Inexpensive toroidal transformer works as a modulation transformer!

[AB2EZ]

Hi!

I was chatting with Peter, K1PHG, today about the feasibility of using a toroidal transformer as a modulation transformer. Tim, WA1HLR, and I had also been chatting about this last week.

As I was talking with Peter, I realized that I had everything I needed to do a quick experiment:

I have an external modulator that I am using with my Ranger, which employs a backward connected Hammond output transformer (model 1629 SEA).

Separately, I have a small toroidal transformer that I recently purchased from Antek (http://stores.ebay.com/Antek-inc) that is the size of a supermarket donut (2 3/4 inches in diameter and 1 inch high), cost $8.00 plus $8.00 shipping, has a pair of 115 volt windings and a pair of 12 volt windings, is rated at 25VA at 50Hz, is specified as having been tested for dielectric breakdown at 3500 volts between the primary and secondary windings, and weighs 1.3 pounds.

http://cgi.ebay.com/12-12V-25VA-Toroid-Toroidal-Power-Transformer_W0QQitemZ370012573775QQihZ024QQcategoryZ73152QQcmdZViewItem

Note: this toroidal transformer employs a high permeability, solid core... and some of our intuition about transformers, based on traditional core materials/structures may not hold for this type of transformer. For example, because of the very high permeability... the number of turns on each winding is much lower than for a traditional transformer, the insulation is probably thicker and the capacitance between turns of the windings is probably less than it is in traditional transformers with the same turns ratios. The breakdown voltage is probably higher than we might expect (see the specification on the Antek web site) because of the thicker insulation. The leakage inductance is also lower than we might expect in traditional transformers whose core materials have lower permeabilities. There are no laminations to vibrate.


Using the 115 volt windings in series (230 volts), and the 12 volt windings in parallel, the turns ratio is 19.2:1. Therefore, the impedance transformation ratio is 367:1.

Using the 8 ohm output of my audio amplifier, the equivalent source impedance at the output of this transformer is less than 8 ohms x 367 ~ 2940 ohms... which is just fine for modulating my Ranger, because I am not concerned about the efficiency of power transfer from the audio amplifier to the Ranger.

Since the 230 volt rms winding corresponds to a peak voltage 230 x 1.414 = 325 volts... I felt that I could fully modulate the Ranger, when I run it at low power, without creating too much hysteresis loss in the transformer. [It actually worked fine (100% modulation) with 600 volts B+ (on the Ranger's 6146) as well, and the transformer stays perfectly cool with normal voice modulation.]

So... I disconnected the Hammond transformer, and inserted the Antek transformer in its place.

The result: it seems to work perfectly. Nice clean 100% sinusoidal modulation from 30Hz to well above 5 kHz, with the Ranger running at 40 watts (carrier level) output.

Amazing... this transformer is 1/10th the cost, 1/10th the weight, and much smaller in physical size than the Hammond transformer.

Antek makes transformers on toroidal cores capable of handling 800VA (32x the VA rating of the little one I am using). Furthermore, if you use a transformer with a turns ratio of 1:1 or 1:1.5 (etc.) you will have a traditional modulation transformer.

Important note: I am running the modulator in a modified Heising configuration, so there is no unbalanced DC going through the high impedance (output) winding of the transformer. I suspect that one would have transformer core saturation problems if one did not use a modified Heising configuration. [I did a quick check by switching my configuration from modified Heising to "standard"... and the output of my off-air monitor sounded very distorted]

Best regards
Stu

[WA1GFZ]

Stu,
Early switchers used tape cores and the best ones were wound on cores with the thinner tapes. I'm pretty sure Variac cores are also tape cores and have the thicker tapes. The thinner cores gave the best high frequency performance. Who knows maybe a variac core is good enough. So you take a dead variac and count the turns for say 130 volts RMS across the whole winding. Now just scale the turns up and the wire size down. The only worry is the DC offset I bet a 20 amp variac could work pretty well at hundreds of watts and have plenty of room to wind wire.

[Steve - WB3HUZ]

Very cool Stu. Great info. I've often wondered about toroidal transformers in modulator service. Just think how great they would work if they were actually designed and wound with audio in mind.

[ka3zlr]

Very Nice Post Stu....good food for thought OM.

[WA1GFZ]

Tape thickness of a tape core is thinner than E-I lamination so there are less eddy current losses. You still have to worry about dc offset saturation. no free lunch
A Toroid are harder to wind and isolating layers takes more work than an E-I bobbin.

