DIY Transformer for use with Bode Plots.

[mawyatt]

This is an extension of the previous thread:

https://www.eevblog.com/forum/testgear/injection-transformers-bode-plots-application/

Awhile back when ordering parts from LCSC we picked up a couple Common Mode Filters C2832691 and C2924780 with the idea to attempt to convert into a Bode Injection Transformer.

Finally got around to fooling around with the smaller filter, and decided to unwind the windings. After removing the center shield and yanking the wires from the black epoxy we were able to remove the windings. Since we didn't have any twisted pair wire, decided to use the wire from the filter and after straightening we used a hand drill to twist the wires in a light twist. This twisted pair was sewn into the core and the ends swapped and mounted onto a fixture we had created with a 3D print. Know our 3D print skills suck, and we need to get better adhesion, but that's another topic!!

Anyway, after printing a lid and base, we mounted some BNC and Banana connectors as shown below. Decided to terminate the transformer with either 100 or 50 ohms, so used 4 2W 100 resistors, and soldered one set of resistors and clamp the other set with the Banana terminal nuts as shown. This allows the remove of a resistor pair for a 100 ohms termination, and 2W were used just in case we provided too much voltage

Here's a plot of the Transformer using a 50 ohm source @ 0.2Vpp (SDG2042) and Siglent SDS2104X Plus under Bode mode.

Note IL at 100Hz is ~ 0.06dBV and at 1MHz 0.12dBV

Not bad for a ~1$ reconfigured CM Filter using same wire, four 100 ohm 2W resistors, couple BNC connectors and 4 Banana terminals, total cost <$10 


Edit: Added another plot out to 100MHz.
Best,

[mawyatt]

Was able to unwind the larger CM Filter and rewind with twisted pair made from the unwind wire. This is a larger core ($3), and larger wire so should be able to handle more DC current in applications where that's required. Performance looks good as does the smaller core. 3D Printed another front cover (Grey), and waiting on a deeper bottom, 12 hr print time

Best,

[RoGeorge]

I wouldn't have expected it to be so linear.
Nice touch 3D-writing on the lid! 

About getting bigger ferrite cores, the biggest yet easy to source (from scraped electronics) would be the magnet from (defective) audio speakers.

Many audio speakers have the magnet made from something that looks like ferrite.  Some can have a very large magnet, much bigger than the usual OTS ferrite cores.  Not to say big ferrite cores tend to be very expensive.  Magnets have a Curie point low enough so they should be easy to demagnetize by slowly heating them with a common stove/oven.

Never tried, but it should work! 

[jonpaul]

Bonjour, BRAVO! nice job,

in 1960s we used the ESI bode plotter with electrostatic paper.
In 1980s we used the fine open loop stability plot system from Deane Venable.

Speaker magnets material will be unsuitable, and  strong BH curve  bias due to the magnetization.

The professional units have much lower LF cutoff than 100 Hz, 1..10 Hz.

Bon chance!

Jon


PS: I might have  old core larger sizes, better material, contact me by PM if interested in better cores

[RoGeorge]

strong BH curve  bias due to the magnetization

Even after de-magnetising the ferrite by heating the magnet above the Curie point in an oven?

[jonpaul]

Hello again RoGeorge...

I have been designing and manufacturing WB trafos for decades....

The ferrite or metallic NiCo speaker magnets are designed to hold the remnant magnetism and not for AC use let alone wideband transformers.
Even after heating above curie the materials properties like u, Bmax, BH curve characteristic is not much changed.
It depends more on the type of ferrite eg MnZn eg and the Ferrite mix, compression, etc.

Recommend the 1980s..1990s TDK and Ferroxcube books notes on ferrite core material properties and selections.

Core  material of WB trafos is generally  high perm (u 10,000) and optimized for freq resp, linearity and not for remnant mag nor power handling.  We used H5C2 TDK, N27 Siemens, etc.


