new "class x" transformers from Edcor

[NewYorkDave]

Aye, there's the rub. The inductance falls to uselessly low levels at the kind of currents you would need in an output stage. Although the (very incomplete) specs on the maker's webpage imply otherwise, this transformer just doesn't seem suitable for single-ended, series-fed use. It'd probably be great in push-pull or single-ended, shunt-fed.

I wasn't sure of the best signal level to use for the test, so I chose +12dBU which would correspond to a fairly "normal" average level (0dBU) at the secondary.

The core is just under 2" wide. Let's say 1-15/16".

 

[CJ]

You could have them lace one up L3. This means 3 lams going one way, 3 lams going the other, instead of 1 by 1. Easier to lace for the production guys, so cost will stay the same. Max inductance goes down, but incremntal graph will shift it's slope. How much is anybody's guess. What is the core material and what is the W and H? What is stack height?

 

[NewYorkDave]

Core material is M6, width is 1-15/16", height is 1-9/16", stack height is 3/4".

 

[CJ]

That's 625 EI, same lam as in the API 2503. It is actually 1/7/8 th's wide, you just have a litt;e vazrnish or paint on there.

You can get thoses lams with a 0.015 gap on the center leg, maybe that would do the trick if they have some.

B max is listed at 149 x 10^3/K1N gausses per volt at 60 hz.

K1 is .95 in your case, N is turns.

H is listed at (0.132 x 10 ^-3N Oersteds

I am still working on using these formulas, maybe PRR could help you figure out what that stuff means.

 

[PRR]

> N is turns

Turns can be quickly "counted", on an open-frame tranny, by feeding say 400Hz at fairly high level on the 600 ohm winding, winding one turn of scrap wire around the coil, and reading the milliVolts induced in it.

> B max is listed at 149 x 10^3/K1N gausses per volt at 60 hz.

So a one-turn winding, with one Volt of 60Hz across it, will induce 149,000 Gauss.

Is that really B max??? Seems to me that "just" 15,000 Gauss is a lot for iron. In that case, 0.1V at 60Hz will give 15,000 Gauss. And indeed "a tenth volt per turn" is a good guess for many small power transformers. Power trannies don't care about distortion (it just puts a funny load on the utility company generators) but most audio sources can't push iron to 15,000 Gauss cleanly. One-third of that is a better first-crack for audio iron: 0.030V per turn. Knowing the turns-counts from above, we instantly have a reference maximum level for 60Hz audio. It may not be "clean enough", but won't be wildly distorted like 15,000 Gauss would be.

 

[NewYorkDave]

I corrected a deficiency in my test setup and did the tests over again. The numbers come out better this time, but still not "great." The graph has been updated.

Graph

I increased the signal level to +16dBU (to correspond to an output level at the seconday of +4dBM) and didn't bother to go past 35mA this time, since the primary L is pretty much in the toilet by then.

 

[bcarso]

[quote author="NewYorkDave"]I corrected a deficiency in my test setup and did the tests over again. The numbers come out better this time, but still not "great." The graph has been updated.

Graph[/quote]

How are you accounting for phase effects? Maybe you just adjust for the deepest null, while never getting to zero?

 

[NewYorkDave]

Yep! You could say I'm using a "null-ish" detector :wink:

 

[CJ]

That steep 20 hz curve is probably what puts the "thump" into variable mu compressors.

 

[PRR]

If you assume that a higher-current triode has a lower plate resistance, then it may matter very little how much current you pull. Using very round numbers:

0mA - 90H
1mA - 65H = 65mAH
2mA - 43H = 86mAH
4mA - 22H = 88mAH
10mA - 7H = 70mAH
20mA - 5H = 100mAH
40mA - 3H = 120mAH

Or: 1mA in 12AX7 gives 60K Rp, 40mA in 6BX7 gives maybe 2K Rp. 60K and 65H is 159Hz, 2K and 3H is 106Hz. Not the same, but awful similar for two very different tubes and currents.

Also very low-fi.

And +4dBm out still isn't "pro". That's what the VU reads; peaks will be +20dBm. If we know performance degrades slowly, we may cite performance at +4dBm and let the peaks be a little bent; if performance goes all to crap at +6dBm, it may be annoying.

You have a lovely test-rig that measures some ideal property, maybe; but there's really only one way to prove this pudding. Put it in an audio amp.

Nearly all transformer-coupled tube amps have to be considered as POWER amps. That means peak current will approach bias current. What peak current? Say +20dBm in 600Ω, 18mA peak. Round up to 20mA, transform 4:1 to the primary: 5mA. Load the transformer, either 600 on the secondary or 10K across the primary. Drive your BJT at 5mA DC with 4.5mA peak AC current, 45V peak. Use an emitter resistor well over 100 ohms, keep the base bias/drive impedance well below 10K, like near 1K (or even the 90Ω of your H-P oscillator).