[Opcom]

variac ratings in their spec sheets indicate that many of them can be used at audio frequencies up to 4Kc.

[AB2EZ]

As an update:

I tested this setup (using the toroidal transformer) in more detail this morning. I measured both the output of the toroidal transformer (driving the plate and screen of the 6146 in the Ranger via a modified Heising configuration... all done using the 9-pin connector on the back of the Ranger, with no mods to the Ranger itself), and the output of my off-air monitor.

Results:

Toroid input: a 30-9500 Hz sine wave audio waveform from the 8 ohm output of a 60 watt (rated) audio amplifier.

Toroid output: 100% modulating the B+ line of the Ranger with 400 volts peak (800 volts p-p) of audio.

Frequency response measured at the output of the toroid (i.e., the modulation on the B+ line): flat from 30Hz to 9500 Hz.

Frequency response measured at the output of the off-air monitor (which includes the effects of the Ranger's screen bypass capacitor on the high frequency response of the Ranger): no low frequency rolloff at 30 Hz, 3dB down at 6 kHz, 6dB down at 9500 Hz.

Best regards
Stu

[TedN]

Stu,

Thanks for the R&D!

After trying to dig up plate iron for a power supply, I decided to purchase two plate transformers from Antek. One of the two was non stock but John at Antek said he could have it in three weeks. Four weeks later both transformers arrived. Having some background in the manufacturing of toroidal transformers, they look well constructed. John left the impression that they will wind whatever You want at little or no additional cost, and a short lead time. 

"Amazing... this transformer is 1/10th the cost, 1/10th the weight, and much smaller in physical size than the Hammond transformer."

Unfortunately for Hammond this is true.

Depending on the application, toroid's can be a better design choice. No monkey business with the assembly of E I sections, You toss the core on the winder and build.

BTW toroidal inductors typically have a higher Q.

Anyway thanks for the research. As it seems, supplies for this wonderful hobby are slowly drying up, it's great to find better alternatives.

73's
Ted / KC9LKE

Also don't buy Antek on Epay. I have seen the bid price exceed the list price.
Antek has a website. http://toroid-transformer.com/

[WA1GFZ]

Not all that simple. You have to be sure there is enough voltage isolation between windings and each end of a winding. Wire usually has a breakdown of a couple hundred volts so start can't sit on finish or you get a flash over.

[AB2EZ]

Addendum

Much higher permeability of the core material (vs. traditional core materials) =>

A. Much fewer turns required [i.e., V = # of turns x d/dt (total flux), and, for a given H-field level, the flux is proportional to the permeability]

B  Much less leakage inductance [i.e., the H field is roughly the same inside the core and just outside the core; but, because of the high permeability of the core material, the B-field is much stronger in the core than in the small gap between the core and the wires wrapped around it. Flux is the integral of the B field over the cross-sectional area.  Thus, the higher the permeability of the core material, the higher the percentage of the total flux contained within the core]

Much fewer turns required =>

C. Much more room for spacing between turns, and less layers (or only one layer)

Much more spacing between turns and less (or no) layers =>

D. Much less capacitance to spoil high frequency performance
E. Much more room for thicker insulation on wires
F. No arcing between layers (if there are no layers)*

The limitation is that these transformers cannot support unbalanced DC, because they will magnetically saturate. Therefore, the use of modified Heising.

Stu

*Even the little transformer that I am using is specified to have been hi-pot tested at 3500 volts.

[WA1GFZ]

Stu,
Tape cores will give you the highest perm. I wouldn't bother with iron or ferrite. Magnetics Co. sells tape cores. There is a lot of data on their site. I seem to remember the flux density is similar to an EI core so again there is no free lunch. Stacking 2 cores cuts the turns in 1/2
I would think magnetic path length. The advantage may be the longer magnetic path length will reduce H for a given number of turns allowing more DC offset so a big toroid could make it easier to handle DC offset.

[VE7 Kilohertz]

Stu,

Nice bit of research!! Thanks!  I have some 300VA toroids here, dual 120VAC pri and dual 24V sec. These may also work.

Refresh my memory..modified Heising is where the mod xfmr has a 1-5uF cap in series with the secondary to ground and the mod reactor feeds the B+ ..right?

Cheers

Paul

[Todd, KA1KAQ]

Despite the noise and fading here last night Stu, you sounded fine. Hope to hear you over the weekend for a better opportunity to hear it in action.

[AB2EZ]

Thanks guys... for the kind words of encouragement!