Another alternate to ferrite is powdered iron, eg from MicroMetals.

What you suggest is an interesting experiment, but AFIK the shapes of the speaker magnets are not suitable.

As toroid's and various split cores shapes are readily available at low cost,   I think its not worth the effort.

Just the ramblings of an old retired EE

Bon Chance

Jon

[RoGeorge]

Thank you for the extra info.  Not trying to argue about that, but you know how wishful thinking is.  I'm still curious how good (or bad) such a demagnetized speaker magnet will work in practice. 

The signal levels for an injection transformer are usually very small (needed to be so in order to keep the amplifier under test to behave linearly with the injected signal).  My hope is the signal would be low enough so the BH curve will stay close enough to the origin, so the magnetic hysteresis will remain negligible.



Aside from the ferrite selection, I've clicked the posted links and read more about injection transformers.  Never used one, so I might be asking wrong questions, but why the secondary is still made of many turns?

If I inject signal in a feedback loop, I'll want minimum perturbation of the feedback loop, with minimum parasitic C or L from the transformer's turns.  Why not having the secondary turn as a single piece of wire passing through the hole of the core (so more like a current transformer)?

- this will give a minimal impedance to the secondary, making it to appear closer to the Ri=0 of an ideal voltage source
- if the induced voltage in a single turn secondary is too small, then we can still get good measurements by synchronous averaging (we have a clean and synchronous reference signal coming from the generator)
- in theory, even a two-halfs core clamped around an existing wire in the DUT should work.  We do not care much about the absolute levels, we only care for the ratio between the input and output signal to remain the same across all the measuring bandwidth
- in fact, having a small gap in the core might be an advantage, by adding the "magnetic resistance" of the gap in series with the (otherwise closed) magnetic circuit of the (former closed) core.  My intuition tells this will reducing the hysteresis even more, and linearize the transformer's response.

I wonder why all these are not the norm for injection transformers.  Most probably because none of these improvements matter/work/can be applied in practice. 

[mawyatt]

@RoGeorge
You bring up an interesting thought about reduced windings in the secondary. With the drive coming from a AWG then the drive level can be much higher, maybe a 2, 4 or even 10X secondary reduction. Maybe an experiment is in order

Thanks for the complement, sometime we'll try and use a different color filament for the 3D printed text. Beginning to understand 3D printing a little better, but still a ways to go. Some folks on here are really good at this!! Certainly fun to watch the 3D object "grow", and highly recommended to any serious DIYer!!

@jonpaul
These DIY injection type transformers have a lower 3dB corner below 10Hz and above 1MHz, even tho we conservatively stated 100Hz to 1MHz.

Best,

[2N3055]

Thank you for the extra info.  Not trying to argue about that, but you know how wishful thinking is.  I'm still curious how good (or bad) such a demagnetized speaker magnet will work in practice. 

The signal levels for an injection transformer are usually very small (needed to be so in order to keep the amplifier under test to behave linearly with the injected signal).  My hope is the signal would be low enough so the BH curve will stay close enough to the origin, so the magnetic hysteresis will remain negligible.



Aside from the ferrite selection, I've clicked the posted links and read more about injection transformers.  Never used one, so I might be asking wrong questions, but why the secondary is still made of many turns?

If I inject signal in a feedback loop, I'll want minimum perturbation of the feedback loop, with minimum parasitic C or L from the transformer's turns.  Why not having the secondary turn as a single piece of wire passing through the hole of the core (so more like a current transformer)?

- this will give a minimal impedance to the secondary, making it to appear closer to the Ri=0 of an ideal voltage source
- if the induced voltage in a single turn secondary is too small, then we can still get good measurements by synchronous averaging (we have a clean and synchronous reference signal coming from the generator)
- in theory, even a two-halfs core clamped around an existing wire in the DUT should work.  We do not care much about the absolute levels, we only care for the ratio between the input and output signal to remain the same across all the measuring bandwidth
- in fact, having a small gap in the core might be an advantage, by adding the "magnetic resistance" of the gap in series with the (otherwise closed) magnetic circuit of the (former closed) core.  My intuition tells this will reducing the hysteresis even more, and linearize the transformer's response.

