Are these isolation transformers any good ?

[Christ778]

Hi
I'm trying to build R.G's transformer isolated ABY box:
http://www.geofex.com/FX_images/TransformerSplitter.pdf
I'm aware there are many good 10k:10k transformers available from Mouser or Digikey etc, but they are not available at where I live now. This 600:600 variety is what I can get most easily, and dirt cheap too:
https://www.aliexpress.com/item/32827431499.html?spm=a2g0o.cart.0.0.7b623c00n6C1vT&mp=1
Someone tested these with good result:
https://www.audiosciencereview.com/forum/index.php?threads/e14-1-1-600-600-cheap-transformers-some-measurements.10438/
Do you think these transformers will work ? How about two of them in series as a 1k2:1k2 transformer for even less low frequency roll off ?
Thanks

[composition4]

These ones are great... available in Australia, but I'm sure they must be available elsewhere too

https://www.altronics.com.au/p/m0706-10k-ohm-10k-ohm-shielded-bridging-transformer/
https://www.altronics.com.au/p/m0705-600-ohm-10k-ohm-shielded-microphone-transformer/
https://www.altronics.com.au/p/m0707-600-ohm-600-ohm-shielded-bridging-transformer/

[composition4]

Just looked at the splitter schematic. You would have to ensure the op amp you use is happy driving a 600 ohm load (most won't be without high distortion). If you can, stick with the 10k primary

[R.G.]

On a philosophical note, transformers don't have impedances, they have ratios.

A transformer can be modeled as a perfect signal ratio from primary to secondary, paralleled on the primary side by a big inductance.
Let's say you have a 1:1 impedance ratio. That could be a 10K:10K, or a 600:600, doesn't matter. It's the ratio that matters.
If you put a 1K resistance on the secondary, the primary looks like a a 1K resistor paralleled by the primary inductance to the driving circuit. 
Here's the asterisk-footnote. This is true for the middle of the frequency range for the transformer, not on the bass end or treble end, where the parasitics come into play.
If you are not into torturing transformers or driver circuits, if you put a resistor equal to the specified secondary impedance on the secondary winding, then the parasitics work with that secondary load to not only give the specified primary inductance, they also give the specified bandwidth.

... bandwidth??

Yep. The hypothetical perfect transformer part truly does transform the secondary load to the primary side, but that appears in parallel with the primary inductance. The lower the frequency you drive the primary with, the lower the impedance of that primary inductance. That means that the driver circuit's output has to drive an increasing load the lower the frequency goes. So the driver circuit's output impedance has to drive a steadily decreasing load. When the driver circuit's impedance is equaled by the steadily decreasing impedance of the primary inductance, you have hit the bass rolloff of the circuit, completely independent of what is happening on the secondary. So the bass end of the frequency response is determined not by what's on the secondary necessarily, but by the primary inductance of the transformer and the imponderable output impedance of the driver circuit.

That leads you around the circle back to impedance ratio. If you load the secondary with the specified impedance, the primary impedance appears in parallel with the primary inductance, and those are roughly equal at the lowest bass rolloff of the specified bandwidth. You can then calculate the primary inductance from the specified primary impedance and low end bandwidth.

This gets quirkier. Run a transformer at lower impedances, and you extend its bandwidth. Running a 10K:10K with a 1K secondary load makes both the low and high end bandwidth bigger, because the low impedance loading and the necessary primary drive mean that the primary inductance doesn't come into play until much lower frequency.

The result is that those 600:600 transformers work fine as long as you can drive the primary inductance and reflected secondary load. The comment that many opamps don't like to drive 600 ohms is accurate. The 5532 does work fine with 600, and so does the LM833. There are probably others.

[PRR]

This has prettier numbers.

https://www.aliexpress.com/item/32811092393.html?spm=a2g0o.detail.1000014.19.7c547fbd4pJgac&gps-id=pcDetailBottomMoreOtherSeller&scm=1007.13338.146400.0&scm_id=1007.13338.146400.0&scm-url=1007.13338.146400.0&pvid=e686c55a-8342-4cf7-8f76-209eba0f0dfd

However it is clear that anybody can put anything on that site, and some of it may be bogus.

[Christ778]

On a philosophical note, transformers don't have impedances, they have ratios.