With respect to modified Heising...

Yes, the B+ is fed through a choke (to block the audio from being detoured into the power supply), and the audio is fed through a capacitor (to block the B+).

More specifically, it is important that the impedance of the choke at the lowest audio frequency of interest should be high enough so that it's impedance, in parallel with the plate resistance of the modulated tube (B+/the plate current), and in parallel with the impedance of the secondary of the transformer (associated with the magnetizing inductance of the transformer) should be greater than the effective source impedance of the modulator:

Example:

In my case, I have the following:

Heising choke: 20 Henries rated for 300ma.

Plate resistance: 600 volts/120ma ~ 5000 ohms

Magnetizing inductance (as reflected to the high impedance output of the transformer): estimated from the Antek specification sheet: 230 volts/.01A at 60Hz => 23000 ohms at 60Hz => 61 Henries (provided you don't saturate the transformer with unbalanced current or too much voltage)

Open Circuit Test (core loss test):
TEST CONDITION: Apply variable voltage to primary coil (115V terminals) from 115-140VAC at 60Hz. No load on secondary coils.
1. Primary V = 115VAC, Primary I = .01A or core loss is 1.0Watt.
2. Primary V = 140VAC, Primary I = .01A

Therefore, not yet including the effect of the capacitor (see below), and at the lowest frequency of interest to me... which is 30Hz...  I have 5000 ohms of resistance in parallel with approximately 30 Hz x 2 x pi x 15 Henries (i.e., 20H in parallel with 61H) ohms of reactance =

5000 ohms in parallel with j2827ohms

[Note that at 30Hz, more audio frequency current is flowing into the power supply or to ground through the reactances than is flowing into the plate of the tube. This is not necessarily the end of the world... but it does increase the burden on the audio amplifier as will be calculated next]

The transformer has a step down ratio of 230/12 = 19.2. Therefore the amplifier sees a load impedance that is 1/(19.2 x 19.2)  of the load on the secondary of the transformer. I.e.

Impedance seen by the audio amplifier looking into the 12 volt windings (in parallel) of the transformer =

[5000 ohms in parallel with j2827 ohms]/ (19.2 x 19.2) = 13.6 ohms in parallel with j7.7 ohms.

The audio amplifier that I am using is a modern amplifier that is specified to be capable of driving a load that is as small as 4 ohms. The fact that the load is mostly reactive makes me tempted to put a swamping resistor across the output of the amplifier... but I haven't done that.

Since the output impedance of my amplifier (4 ohms) is lower than the impedance of the load (down to 30Hz), the voltage that my amplifier develops across the load, for a fixed level of audio input to the amplifier, does not roll off for frequencies as low as 30Hz.

This is a well known aspect of the theory and practice of modulation (etc.)... if you drive a load from a relatively low impedance voltage source, the voltage developed across the load will not roll off (with frequency) until the load impedance drops below the source impedance.

From the perspective of the load (including the 6146 in the Ranger), the audio source impedance is 4 ohms x 19.2 x 19.2 = 1469 ohms.

With respect to the capacitor... there are some very enlightening discussions of what its value should be in the literature.

One school of thought is that the capacitance should be large enough to be invisible at the lowest audio frequency of interest. [Note: This will lead to a very low frequency, possibly high Q,  series resonance between the capacitor and the Heising choke (which are effectively in series, and across the transformer's secondary). This could create problems of oscillations at that very low frequency if there is enough of a feedback path at that frequency. An example of such a feedback path (for illustration) might be "talk-back" from the choke or the transformer into the microphone... and through the audio chain... if the audio chain passes such low frequencies.]

Another school of thought, that is particularly applicable to the case where the Heising choke's inductance is equal to the magnetizing inductance of the modulation transformer (which is not the case here), is that you should choose the capacitor to resonate with the Heising reactor at the audio frequency where the reactance of the inductor equals the plate resistance of the tube being modulated (Wow! That's a mouthful!). This will ensure that at the series resonant frequency, the Q of the resonant circuit is low (because of the loading associated with the plate resistance). It will also extend (somewhat) the low frequency rolloff of the total impedance across the secondary of the transformer. It's interesting to note that when the Heising choke's inductance equals the magnetizing inductance of the modulation transformer, the load on the transformer at the resonant frequency will be purely resistive, and equal to the plate resistance of the r.f. circuit being modulated.

If I were to use the above design rule... even though the inductance of my Heising choke (20H) is 3x lower than the magnetizing inductance of my modulation transformer (~60H)... then I would proceed as follows.