I wonder why all these are not the norm for injection transformers.  Most probably because none of these improvements matter/work/can be applied in practice. 

I agree with Jon, those ferrites in loudspeakers are made of hard remanence materials-  They might work to some point as you correctly say, but I would be curious about it just for an experiment.

[2N3055]

@RoGeorge
You bring up an interesting thought about reduced windings in the secondary. With the drive coming from a AWG then the drive level can be much higher, maybe a 2, 4 or even 10X secondary reduction. Maybe an experiment is in order

Thanks for the complement, sometime we'll try and use a different color filament for the 3D printed text. Beginning to understand 3D printing a little better, but still a ways to go. Some folks on here are really good at this!! Certainly fun to watch the 3D object "grow", and highly recommended to any serious DIYer!!

@jonpaul
These DIY injection type transformers have a lower 3dB corner below 10Hz and above 1MHz, even tho we conservatively stated 100Hz to 1MHz.

Best,

Someone else had an idea before (can't remember who, maybe Tim?) to make a single braid pentafilar winding, and connect 3 windings in series and 2 in parallel for secondary.
Usually trafos are terminated into something like 10-20ohms for injection in control loop (not to make big difference in feedback resistors value).

I also have some CoolMu cores somewhere that  I wanted to try....

I had best intention to try it but life had other plans...

[jonpaul]

Rebonjour...wideband transformer design is a compromise and tradeoff, LF ..HF BW, ratio, # turns Lp, Llkg, Cp-s, Cshunt.

See many fine refs on WB trafo design, one example is /Magnetic Components   from our old friend Steve SMITH,

https://www.amazon.com/Magnetic-Components-Applications-Steve-Smith/dp/0442203977

RE stability/bode polts/injection trafos, i suggest   original papers and hardware from Deane VENABLE in 1980s

https://www.venableinstruments.com/loop-stability

 the excellent modern systems and transformers from Omicron Labs

https://www.omicron-lab.com/fileadmin/assets/Training_and_Events/Webinar/2014-11_Webinar_LoopGain.pdf

https://www.omicron-lab.com

Bon chance!

Jon

[mawyatt]

We decided to make another transformer out of the last CM Filter we have, and removed the windings then twisted the pair together as before. This time we cut the twisted pair in the middle and created two twisted pair windings on the core. The two windings were arranged as primary in series and secondary in parallel. This should yield a Vo/Vi transfer of 1/2 or -6dBv, and we measured -6.85dBv.

For sanity checks on all 3 transformers used as Injection Bode types, we setup a simple circuit for Closed Loop Bode evaluation. The circuit is a simple Op-Amp (LM358) non-inverting gain of 21 with the feedback resistor of 20K and the shunt to ground resistor of 1K. The + Op-Amp input was grounded thru a 1K resistor and the transformers were connected from the 20K and 1K junction to the Op-Amp - input. The Op-Amp was power by +-10V.

A Siglent SDG2042X AWG was used as the input (we could have used the scope built-in AWG). This was on a LAN with the Siglent SDS2104X Plus scope. Amplitude was set to 0.05Vpp so as not to saturate the small trasnfomers, and confirmed by monitoring the transformer output waveforms.

Thanks to tautech for helping us in the past to get familiar with LAN use (and earlier the SDS2104X Plus), as we didn't know "Diddley Squat" about LANs. BTW Dave used this old redneck engineering term "Diddley Squat" meaning "not much", in his latest How a LCR Meter Works video! Guess this term found it's way to the Land Down-Under 


Here's what the latest transformer rendition looks like with the series/parallel windings.

