An inexpensive audio transformer with DIY projects

[boomertech]

It is in their catalog http://triadmagnetics.com/pdf/13.pdf

BTW, another option is to drive it with JFET/BJT differential pair, since primary is center tapped (with problem of maintaining balanced idle current without tight matching and temperature coupling of pair). With JFET pair it would sort of be like push-pull tube poweramp , but JFET output characteristic is not the same as pentode, and distortion would not be as dominantly low frequency any more as compared to voltage drive.

That's the same data that is available from the link in the first post. I was hoping that their site had a bit more info about the core materials.

I was thinking that if this circuit worked out well, that we could expand and try different driver configurations. For the last week or so I have been itching to do something with MOSFET drivers. But to keep John sane, I thought this would be a good start.

If you have any circuits that you would like to share with us, they would be more than welcomed. I would like to keep this thread focused on the TY250. I have absolutely no tube design experience, so that would be something interesting to mix with this tranny.

-Frank

 

[boomertech]

that's awesome Frank. I really like this idea.
i'll build it up soon. hopefully tomorrow.

I just added C11 to prevent any DC issues between the drivers.

-Frank

 

[nshuman]

Cool, yet another fun project. Can't wait to see the results!

 

[andrew]

Subscribed.

 

[blindrabbit]

Subscribed.

Ditto.

 

[boomertech]

Core saturation is determined by maximum flux density for given material, and flux is determined by transformer geometry and current trough windings. If you check table 3. here you'll notice that "expensive" nickel alloys have smaller max flux density than steel (most of audio output transformers are plain silicon steel core, mic/low-level transformers are often nickel alloys because of low level performance). So, IMHO key issue with this transformer is size/geometry, it's most likely made of silicon steel like rest of output/matching transformers, nonscientific rule of the thumb would be smaller transformer -> higher flux density for given current -> lower max current/power.

BTW, Edcor makes great low cost audio transformes (https://www.edcorusa.com/matchers)

I was looking at this COM23 1:1 10K pri and sec for $5.01: [Invalid or Expired Link Removed]

It's nice that they provide the maximum input voltage level and core material, but the frequency response isn't as good as this Triad TY250.

-Frank

 

[boomertech]

Updated the schematic.

I have changed how the level control works in the output amp. There was a risk that a hot output guitar could possibly cause clipping in the output amplifier with a fixed 20dB gain. Now the level control adjusts the output gain from -6dB to +15dB, this will prevent the output amplifier from clipping under normal use.

-Frank

 

[johnk_10]

cool. as soon as I get the GK sim pedal completed and boxed up, i'll start building this one .

 

[recnsci]

It's nice that they provide the maximum input voltage level and core material, but the frequency response isn't as good as this Triad TY250.

Problem is that there is no standard for response, one manufacturer might use +/-1dB at nominal source and load impedance, another might use -3dB from 1KHz level at no load.

------

One comment about circuit you posted (and voltage transformer drive in general). When xformer is in linear region, to the source it looks like primary inductance in paralell with reflected load impedance on secondary. Ballpark estimate for this xfromer is 8H of primary inductance, in paralell with 1Kohm. That's ok for 2134 as long as there is no lot of sub 10Hz stuff, so filtering out thumps and stuff before hitting transformer would be required (nothing dramatic, scaling two DC blocking caps to cutoff at 10Hz would be good enough).
However, there might be a problem when transformer is driven into saturation. When core saturates two things happen: first primary and secondary coupling starts to drop (depending on geometry) and primary inductance starts to drop as well. At full saturation, source sees just DC resistance of primary. Since you used differential drive, primary center tap is at 0V, so each amp sees just half of DC resistance, or about 75ohms. Together with 100ohm resistor, each amps have to drive about 180ohms at full saturation. This is too much of load for 2134, so aon top of transformer distortion, some opamp distortion would happen. If this should be avoided, LM6172 has no problems driving loads as low as 100ohm.
There is one mor thing that should be mentioned. When transfomer saturates, change in primary current does not change flux, and since voltage generated on secondary is proportional to flux change, output drops to zero after some time. If you drive transformer very hard with 100Hz sine, you get bunch of spikes at output. This sounds very ugly, so that's why hard transfomer saturation is avoided in audio at any cost. So, this problem with too much load for 2134 might be moot point since things might start to sound very ugly before 2134 is overloaded.