The frequency at which the Heising inductor's reactance equals the plate resistance is given by: 20 Henries x 2 x pi x f = 5000 ohms.

Therefore f= 40Hz.

If I wanted the capacitor to resonate with the inductor at f=40Hz, then I would need a capacitor whose impedance at 40 Hz is 5000 ohms. Thus I would need a capacitor whose value is 0.8uF. If I used a value of 1uF, then the resonant frequency would be a little lower... but this would have very little effect on the behavior of the circuit.

All of the measurements I described in my earlier posts in this thread were made with a capacitor having a value of 2.5uF, rather than the 1uF value calculated above.

[Note: I tried this, subsequent to my earlier posts, on January 19, 2008, by using four (4) 4.7uF capacitors in series, with a 100,000 ohm balancing resistor across each capacitor... to get a total value of a little more than 1uF.

The impact of using a 1uF capacitor in this application (where the magnetizing inductance of the transformer is 3x the inductance of the Heising choke) is that the load on the secondary of the transformer looks like the plate resistance (5000 ohms) in parallel with a capacitive reactance (instead of an inductive reactance, as would be the case if the capacitor were chosen to be larger) at frequencies near 40 Hz. The audio power amplifier seems to be happier with a resistive load in parallel with a capacitive load at frequencies near 40Hz... rather than a resistive load in parallel with an inductive load.... but there is some low frequency roll off in the modulation at frequencies between 50Hz and 100Hz that wasn't present when I used a 2.5 uF capacitor. The modulation frequency response is down around 3dB at 50Hz, and down about 6dB at 30Hz when I use a 1uF capacitor. It is essentially flat down to 30Hz when I use a 2.5uF capacitor.]

If you want to minimize the voltage between the capacitor and ground, the capacitor should be placed between the transformer's secondary and ground.

WARNING: The peak voltage between winding of the Heising choke (if it is a single unit, rather than multiple units in series) and its case (if the case is grounded) will equal the peak voltage on the plate of the modulated tube(s)... i.e., more than 2x the B+.

WARNING: The voltage across the capacitor, and also across the secondary of the modulation transformer can be several times the B+, under certain conditions, as described immediately below.

There is an important (to take into account) series resonance effect that will occur in a modified Heising configuration at low audio modulation frequencies. I.e., 60H of transformer magnetizing inductance in series with a 1uF capacitor forms a series resonant circuit at 20 Hz. Therefore, you should assume that the voltage across the capacitor, and the voltage across the secondary of the modulation transformer, may each be several times as large as the modulation being superimposed on the B+ (plus the B+ itself, in the case of the capacitor)... particularly if one tries to modulate the transmitter with a high percentage of modulation at very low audio frequencies. Given the effects of: the plate resistance of modulated the r.f. tube,  the series resistance of the transformer's secondary, plus the loading effect of the audio source across the primary of the modulation transformer (which acts like a load on the secondary of the transformer), and with a suitable bleeder resistor across the capacitor, one can keep the Q of this series resonant circuit low enough to keep the associated voltages within reason. For example, if the Q of the series combination at the series resonant frequency is 5, then (with 100% modulation with a sine wave at that frequency) the peak voltage across the capacitor, when it is series resonating with the magnetizing inductance of the transformer may be more than 5x the B+; and similarly for the voltage across the secondary of the transformer.

Given the above, it is a good idea to not apply a 100% sine wave modulating signal whose frequency is near the series resonant frequency discussed immediately above!

Best regards
Stu

[Steve - WB3HUZ]

Typica values found in 250-1000 watt BC rigs are 40-60 Henry choke and 1-2 uF blocking cap. It's usually better to go with a larger amount of inductance and a smaller value of capacitance to avoid overshoot.

[WA1GFZ]

I notice the Tom Vu 813 put the ckoke across secondary in series with the cap. This could limit the voltage across the cap since it is floating at B+. I'm not smart enough to know which configuration is best.

[AB2EZ]

Steve

Hi!

I agree that if I wanted to build a technology-faithful replica of a traditional broadcast transmitter, that I would follow the design rules that were used by the engineers who built broadcast transmitters in that era. They certainly knew what they were doing!

Some of us are trying to built plate modulated transmitters that are, in a manner of speaking, a hybrid between traditional designs and modern day technologies.