The Bode plots are shown for the mentioned Op-Amp circuit above, measuring the Open-Loop Gain and Phase of a Close-Loop-System. The plots are for the 1st & 2nd transformers with the different cores, and the last is with the split winding transformer. Please note the indicated Gain and Phase Margin indicated by P1 and P2, you see the frequencies at Markers X1 and X2.

Not sure what's going on around 100~200Hz on the left, likely some interaction with the transformer, this should be a smooth plot coming away from 100Hz.

Note the use of the 10 bit mode in the scope, big help with the large dynamic range involved, and with the dynamic channel input scaling the scope performs in Bode Mode, extending the DR even further.

Anyway, this simple, cheap Isolation Transformer seems to work quite well with the SDS2104X Plus and the built in Bode Plot capability. Very powerful feature being able to do Open-Loop measurements on Closed Loop systems without any additional equipment (you can use the scope built-in AWG) other than a DIY low cost Isolation Transformer 



Best,

[T3sl4co1l]

Yeh specifically they're strontium ferrite, a magnetically hard material (wide BH loop); you need MnZn (LF/broad) or NiZn (high freq), soft materials with small loop.

Also amorphous/nanocrystalline, which show up from time to time in pulse transformers, CMCs, etc.

Tim

[mawyatt]

Have no idea what these CM Filter cores are made of, actually surprised they worked as well as they did. Noted on the other thread linked in first post that some folks had used Vacuumschmelze Nanocrystal Core type toroidal cores, only cores I recall using were from Micrometals and TDK, and that was eons ago, never heard of these!! Digikey has a few in stock that run from $10~20 per core.

Has anyone had success with these Vacuumschmeize cores, or other types?

One thing that Jay_Diddy touched on in some of the other threads is that in very high open-loop gain circuits (like op-amps) the negative feedback return signal creates an almost null voltage at one of the transformer terminals. This gives the scope input a difficult time with accurate amplitude and phase measurements, which contributes to the uncertainty at lower frequencies, and may be more of an issue than the transformer core material.

Open Loop response shown below with a OP-07 (Closed-Loop non-inverting gain of  ~17.7dBv) and the region between 100 and 1KHz is chaotic, thus showing from 1KHz to 1MHz which looks good considering it's just plug-in on a board with jumpers. Just for fun swapped the transformer leads and you should see the inverted response (negative in dB & phase response) which is shown in 2nd image.


BTW whether you like the Siglent or not, that's pretty impressive on screen dynamic range for a mid-level DSO that's based upon a core 8 bit ADC, can't wait to see how well the 12 bit version behaves!!


Best,

[T3sl4co1l]

Yes, they live up to the hype -- check datasheets, they're pretty solid.  The best example is the appnote comparing CMCs with equivalent size ferrite parts, the Am./NC are a fair cut above at low frequencies, and remain above even into the skin effect and capacitive cutoff regions.  (As laminated metal, they do exhibit skin effect starting around 20kHz or so.)  Likewise, keep in mind that the A_L figure varies with frequency -- for the same reason, it goes quite lossy at that point (R = X) and so L is decreasing with frequency in that range.  So keep that in mind.  You're just designing for impedance, and they're great for pulse transformers and CMCs because that impedance is just so high, even as it drops off through the capacitive region.

An important but now-archaic application was ISDN transformers, something like DSL but shifted lower I think?  Very wide.  Not sure what's commercially available now, or if it's basically going to be homemade, but anyway for stuff like this, homemade is fine.

Tim

[TopQuark]

I have been down this path of building wide BW transformers before as well.

I tried the Vacuumschmeize cores, while they are good, it likely won't improve things beyond what you have now. My only issue with them is they are quite fragile, I believe my Vacuumschmeize core shattered after living on my bench for a while.

My current build uses a ~ 3 USD high permeability core from Taobao (https://item.taobao.com/item.htm?id=561000102517), works quite well as shown in measurements.