 

[recnsci]

As far as current dfferential drive, I had something like this in mind.

Basic idea is to strap transformer on top of differential pair, one leg to each end of primary, and power supply at center tap of primary. If primary is symetric, and diff pair is balanced, same currents go trough each half of primary, but since they are of oposite direction, their flux cancels, so net DC flux is 0 and transformer will work happily. Primary should be very symetric, so remaining problem is to maintain diff pair balance

Each tranistor in pair should be biased ta raound 18mA, so quite hot. Xfomer is speced to saturate at around 4.5mA at 1Kohm load. Pair could be biased at 5 or 10mA, but at lower current, when you drive enough current trough xformer to saturate it, there is large swing in pair so we get transistor saturation as well. At 18mA bias, pair it self will generate around 0.2-0.3% THD when transformer is diven to saturation (if more MOSFET tone is desired, pair should be biased lower).

Also, transformer is driven into saturation with around 150-200mV at diff. pair input. I'm not sure, but I guess it's a bit too much for low output passive pickups, so input stage is buffer and has fixed gain of around 3.5.

There are 3 trimmers. TRIM_INPUT is used to set Q1 drain at around 5.5v. TRIM_IDLE should set Q2 source at around 1.2V (this will set tail current at 36mA). Finaly TRIM_OFFSET should be used to null offset of pair. Offset should be measured across primary (there should be 0 volts when pair is balanced). This trimmer does not cover all possible variations in Q3 and Q4 params (when one Vt is 0.8V and other is 3V for example), but should do the job in 99% of situations. Im note sure what si temperature dependance of offset, BJTs are cool with offset dependancy but dont know about fets, so this offset nulling might not work. Also, nulling would be very delicate across variations, so multiturn trimmer migh be usefull or even required here. OTOH, transfomer is ok with up to 4mA of DC current, so thighs might work fine with just ok (instead of perfect) null.

 

[NKBassman]

About the only thing I understood from this thread was the phrase "transformer coloration".
Is this something that could be built into a little studio gadget? For instance, to reamp a clean bass/guitar/snare/etc DI track through to add some "transformer mojo" to the signal?

 

[boomertech]

Problem is that there is no standard for response, one manufacturer might use +/-1dB at nominal source and load impedance, another might use -3dB from 1KHz level at no load.

That's true.

The Triad data sheet has frequency plots with various secondary loads which makes it easy to see how the transformer performs under load. It has been my experience that manufacturers manipulate data to put the best face on their products and let the designer 'discover' their short comings. So when I read 50~15K Hz., <1dBu on the Edcor COM23, that tells me that it is nice and flat from 50-15KHz under whatever undisclosed conditions that they tested it with. Since the COM23 is an audio transformer and most people look for full spectrum (20-20KHz) response, my first impression with the data is that it falls short of the full bandwidth criteria that I was comparing against the TY250.

One comment about circuit you posted (and voltage transformer drive in general). When xformer is in linear region, to the source it looks like primary inductance in paralell with reflected load impedance on secondary. Ballpark estimate for this xfromer is 8H of primary inductance, in paralell with 1Kohm. That's ok for 2134 as long as there is no lot of sub 10Hz stuff, so filtering out thumps and stuff before hitting transformer would be required (nothing dramatic, scaling two DC blocking caps to cutoff at 10Hz would be good enough).
However, there might be a problem when transformer is driven into saturation. When core saturates two things happen: first primary and secondary coupling starts to drop (depending on geometry) and primary inductance starts to drop as well. At full saturation, source sees just DC resistance of primary. Since you used differential drive, primary center tap is at 0V, so each amp sees just half of DC resistance, or about 75ohms. Together with 100ohm resistor, each amps have to drive about 180ohms at full saturation. This is too much of load for 2134, so aon top of transformer distortion, some opamp distortion would happen. If this should be avoided, LM6172 has no problems driving loads as low as 100ohm.
There is one mor thing that should be mentioned. When transfomer saturates, change in primary current does not change flux, and since voltage generated on secondary is proportional to flux change, output drops to zero after some time. If you drive transformer very hard with 100Hz sine, you get bunch of spikes at output. This sounds very ugly, so that's why hard transfomer saturation is avoided in audio at any cost. So, this problem with too much load for 2134 might be moot point since things might start to sound very ugly before 2134 is overloaded.

