Mixed feedback drive to extend bandwidth of cheap transformer

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

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

Studio Mollan

Well-known member
Joined
Oct 26, 2011
Messages
323
I'm investigating this for a project I'm currently working on.
I want galvanic isolation from a reamp interface to a fx-pedal/tape-echo/etc.

I need a 10k 1:1 transformer and a"NICE" one would set me back about 100€.

So on to the question.
These cheap Xicon transformers have a rated +-3db frequency response between 300Hz and3.5kHz.
Would driving the transformer via mixed feedback drive extend the response to something more useable?

Transformer datasheet:
https://www.electrokit.com/uploads/productfile/41016/41016665.pdf

Mixed feedback drive paper:
https://www.lundahltransformers.com/wp-content/uploads/datasheets/feedbck.pdf
/
Emil
 
Back in the 70s I spent some bench time seeing if I could use active feedback to linearize a cheap transformer to get the transformer isolation, but solid state performance. Admittedly I experimented with really cheap small transformers and had zero success.

I suspect you can make a good transformer better, but even back then I had the option to lose transformers entirely.

YMMV

JR
 
If it's an output transfomer you don't want 10K. And you're going to need something that goes to 10kHz or better. The negative impedance converter (NIC) is going to help primarily (only?) with LF. So the Xicon is just not going to cut it. Something like an OEP that is 600:600 with a primary that is >1H is probably about right and shouldn't set you back too much.

Otherwise, yes, according to this page:

  https://sound-au.com/articles/audio-xfmrs.htm#s5

it sounds like you can significantly improve a small transformer's LF response with a NIC. But clearly using values that are stable is quite important. You'll need to figure out what the primary resistance is and the right resistor to use in the positive feeback network. Lot's of testing is required.

But for a one-off project, it's probably not even worth it. A JT-123, which can be had for ~$40 USD, driven with an RC4580 (either diff or para) is probably about as good as it gets under +20dBu, a THAT1646 is pretty great too and a simple a impedance balanced is just dandy IMO.
 
squarewave said:
If it's an output transfomer you don't want 10K. And you're going to need something that goes to 10kHz or better. The negative impedance converter (NIC) is going to help primarily (only?) with LF. So the Xicon is just not going to cut it. Something like an OEP that is 600:600 with a primary that is >1H is probably about right and shouldn't set you back too much.

Otherwise, yes, according to this page:

  https://sound-au.com/articles/audio-xfmrs.htm#s5

it sounds like you can significantly improve a small transformer's LF response with a NIC. But clearly using values that are stable is quite important. You'll need to figure out what the primary resistance is and the right resistor to use in the positive feeback network. Lot's of testing is required.

But for a one-off project, it's probably not even worth it. A JT-123, which can be had for ~$40 USD, driven with an RC4580 (either diff or para) is probably about as good as it gets under +20dBu, a THAT1646 is pretty great too and a simple a impedance balanced is just dandy IMO.

I second the That 1646. I've been using it in all sorts of line level gadgets lately. Feels like cheating sometimes.

Edcor has some budget options that perform admirably at roughly $12/piece https://edcorusa.com/pc_series
 
squarewave said:
If it's an output transfomer you don't want 10K. And you're going to need something that goes to 10kHz or better. The negative impedance converter (NIC) is going to help primarily (only?) with LF. So the Xicon is just not going to cut it. Something like an OEP that is 600:600 with a primary that is >1H is probably about right and shouldn't set you back too much.

Otherwise, yes, according to this page:

  https://sound-au.com/articles/audio-xfmrs.htm#s5

it sounds like you can significantly improve a small transformer's LF response with a NIC. But clearly using values that are stable is quite important. You'll need to figure out what the primary resistance is and the right resistor to use in the positive feeback network. Lot's of testing is required.

But for a one-off project, it's probably not even worth it. A JT-123, which can be had for ~$40 USD, driven with an RC4580 (either diff or para) is probably about as good as it gets under +20dBu, a THAT1646 is pretty great too and a simple a impedance balanced is just dandy IMO.

The chips won't cut it. I need the output to be isolated, and unbalanced. This is supposed to feed a pedal and then there will be an unbalanced to balanced interface in the same box that returns the output from the pedal.
I have another thread open on that design, should be right above this.


Interesting with the transformer impedance? I'm trying to simulate the output from a guitar or such. 10k feels like "ball park" to me?
 
The Edcor is absolutely an option. I will look into that. They are harder to come by in Europe but if DonAudio carries them it should be fine.

The reason I choose the small, cheap ones is that that's what radial is using in their boxes. Im not cloning it AT ALL, but I own one and figured I could take a peek inside.

RSRecords said:
I second the That 1646. I've been using it in all sorts of line level gadgets lately. Feels like cheating sometimes.

Edcor has some budget options that perform admirably at roughly $12/piece https://edcorusa.com/pc_series
 
CJ said:
do  not believe those freq specs, it will do a lot better than that,

it is almost impossible to wind a 1:1 with limited bandwidth.

Thanks!
Somehow I had a hunch that was the case.
/
Emil
 
10K:10K implies a LOT of inductance, thus many-many windings. Capacitance may cause imbalance or even top loss.

What quality do you NEED? If you're going the guitar-pedal route anyway, you may be happy with the response of some very-cheap isolationists, like the monacor fga40 - https://www.monacor.com/products/pa-technology/signal-processing/signal-optimisation/fga-40/

/Jakob E.
 