[mawyatt]

Thanks, but that link doesn't work! Do you have another in English, or the part number and OEM, and what size is the core, looks large?

Performance looks good, have you done any open-loop measurements with an op-amp in a closed loop amplifier?

The limiting factor maybe not be the core, but the ability of the DSO to "pull out" the response where the overall loop gain is very high a low frequencies. However, the larger core may have a benefit of allowing a larger DC thru current while still remaining somewhat linear.

Thanks for the response.

Best,

[mawyatt]

I tried the Vacuumschmeize cores, while they are good, it likely won't improve things beyond what you have now. My only issue with them is they are quite fragile, I believe my Vacuumschmeize core shattered after living on my bench for a while.

Please explain the core shattering, do you mean actually cracking while overheating, or physical stress (dropped)?

Best,

[TopQuark]

Yea Taobao is a pain to work with, I've attached a screenshot of the item listing.

It is quoted as a nanocrystalline core with 50mm OD, 32mm ID, 15mm Height. No mention of who "AT&M" makes it or any part number, as per usual for stuff listed on Taobao.

I haven't used it for opamp open loop measurements, but have used it in SMPS loop response measurements and it works well, can't find the screenshot at the moment, might run a test and post the loop measurement later.

TBH as long as enough signal goes through the transformer, the actual response of the transformer does not matter that much to me, as the measurement of the loop response is always done after the injection signal goes through the transformer, not before it.

Cheers

[TopQuark]

I tried the Vacuumschmeize cores, while they are good, it likely won't improve things beyond what you have now. My only issue with them is they are quite fragile, I believe my Vacuumschmeize core shattered after living on my bench for a while.

Please explain the core shattering, do you mean actually cracking while overheating, or physical stress (dropped)?

Best,

It was working fine when I built the transformer, but measuring the transformer response after a few months yielded very poor results. I could hear something loose shaking the core, but the core looks fine externally, as there's a plastic casing around the core material.

I am not 100% certain it was shattered, but re-winding the transformer did not improve things. Most likely it shattered internally when I bumped it around / dropped it off my shelf etc. 

[_Wim_]

The professional units have much lower LF cutoff than 100 Hz, 1..10 Hz.

That quite achievable DIY also. Quite a while ago I made these following these excellent instructions: (http://www.simprojects.nl/images/DIY_signal_injection_transformer.pdf)

My results were posted here:
https://www.eevblog.com/forum/testgear/looking-for-a-low-cost-way-of-measuring-dc-dc-converter-control-loop-response/msg1738013/#msg1738013

[joeqsmith]

Looking at the Jay link, I can't understand why he was showing 10dB/div.   Interesting is he shows a // resistor with the secondary.   I wonder if he was driving a 1M source.

Here's a plot of the Transformer using a 50 ohm source @ 0.2Vpp (SDG2042) and Siglent SDS2104X Plus under Bode mode.

Your source is 50ohms and you are driving your transformer that has a 50ohm in // with the primary?   Then the secondary is also in // with a 50ohm which drives a 50ohm input? 

When you built the second transformer (series parallel) you did not use the // resistors and you also changed how you measured it.  Could you go back and measure it the same way as the first?   

I have been down this path of building wide BW transformers before as well.

Could you change the vertical scale to 0.2dB/div and repost?

[mawyatt]

Looking at the Jay link, I can't understand why he was showing 10dB/div.   Interesting is he shows a // resistor with the secondary.   I wonder if he was driving a 1M source.

Here's a plot of the Transformer using a 50 ohm source @ 0.2Vpp (SDG2042) and Siglent SDS2104X Plus under Bode mode.

Your source is 50ohms and you are driving your transformer that has a 50ohm in // with the primary?   Then the secondary is also in // with a 50ohm which drives a 50ohm input? 

When you built the second transformer (series parallel) you did not use the // resistors and you also changed how you measured it.  Could you go back and measure it the same way as the first?   