Sweep down from 2KHz. Either the sine will drop in level, or the sine will distort bad. If it distorts, note that and reduce the AC level to maybe 1mA peak, keep going lower. You will get the envelope of either small-signal droop or power bandwidth, whichever is more limiting.

You will get slightly different results if both sides of the transformer are terminated. Small-signal response will extend lower, but also at lower level. Since un-terminated connections are common, I'd get those numbers first. Then you can try adding shunt resistance to see what it takes to flatten the bass, but added shunt resistance will reduce peak output.

If we could get a very low plate resistance at 5mA, we could do better. But 12AU7 or 6DJ8 gives around 6K Rp at 5mA... at 16H this works out to 60Hz unterminated, 38Hz terminated.

I think you have a choice of no-DC or low-fi. Is this "Class X" iron any better in no-DC use than their generic iron?

 

[NewYorkDave]

You've detailed some really good methods to characterize a transformer, and I hope people are paying attention and taking it in. But as you say:

You have a lovely test-rig that measures some ideal property, maybe; but there's really only one way to prove this pudding. Put it in an audio amp.

For sure, and that's my next step. But I figured a test jig I could assemble in a few minutes from cheap parts would help me to decide if it were even worth pursuing before I built an amp, even just a breadboard amp.

With my usual disclaimer that I don't really trust simulations... Using a 12AV7 model that has proven very accurate so far, and a crude first-approximation low-frequency transformer model based on what I know about this Edcor, I'm showing an open-loop response of -3dB at 50Hz in an altered version of my "One Bottle Preamp" (with the xfmr taking the place of the output stage plate resistor, and a 1K cathode resistor instead of the original 330 ohms). With feedback the low end flattens out nicely down to 15Hz, although there's an abrupt phase shift and real steep dropoff around 12Hz, which may or may not cause a stability issue with the loop closed. But that's a simulation. We'll see.

Is this "Class X" iron any better in no-DC use than their generic iron?

The primary inductance is quite a bit higher, with slightly over half the DCR of the equivalent WSM-series model. I think it would make a very good matching transformer in a no-DC application.

 

[CJ]

remember, if this is not push pull, and you are using a non gapped transformer, you are going to have to scrape paper clips off the core every month.

 

[NewYorkDave]

Yeah, their webpage kind of implies that it's gapped (it gives a maximum "AC + DC Current" spec) but after looking at these numbers, I'm starting to wonder. I only ordered one sample of each type, so I don't want to tear it apart to check.

Hey, wait a minute. If it wasn't gapped, the core would become magnetized and even after I turned off the juice, I could literally pick up paperclips with it (right?)... Which I couldn't do if the core were gapped (right?).

CJ, since you've tested many more xfmrs than I have, is it normal for plate-coupling transformers designed for this kind of service to show such a steep drop in primary L as soon as any DC is applied?

 

[CJ]

You can tell if it's gapped just by looking at how the lams are stacked. Sice the lams are not perfectly square, and the assembler is not perfect either, there will be tiny gaps that you can see where the E lams butt up against the I lams.
If the core is laced lap one, which is a non gapped transformer, you will see these tiny holes in the side of the transformer alternate between top and bottom. You might have a U channel mounting system on there, which will prohibit seeing anything but the bottom, which will not tell much.

You can do a search on Hanna Curves to get the DC design parameters. Look up your 625 EI lam on the mag met site, should be able to downlaod the data sheet, which will tell you your magnetic path and all that good stuff.

 

[NewYorkDave]

[quote author="CJ"]You might have a U channel mounting system on there, which will prohibit seeing anything but the bottom, which will not tell much. [/quote]

Exactly. That's why I was thinking of another way to tell without having to rip off the mount.

 

[bcarso]

When I'm wondering about the saturation characteristic of an inductor I set up a jig and hit it with a low-duty-cycle variable voltage pulse (function gen driving a big MOSFET, hefty well-bypassed power supply) and look at the current through something of the order of 100mohms on a scope. The current will be a linear ramp until saturation begins to set in, and the abruptness will tell whether there is gapping, whether discrete or distributed. A reverse-connected diode across the DUT will prevent MOSFET breakdown when the FET switches off.

You can also infer the inductance directly this way.

 

[CJ]

This looks pretty close to what you were seeing:

Those different lines represent different gap thicknesses.

 

[NewYorkDave]

F*ck it. This transformer is just not suitable for DC on the primary. I won't waste any more of my time (or yours) on this just because of a misleading spec on the product webpage. I do think it is a good candidate for an AC-coupled output, though, due to its interleaved windings and heftier core and higher primary L (with no DC) than the WSM series. So I am going to try it that way--I'll probably slap it on the output of my One-Bottle Pre.

I really like Edcor for bringing us a good, affordable US-made line of audio transformers, but I really wish they'd get it together with their specs.

 

[tk@halmi]

Dave,

Would it work better using it in a push-pull configuration like the output of the AM864 limiter?

Thanks,
Tamas

 

[NewYorkDave]

Yes, it should work just fine for that. Just be sure to keep the imbalance within a couple of mA.