 (250 nominal on primary)
Sescom MI-44 Input/Splitter 200:600
Whether you think you can or you think you can't, you're probably right.

warpie

Re: A bunch of transformer freq response test curves
« Reply #2 on: June 07, 2017, 04:38:33 PM »
Thanks for doing that!  :)

rackmonkey

Re: A bunch of transformer freq response test curves
« Reply #3 on: June 09, 2017, 04:57:24 PM »
Added:

UTC A-11 (200:50k and 500:50k)
UTC A-22 (500:125 and 500:500)
Whether you think you can or you think you can't, you're probably right.

rackmonkey

Re: A bunch of transformer freq response test curves
« Reply #4 on: June 13, 2017, 02:07:08 AM »
Added:

Akai M8/Roberts 770x Output Transformer (6.5k - 8.5k: 8ohm)
RCA 12301-A Input Transformer (~30 - 50ohm: 50k (probably for dynamic, crystal or carbon mic (?) - grid); tested multiple primary configs at 200:50k and 50:50k.
Whether you think you can or you think you can't, you're probably right.

rackmonkey

Re: A bunch of transformer freq response test curves
« Reply #5 on: June 14, 2017, 01:58:17 PM »
Last post of just a couple of transformers. I'll save up and post batches of several in the future to keep this thread from jumping up top so frequently. Wanted to get these out though because the Ampex is so common and the Protec shows up on Ebay quite often as part of a line splitter card (or something like that).

Added:

Ampex 4580200-01 Bridging input Transformer (Green base, 20k-20k)

Ampex 4580200-01 (Brown base, also 20k-20k, just to show there's no differences between base colors alone. The 4580200-02 is a step up input transformer (line, I think) and has both green and brown base versions as well. No difference between those, either. Will test again and post when I have a chance)

Protec RÜP 1401M Input Transformer
Whether you think you can or you think you can't, you're probably right.

rackmonkey

Re: A bunch of transformer freq response test curves
« Reply #6 on: July 06, 2017, 12:14:46 PM »
Added:

As usual, performed several tests based on the measured or spec'ed impedances, often with one or two "what if's" to test the effects of, for instance, using a 600ohm mic with an input transformer that measures 50 or 200 ohms on the primary. When this was done, in most cases the reflected impedance of the higher (or lower) "what if" impedance on the opposite winding was used in the test.

JBL 5195 Plug-In Input
Raymer MT-3 Plug-In Input
Freed 39943 Input (apparently for HAM radio or consumer-grade tape electronics. Pretty useless for our typical purposes)
Kearfott 215820 (made for the aviation industry, I think)
Rauland 44-1002 Plug-In Input
Rauland R-1002 Plug-In Input
Saratoga Industries TF4R21AJ316 (looks like possibly a UTC or maybe Microtran. Military designator.)
McMartin MT-3 Plug-In Input
Whether you think you can or you think you can't, you're probably right.

emrr

Re: A bunch of transformer freq response test curves
« Reply #7 on: July 09, 2017, 02:17:14 PM »
Thanks for posting these.  Couple thoughts.

Maybe I missed it, but with grid impedance transformers (over 10K), how are you dealing with load between the R load on the secondary and the sound card?   I use a flat response active DI box with 2M input Z. 

I always use pink noise, seems more representative of real world sources than sweeps or white. 

I take all sweeps 0-48kHz, lots to be gleaned from the out of band info, informs phase response if you aren't able to plot that.  Response below 20 informs distortion, above 20k informs need for specific loading requirements. 

Lots of surprises to be discovered testing off the shelf UTC's/etc which have catalog plots.  Frequently there are differences to be discovered. 

Look at grid transformers with an 'unloaded' sec, as would be typical of most circuits, load only the primary. 

Hi-Z secondaries, in absence of specified grid/ground terminals, also look at it reversed, as C to ground can vastly alter response. 

Single ended versus push pull connection with anything having a split windings like HA-100 etc, sometimes very different response result for each condition. 


Best,

Doug Williams
Electromagnetic Radiation Recorders

"I think this can be better. Some kind of control that's intuitive, not complicated like a single knob" - Crusty

"Back when everything sounde

rackmonkey

Re: A bunch of transformer freq response test curves
« Reply #8 on: July 10, 2017, 01:57:52 PM »
Thanks for the advice, Doug. Great stuff to consider.

"Maybe I missed it, but with grid impedance transformers (over 10K), how are you dealing with load between the R load on the secondary and the sound card?   I use a flat response active DI box with 2M input Z."

I didn't mention it. Good catch. I think my soundcard's input is only 2.5k, so I've just left my Countryman type 85 DI hooked up as I do for just  about everything unbalanced that goes into the SC. It sits behind the computer so I just forget it's there. I don't know the input impedance of the Countryman, but I know it's FET-based so it's got to look like infinite input impedance to pretty much any source. I've wondered though if some aspect of this signal chain might somehow be responsible for the low frequency hump around 50 Hz that I often see with HiZ input transformers. Can't think of a reason why, but maybe I'm overlooking something.