The TY250 has a measured primary inductance of 3.9H (we measured this for another project that we were working on).

This circuit has two HP filters formed by C5/R6 and C11/R9 both at approx 16Hz.

When I was throwing this circuit together, I assumed worst case for the OPA2134 with the TY250 providing only DCR across the primary coil. I put the 100 ohm resistors between the driver and primary coil to provide a little more resistance seen by the driver outputs. I fully understand that it will tax the 2134 under heavy transformer saturation. This overload response of the 2134 isn't a problem for this circuit, instead it's an intended feature.

Part of the sound that I am trying to recreate with this concept circuit is the output of a pushed tube amp. A lot of older power amplifiers experience power supply sag under extreme conditions and I am hoping the drive limitations of the 2134 would operate in favor of this 'sag' response under saturation. This 'amp' circuit is intentionally designed to become dirty... I am looking for the opposite of high fidelity here, an effect if you will. That is why the gain structure has been set up so that the amount of distortion is adjustable by the user so that a sweet spot can be found.

You are saying that because there will be no flux changes on the secondary under saturation, that the output of the TY250 will drop to zero. I agree that there will be no more changes to the signal at the output of the transformer when the flux capacity of the core has been reached, but I have never seen the output of a transformer drop to zero in this case. Would there not be only signal distortion observed from the lack of flux transmitted to the secondary windings?

-Frank

 

[boomertech]

View attachment 536655

As far as current dfferential drive, I had something like this in mind.

Basic idea is to strap transformer on top of differential pair, one leg to each end of primary, and power supply at center tap of primary. If primary is symetric, and diff pair is balanced, same currents go trough each half of primary, but since they are of oposite direction, their flux cancels, so net DC flux is 0 and transformer will work happily. Primary should be very symetric, so remaining problem is to maintain diff pair balance

Each tranistor in pair should be biased ta raound 18mA, so quite hot. Xfomer is speced to saturate at around 4.5mA at 1Kohm load. Pair could be biased at 5 or 10mA, but at lower current, when you drive enough current trough xformer to saturate it, there is large swing in pair so we get transistor saturation as well. At 18mA bias, pair it self will generate around 0.2-0.3% THD when transformer is diven to saturation (if more MOSFET tone is desired, pair should be biased lower).

Also, transformer is driven into saturation with around 150-200mV at diff. pair input. I'm not sure, but I guess it's a bit too much for low output passive pickups, so input stage is buffer and has fixed gain of around 3.5.

There are 3 trimmers. TRIM_INPUT is used to set Q1 drain at around 5.5v. TRIM_IDLE should set Q2 source at around 1.2V (this will set tail current at 36mA). Finaly TRIM_OFFSET should be used to null offset of pair. Offset should be measured across primary (there should be 0 volts when pair is balanced). This trimmer does not cover all possible variations in Q3 and Q4 params (when one Vt is 0.8V and other is 3V for example), but should do the job in 99% of situations. Im note sure what si temperature dependance of offset, BJTs are cool with offset dependancy but dont know about fets, so this offset nulling might not work. Also, nulling would be very delicate across variations, so multiturn trimmer migh be usefull or even required here. OTOH, transfomer is ok with up to 4mA of DC current, so thighs might work fine with just ok (instead of perfect) null.

Thank you for the circuit and description.

-Frank

 

[boomertech]

About the only thing I understood from this thread was the phrase "transformer coloration".
Is this something that could be built into a little studio gadget? For instance, to reamp a clean bass/guitar/snare/etc DI track through to add some "transformer mojo" to the signal?

It could be. But right now it is a wild concept that I had... it may work as intended, we most likely will have to make modifications to it or it could be a bust. At this point we don't know.

-Frank

 

[boomertech]

John, hopefully you haven't laid down the copper for this circuit yet. I want to make another change to the clippers in the driver circuit.