A transformer can be modeled as a perfect signal ratio from primary to secondary, paralleled on the primary side by a big inductance.
Let's say you have a 1:1 impedance ratio. That could be a 10K:10K, or a 600:600, doesn't matter. It's the ratio that matters.
If you put a 1K resistance on the secondary, the primary looks like a a 1K resistor paralleled by the primary inductance to the driving circuit. 
Here's the asterisk-footnote. This is true for the middle of the frequency range for the transformer, not on the bass end or treble end, where the parasitics come into play.
If you are not into torturing transformers or driver circuits, if you put a resistor equal to the specified secondary impedance on the secondary winding, then the parasitics work with that secondary load to not only give the specified primary inductance, they also give the specified bandwidth.

... bandwidth??

Yep. The hypothetical perfect transformer part truly does transform the secondary load to the primary side, but that appears in parallel with the primary inductance. The lower the frequency you drive the primary with, the lower the impedance of that primary inductance. That means that the driver circuit's output has to drive an increasing load the lower the frequency goes. So the driver circuit's output impedance has to drive a steadily decreasing load. When the driver circuit's impedance is equaled by the steadily decreasing impedance of the primary inductance, you have hit the bass rolloff of the circuit, completely independent of what is happening on the secondary. So the bass end of the frequency response is determined not by what's on the secondary necessarily, but by the primary inductance of the transformer and the imponderable output impedance of the driver circuit.

That leads you around the circle back to impedance ratio. If you load the secondary with the specified impedance, the primary impedance appears in parallel with the primary inductance, and those are roughly equal at the lowest bass rolloff of the specified bandwidth. You can then calculate the primary inductance from the specified primary impedance and low end bandwidth.

This gets quirkier. Run a transformer at lower impedances, and you extend its bandwidth. Running a 10K:10K with a 1K secondary load makes both the low and high end bandwidth bigger, because the low impedance loading and the necessary primary drive mean that the primary inductance doesn't come into play until much lower frequency.

The result is that those 600:600 transformers work fine as long as you can drive the primary inductance and reflected secondary load. The comment that many opamps don't like to drive 600 ohms is accurate. The 5532 does work fine with 600, and so does the LM833. There are probably others.

Thank you for the explanation, it is really helpful 

So in order to extend the bandwidth, is it a good idea to connect a 10k-20k resistor across the secondary as a way to lower the primary's load when ABY box output is plugged into an amp's 1M input impedance ? (as I assume the test result's relative flat frequency response was obtained with the transformer hooked up into Steinberg UR22 line in's 20k input impedance).
Or is it what the output section's 10k resistors and 0.001uF caps in your schematic are doing already ?

What about core saturation/distortion when this tiny transformer is blasted with full signal from a 12V-18V pedal? (I guess there's no way around physics) Can I prevent core saturation by using two transformers instead of one (as they cost next to nothing), wire the the primaries in parallel and the secondaries in series (so the ratio becomes 1:2 ?), trim the input voltage before it enters the opamp to 1/2 so the gain is unity on the outputs?
 
Maybe more hassle than it worths but this could be interesting on how can you get away with the cheapest transformers possible 

[Christ778]

This has prettier numbers.

https://www.aliexpress.com/item/32811092393.html?spm=a2g0o.detail.1000014.19.7c547fbd4pJgac&gps-id=pcDetailBottomMoreOtherSeller&scm=1007.13338.146400.0&scm_id=1007.13338.146400.0&scm-url=1007.13338.146400.0&pvid=e686c55a-8342-4cf7-8f76-209eba0f0dfd

However it is clear that anybody can put anything on that site, and some of it may be bogus.

They have 10K:10K for the same price too, but yeah they costs a lot for something without datasheet or review/report 

[merlinb]

So in order to extend the bandwidth, is it a good idea to connect a 10k-20k resistor across the secondary as a way to lower the primary's load when ABY box output is plugged into an amp's 1M input impedance ?

Yes, and you would typically add a capacitor or Zobel network across the secondary too, to tame any high frequency resonance.

The transformer you linked to claims a primary inductance of 290mH, which is 91ohms at 50Hz. So if you want flat bandwidth down to 50Hz (basically the whole guitar range) then your driver circuit would need to be able to drive ~90ohms or less. On the other hand, part of the charm of transformer coupling is its defects, so maybe just try a 5532 and see how you like it. Two of those transformers in series is probably a good compromise.

Can I prevent core saturation by using two transformers instead of one, wire the the primaries in parallel and the secondaries in series

Wrong way around. Saturation is caused by too much primary voltage, so you would need to put the primaries in series (secondaries however you want) so each one sees half the primary voltage.