I always use pink noise, seems more representative of real world sources than sweeps or white. 

As for white vs pink noise, this is one I've pondered and maybe you can set me straight. I've always found the technical explanations a little cloudy. As I understand it, white noise contains equal energy across all frequencies across the band of interest. Pink noise drops off at 3db per octave. This dropoff better aligns with human hearing, which is more sensitive to mid-high frequencies than low frequencies. So humans perceive pink noise as relatively flat, making pink noise ideal for room equalization, where you want to compensate eq to balance perceived sound levels. White noise OTOH contains frequencies of equal amplitude across the spectrum, which it seems to me to be closer to a frequency sweep. I've read that mic manufacturers use white noise to measure the response of mics. Given that humans are the ultimate target of the sound produced, maybe pink tells us something more useful, I don't know. I just assumed that since white noise contains equal energy across all frequencies that it would be the most useful noise type for just showing relative response levels within an electronic network. Maybe I have this stuff all wrong, though. 

I take all sweeps 0-48kHz, lots to be gleaned from the out of band info, informs phase response if you aren't able to plot that.  Response below 20 informs distortion, above 20k informs need for specific loading requirements. 

I've seen some references here on the usefulness of OOB frequencies, but never understood the reason.  So by saying >20k informs loading requirements, are you referring to resonant frequencies and zobel networks? What does it tell about phase response? If both of these, what's the relationship?

Lots of surprises to be discovered testing off the shelf UTC's/etc which have catalog plots.  Frequently there are differences to be discovered.

I'm with you here. I've been pleasantly and unpleasantly surprised by Triad, UTC etc actual responses. It also helps to understand how sensitive specific transformers are to things like shield grounding, how sensitive they are to using (abusing?) with impedances that are close but don't quite match spec'ed, etc. I just didn't make measuring these as much of a priority as measuring the pulls/oddballs/unknowns but I am making progress here. Will post more as I test them.

Look at grid transformers with an 'unloaded' sec, as would be typical of most circuits, load only the primary. 

This is really interesting. I assumed that by applying a load on the decade box that I was just simulating the load the grid presents. But I'm actually just basically adding a load resistor across the secondary as some input circuits have? I'll definitely start testing those unloaded as well. Cool.

Hi-Z secondaries, in absence of specified grid/ground terminals, also look at it reversed, as C to ground can vastly alter response. 

Have definitely seen this and I document it in the accompanying notes I keep on each transformer, but don't have this info in electronic form yet. My intention is to have measured DCR, pinout, measured impedance and impact of reflected impedance on the measurements, winding ratio, and primary inductance all captured in a PDF file for each transformer along with these white noise tests. I'll eventually get that done and share it here. Maybe even a picture of the transformer with it. Kind of like Match.com for transformers... ;-)

Single ended versus push pull connection with anything having a split windings like HA-100 etc, sometimes very different response result for each condition. 

This is another one I've seen referenced here a lot.  I really should go back and check this out with transformers I've already tested. With these WN tests, just assume the test you see for any given OPT is for single-ended use, because that's the only way I've had things set up so far. I'm going to change that from here out. Thanks for the reminder!

Thanks again, Doug! Appreciate all you do for this forum!

BT
Whether you think you can or you think you can't, you're probably right.

emrr

Re: A bunch of transformer freq response test curves
« Reply #9 on: July 10, 2017, 04:13:04 PM »
I'm sure the Countryman is a fine Z, could be responsible for the humps you see, maybe you already ran a flat test?  I can't remember. 

White versus pink:  I don't have any great logic to apply other than 1) if I gotta listen to it, and 2) distortion point is at lowest frequency in a transformer, so weighting in that direction doesn't hurt that I know of. 

I can't quote specifics offhand, but there's a direct relationship between roll-off points and phase response.  Daisy chain some amps together and you can end up with in-band phase reversal at some parts of the spectrum.  Yes regarding resonances, re: Zobel's, etc. 

Yes, most tube preamps at least use unloaded secondaries, tube load is capacitive at the grid so high end roll-off is the usual result.  Unloaded is 3dB more 'free' gain than loaded.  Resonances may look different. 

PP tests get tricky, lots of transformer don't look the same with the windings reversed, so the though of injecting signal backwards through a grid transformer so you can take the low Z side in your sound card doesn't always give a meaningful reading.  I have used two active DI boxes in a mixed array to achieve a 'balanced Hi-Z input' that resembles a PP tube amp for these tests.   Outputs are more simple to test!

Best,

Doug Williams
Electromagnetic Radiation Recorders

"I think this can be better. Some kind of control that's intuitive, not complicated like a single knob" - Crusty

"Back when everything sounde


rackmonkey

Re: A bunch of transformer freq response test curves
« Reply #10 on: July 10, 2017, 05:16:53 PM »
I did run a flat test - it's in the folder and other than the dip between about 16k - 19k, it's flat.

Thanks for the info. Gonna have to play with the PP thing a bit.  I like your DI mixed array idea.

BT
Whether you think you can or you think you can't, you're probably right.