Thats actually a great alternative! A bit rough to mount on a pcb, might be doable though.

The first stage is basically a transformer isolated reamp box. I figured a 10k 1:1 transformer would do that best. But I'm NO expert. Any input on this is appreciated.

Right now there's a Edcor PC 10k/10k in my schematic. Theres no screen on that one though. Do you think that will be an issue Jakob?


gyraf said:
10K:10K implies a LOT of inductance, thus many-many windings. Capacitance may cause imbalance or even top loss.

What quality do you NEED? If you're going the guitar-pedal route anyway, you may be happy with the response of some very-cheap isolationists, like the monacor fga40 - https://www.monacor.com/products/pa-technology/signal-processing/signal-optimisation/fga-40/

/Jakob E.
 
And it does not have to be a straight line 20-20k response out from the transformer. the rated bandwidth of some cheap transformers are a bit rough though, 300hz ot 3400hz is to limited for me. 100hz to 10k would be acceptable as a hypothetical figure.

gyraf said:
10K:10K implies a LOT of inductance, thus many-many windings. Capacitance may cause imbalance or even top loss.

What quality do you NEED? If you're going the guitar-pedal route anyway, you may be happy with the response of some very-cheap isolationists, like the monacor fga40 - https://www.monacor.com/products/pa-technology/signal-processing/signal-optimisation/fga-40/

/Jakob E.
 
Studio Mollan said:
the rated bandwidth of some cheap transformers are a bit rough though, 300hz ot 3400hz
It's a sequel of telephony; all analog telephone systems were designed with this restricted BW. Even DTMF and early modems.
So in order to sell products to the telephone industry, that's what was taken in consideration. But the actual response is probably much better.

100hz to 10k would be acceptable as a hypothetical figure.
The actual HF response for electric guitar is much less than 10k. Typically 4-5kHz is enough.
The LF response is very much dependant on the drive impedance. Driving a nominal 10k with a typical 100-200r source should shift the LF point to less than 50 Hz.
 
thanks Abbey!
I'll compare a couple oc transformers and see which one I like most.
Very good to know about the LF response!! This makes me think I could actually do this with a cheap PCB-style transformer.

Regarding HF. 10k is probably overkill for guitars, I agree, but I intend to use these on a variety of sources. A HF roll of would generally work for this box but If I can have a somewhat extended high end that's what I want.
Thanks!

abbey road d enfer said:
It's a sequel of telephony; all analog telephone systems were designed with this restricted BW. Even DTMF and early modems.
So in order to sell products to the telephone industry, that's what was taken in consideration. But the actual response is probably much better.
The actual HF response for electric guitar is much less than 10k. Typically 4-5kHz is enough.
The LF response is very much dependant on the drive impedance. Driving a nominal 10k with a typical 100-200r source should shift the LF point to less than 50 Hz.
 
gyraf said:
10K:10K implies a LOT of inductance, thus many-many windings. Capacitance may cause imbalance or even top loss.

What quality do you NEED? If you're going the guitar-pedal route anyway, you may be happy with the response of some very-cheap isolationists, like the monacor fga40 - https://www.monacor.com/products/pa-technology/signal-processing/signal-optimisation/fga-40/
You're worried about HF loss of 10K so you recommend a 15K?

That thing can probably only take 2Vpp before it starts to sound like you're playing a gong and not a guitar.
 
It will rarely get into those levels. More like 0.5 would be extreme. Like the signal from an electric guitar. Thats the gist of it, however I will try to get the maximum useable level out of it, which hopefully won't make everything sound like a gong.


squarewave said:
You're worried about HF loss of 10K so you recommend a 15K?

That thing can probably only take 2Vpp before it starts to sound like you're playing a gong and not a guitar.
 
Studio Mollan said:
I want galvanic isolation from a reamp interface to a fx-pedal/tape-echo/etc.

I need a 10k 1:1 transformer and a"NICE" one would set me back about 100€.
if you can accept the loss of a step-down,
2 of the Xicon parts have remarkably good frequency response.

test conditions:
300 ohms source Z
600 ohms load Z
0 dBu source




Mouser part number TU019 (10,000 ohms : 600 ohms)

Loss: 13.9 dB
resulting in the following tests performed at a output amplitude of -9.9 dBu

Frequency Response
20 Hz -0.5 dB
50 Hz -0.2 dB
1000 Hz 0 dB
10,000 Hz 0 dB
20,000 Hz -0.1 dB
50,000 Hz -0.6 dB

Total Harmonic Distortion
50 Hz << .15% 30 kHz filter
1000 Hz << .033% 400 Hz and 30 kHz filters
10,000 Hz << .005% 400 Hz and 80 kHz filters




Mouser part number TM017 (25,000 ohms : 600 ohms)

Loss: 17.9 dB
resulting in the following tests performed at a output amplitude of -13.9 dBu

Frequency Response

20 Hz -0.8 dB
50 Hz -0.3 dB
1000 Hz 0 dB
10,000 Hz 0 dB
20,000 Hz -0.1 dB
50,000 Hz -0.7 dB

Total Harmonic Distortion
50 Hz << .36% 30 kHz filter
1000 Hz << .017% 400 Hz and 30 kHz filters
10,000 Hz << .036% 400 Hz and 80 kHz filters
 

Latest posts

Back
Top