Your 18V circuit can do 6.4Vrms at most. Most of those small coupling transformers can handle 1W into 600ohms at 300Hz, which is 25Vrms. At 50Hz that would be 25*50/300 = 4Vrms, so maybe you'd hit saturation at max output and the lowest frequencies. But barely. And again, that may be a desirable sound.

[R.G.]

So in order to extend the bandwidth, is it a good idea to connect a 10k-20k resistor across the secondary as a way to lower the primary's load when ABY box output is plugged into an amp's 1M input impedance ? (as I assume the test result's relative flat frequency response was obtained with the transformer hooked up into Steinberg UR22 line in's 20k input impedance).
Or is it what the output section's 10k resistors and 0.001uF caps in your schematic are doing already ?

Now it gets a bit complicated. The short answers are that yes, a 10K or so across the secondary might help, and that the output section's 10K and a cap are to kind of do that. But it's for reasons of high frequency response.
The simplest model of the transformer is that it simply transfers voltage at the transformation ratio, period. The next level puts in wiring resistance. The next level is to add the bandwidth determining components: the primary inductance, the leakag inductances, and the interwinding capcitances.
The leakage inductances and interwinding capacitances are effectively not there for the low end of the bandwidth. However, they and the loading determine the high end of the bandwidth. Since you have a series leakage inductance and a parallel and distributed winding capacitance, you have the makings of a resonant circuit. Keeping this circuit from ringing and having a funny frequency response peak at the top end, you have to damp it with loading. The 10K and a cap is a way to add damping, but only above freqeuncies where the cap is less than a 10K impedance. It's a high-end-only load, and if the resistor and cap are the right values, they can, in concert with any actual loading, correctly damp the ringing tendencies and give a smooth, not peaky rolloff on the high end. But that IS a form of bandwidth control.

What about core saturation/distortion when this tiny transformer is blasted with full signal from a 12V-18V pedal? (I guess there's no way around physics)

Yep - no one has ever found a way around physics. How soon the core saturates is related to two things - the combination of core material and core volume. Core materials can only support a mag field up to some maximum. Over that, they revert back to acting like air or vacuum for even more magnetic field strength. You can only fill them so full of M-field. Different materials have different field maximums, and so how much M field you can pour into a core depends on the material and how much core you have to fill. A physically bigger core can hold proportionately more total M field before it saturates. So looked at one way, the bigger the core, the more M field you can get.
Since M field is done by ampere-turns, you can more or less say that a given core will hold so many ampere turns without saturating, and a single coil with current in it is how you make an electromagnet. In a transformer, you're feeding an AC signal into the primary, and there's an inductor there to be driven. An inductor opposes any change in current, so if you put a voltage V on the core, the current ramps up as V = Ldi/dt, L being the inductance and di/dt is the rate at which current rises.
With an AC voltage on the core, the current ramps continuously chasing the voltage. Doing some simple calculus you get that the current is the integral of the voltage times the time difference you measure over, times the inductance.
I can feel your eyes glazing at this point.    The central point is that a fixed size core can be thought of as having a maximum voltage times time rating. Get over that, it saturates.
So saturation is both a voltage thing and a low frequency thing. If you feed a transformer DC there is always some time where it will saturate. And as you raise frequency from DC there will come some combination of voltage times time on the winding that will not quite saturate.
Put another way, you can't saturate a transformer with a frequency above X. Where X is depends on both how big the voltage is and how big the core is.

Low frequency response requires big inductance to not suck more signal current than the source can provide and that same approach happens to lower the frequency for a fixed voltage that will saturate the transformer.

So saturation is a low frequency issue. You can extend the low frequency response at lower drive levels so the frequency response is flat to a lower frequency, but you can only do this at lower voltage drive levels, because there is a maximum V*T level the core can stand. You can drive flatter and wider, but the more voltage drive, the less bass frequency you can have.
 

Can I prevent core saturation by using two transformers instead of one (as they cost next to nothing), wire the the primaries in parallel and the secondaries in series (so the ratio becomes 1:2 ?), trim the input voltage before it enters the opamp to 1/2 so the gain is unity on the outputs?

Better to drive both primaries in series to divide the voltage across each one in half. That puts off saturation by a factor of two. Secondaries in series then give you 1:1.  You may start having issues with high end nonlinearities. Each transformer is (at least) a resonant circuit all by itself, so some tinkering with them to keep ringing down may be needed.

If this seems scattered, or I've missed something already posted, forgive me. I actually wrote it scattered over the day.