-Frank

 

[recnsci]

You are saying that because there will be no flux changes on the secondary under saturation, that the output of the TY250 will drop to zero. I agree that there will be no more changes to the signal at the output of the transformer when the flux capacity of the core has been reached, but I have never seen the output of a transformer drop to zero in this case. Would there not be only signal distortion observed from the lack of flux transmitted to the secondary windings?

Yes, something like this http://www.eevblog.com/forum/beginn...ransformer-help-interpreting-waveform-sought/ but you have to drive really hard to see effect to full extent.
However, now you got me thinking, I'm quite sure that effect is highly dependant on actual transformer. For example, I've got here some Carnhill output transfomers for class A Neve modules. Those are big beasts with so called air gap, designed to work with very high DC bias current trough primary (similar to an extent to output transfomers used in small single ended amps like Fender Champ). I'm not sure at all how those saturate (and if they hard saturate at all). So, I guess that only way to know for sure about this Triad is to drive one hard with some kind of poweramp and see what happens. I guess in theory you might see these intense waveform drops if you drive it with 50V sine at 50Hz (but that should fry transfomer it self; they don't have max ratings speced, but at 50V that should be 14W, and one melted transfomer)

 

[boomertech]

Yes, something like this http://www.eevblog.com/forum/beginn...ransformer-help-interpreting-waveform-sought/ but you have to drive really hard to see effect to full extent.
However, now you got me thinking, I'm quite sure that effect is highly dependant on actual transformer. For example, I've got here some Carnhill output transfomers for class A Neve modules. Those are big beasts with so called air gap, designed to work with very high DC bias current trough primary (similar to an extent to output transfomers used in small single ended amps like Fender Champ). I'm not sure at all how those saturate (and if they hard saturate at all). So, I guess that only way to know for sure about this Triad is to drive one hard with some kind of poweramp and see what happens. I guess in theory you might see these intense waveform drops if you drive it with 50V sine at 50Hz (but that should fry transfomer it self; they don't have max ratings speced, but at 50V that should be 14W, and one melted transfomer)

I ran some FFT tests with 20Vp-p, 10Vp-p and 5Vp-p with a 10K load on the secondary in this post:

I ran tests on the TY250 tranny and here are the results with 20Vp-p, 10Vp-p and 5Vp-p signals feeding the primary. I used a USB sound module as the signal source into my SWR power amp to drive the primary side of the tranny. I used a 10K pot on the secondary coil configured as a standard volume pot. I set the volume level on the secondary pot to make sure that no levels would exceed 2Vp-p to keep the sound module input from clipping. The FFT results are interesting:

View attachment 535373

You can see how wavy the trace for 20Vp-p is from 20 to 200Hz. This is what signal distortion looks like in an FFT analysis. This transformer only cost $5 and because it is so inexpensive, I believe that the core is made of low cost materials and easily saturates with this level of signal at low frequencies.

The 10Vp-p trace is a bit cleaner, but there is still a little bit of distortion under 200Hz. The 5Vp-p trace is even cleaner.

-Frank

Although the FFT test is a quick sweep and there was no chance for a meltdown, I don't think I have to worry about the OPA2134 supplying meltdown levels of current with a 40mA maximum.

-Frank

 

[boomertech]

*** IMAGE REMOVED - UPDATED IMAGE LATER IN THREAD***

After putting a little more thought into this concept circuit, I thought that users may want to have the ability to soft clip at lower levels of gain. This additional circuitry will allow for approx 4Vp-p clipping to approx 12Vp-p clipping on U2a driver output.

BTW John, you may want to socket R18 and R19 so we can adjust trans saturation with soft clipping level ratios.

-Frank

 

[boomertech]

I changed VR1 to a dual ganged A500K pot for more available gain to accommodate low output instruments. Also noted the audio pot taper for VR2 (A1K).

*** IMAGE REMOVED - UPDATED IMAGE LATER IN THREAD***

-Frank

 

[johnk_10]

Hi Frank,
I haven't started etching this one yet. do you think it's ready for 'real-world' testing now? if so, i'll start drawing up a PCB for it.