We actually removed the all termination resistors on all the transformers, these were originally included to save from having an external termination for some other tests. The series/parallel version would require a termination of 12.5 ohms from proper termination with a 50 ohm source drive.

Here's what I think you are asking for, these are with 50 ohm source drive, transformer input measured with DSO Hi Z (1M) and output terminated with 50 ohms (DSO). First plot is larger core mentioned, second is smaller core and third is with series/parallel transformer externally terminated with 50 ohms (DSO) and a pair of 33 ohm shunt resistors, for a combo of 12.41 ohms (close to ideal 12.5 ohms for series/parallel transformer).

I'll post the primary and leakage inductance of the transform later.

Edit: Here's the transformer measured details measured @ 10KHz.

Large Core in Grey Box Lp 8.2mH, Ll 4.5uH, Cc 122pF
Small Core in Blue Box Lp 8.0mH, Ll 8uH, Cc 71pF
Series/Parallel with Small Core in small Grey Box Lp 5.4mH, Ll 3.6uH, Cc 48pF

Anyway, hope this helps and what you are looking for.

Best,

[joeqsmith]

Thank you.  This was very helpful and makes much more sense.   Could you display them with 0.2dB/div rather than 1 like you show with your first plots.    Maybe 2deg/div.   I'm not sure what the Siglent can show.

It sounds like the Siglent can't route these signals internally so you have some output  that is 50 ohms that you connect to a T.  One leg of the T goes to one channel of the scope set to 1M.  The other leg of the T goes to the transformer primary.  The secondary goes to a second channel of the scope set to 50 ohms.    Is that correct?

When you run these tests, how do you calibrate the system?   Do you just use a section of coax for the thru and normalize to that?   

***
Sorry, but I had one other question.   Assuming the circuit you are testing applies some DC bias, are you testing this effect?   What sort of breakdown voltage do you need between the two windings?   Things go wrong, seems like it could go really bad.

[mawyatt]

There's no way to calibrate the system we know of other that taking cal measurements and then creating plots from collected data with cal corrections. These Siglent DSOs are relatively new to our lab, so maybe someone with more experience with these can chime in.

I would not place much emphasis in the plots supplied beyond 1MHz, since this is the area of our interest resides. We made no special attempts to minimize or equalize cable lengths, or even include cables effects, or low loss cables....just some stuff we had laying around, even had scope inputs set on 20MHz BW! Likely much of the "artifacts" past 1MHz are due to the scope BW, crude setup, cabling and such. For a more thorough test one would use full scope BW, controlled setup, quality short cables, connectors & such as you would expect for proper RF measurements...which this thread wasn't intended for!!

Our intent was to show the CM Filter cores in a DIY configuration utilizing the wire from the CM Filter can be useful for Bode type Closed Loop Measurements within a 1MHz frequency range, thus frequency range note on case covers. I would not use the CM Filter wire if the use entailed a large voltage across the transformer, this would require a proper insulated twisted pair with thick insulation, again our usable is for Closed Loop measurements for Op Amps and such, not high voltage circuits like SMPS.

DC bias is a concern and why we got the larger core, thinking a larger core should have a higher DC saturation capability vs. the smaller core, even though we don't know what the core details are. Likely we'll get a proper core and windings later if need be, thus questions to others that have utilized proper cores.

This effort was just a quick DIYer to "see" if these cores would work and get familiar with the Closed Loop capability using the SDS2104X Plus and especially to "see" if this capability could be applied to high loop gain circuits such as Op Amp based circuits.

The setup has been torn down, but if we get some time will try and rerun with the tighter dB and phase scales.

Anyway, hope this helps explain the DIY effort behind these CM Filters, reconfigured as Isolation Transformers for Bode Injection use.

Edit: Added finer resolution plots for a Thru, Large Core, Small Core and Series/Parallel with Small Core and scope BW is 100MHz.

Best,