For example, we might be willing to use a modulator that is capable of delivering a lot more audio power than is required to modulate the rf tube(s)... in trade for the ability to use a modulator whose effective output impedance is lower than the plate resistance of the modulated rf stage. Perhaps not cost-effective if you are going to make a lot of these... but its just a hobby. In the grand scheme of things, building a home-brew a plate modulated vacuum tube transmitter serves no business purpose. 

By using a lower impedance modulating source, with plenty of reserve audio power... I don't need to worry about the reactance of my Heising choke being lower than the plate resistance of my Ranger's output stage at 30Hz.

I do agree that if I had a few more 10H 300mA Hammond chokes... I might use four in series, rather than two, just for the 'heck" of it.

As another example, we may want our transmitters to handle (with good linearity and modest roll-off) frequencies that are lower or higher than the frequencies that were handled in traditional broadcast transmitters (for no good reason, other than wanting our transmitters to have that capability).

As another example, we may be using RF decks that have a different plate resistance from that which was used in traditional broadcast transmitters. For example, just for fun, I might want to use a 4-1000A at 6000 volts and 150 mA, instead of a pair of 4-400A's at 2000 volts and 450 mA. I think at 4-1000A is a really neat looking tube.

Therefore, I think it is helpful to go back to the theory, rather than relying on the design rules that were based on assumptions which might not be valid for our objectives (which are to have fun).

Best regards
Stu

[kc2ifr]

Stu,
Thanks a bunch.....as usual u look at  "REAL WORLD" conditions.
The nay sayers are also noted.......
Bill

[KA8WTK]

While doing a little web surfing today I found this article. It talks about micro-gapped toroids that are resistant to DC saturation. ( Idc < 2A)

http://www.cmi-ferrite.com/News/Papers/PCIM1098-1.pdf

Makes you wonder if you could take a large toroid, saw it in half and glue it back together as a core for a modulation transformer.

Bill

[WA1GFZ]

I've seen toroid sold as 1/2 circle so you would band them. Hyperseal is another option and it is made the same just a different shape making it easier to wind a bobbin. Hey guys how about a microwave transformer core???

[Steve - WB3HUZ]

Therefore, I think it is helpful to go back to the theory, rather than relying on the design rules that were based on assumptions which might not be valid for our objectives (which are to have fun).

I was going back to the theory. It shows that using a larger cap produces overshoot. It has little or nothing to do with frequency response requirements.

[WBear2GCR]

I've seen toroid sold as 1/2 circle so you would band them. Hyperseal is another option and it is made the same just a different shape making it easier to wind a bobbin. Hey guys how about a microwave transformer core???

Are you thinking of "C cores"?

Dunno of any toroidal transformer coils that can be cut in half...

I know about "Hypersil" which is a high silicon content alloy...?
Then there is Permalloy, which is high nickel content.






      _-_-bear

[WBear2GCR]

At the risk of raining slightly on the parade...

The experiment that Stu did does appear to work sufficiently well.
However, Stu's experiment was using a very small (low VA) transformer.

My estimation is that it makes it to 5kHz ok, and that is quite good enough for what we are doing. It may go higher in freq, but it did not appear linear higher up watching the output of my IF section on my scope - but then many ham transmitters run out of gas there too! Not a problem.

The limitation on using power toroids as transformers for audio comes when you try to use larger VA ones!

In general the limitation on transformers for audio is from Leakage Inductance, and Interwinding Capacitance. The LF limit is strictly a function of having an adequate number of turns and an adequate amount of inductance. That's usually not a big problem on power iron.

As your transformer size goes up, the two parameters above tend to get worse quickly. I've measured quite a few toroidal power transformers for frequency response, and they are not good. This includes some 125VA size and some 700VA size that I recall clearly.

One can usually judge the quality of a transformer by loading it properly, and then looking for the "bump" in response on the high frequency side. In the case of a good audio output or interstage transformer, one looks for the "bump" to be at 2X the max freq that you want to pass or better. That puts the bump up in the ultrasonic for a high quality transformer, somewhere around > 70kHz.,for a darn good output transformer.

There's no reason not to use a transformer that gets you only as far as the one Stu is using... but I'll be very surprised if larger units will get you past 2500Hz before phase shift sets in, and the response starts to droop.

This is why mod iron is different than power iron. And why mod iron is wound very differently than power iron. And also why building high power mod iron was very difficult and took many years and iterations before good and useful designs were found. In general, the higher the power, the harder it is to make a transformer remain broadbanded and flat out to and past 20kHz. (In this case we're looking for flat and smooth at least to 5kHz.)

For some interesting reading on output transformers, it is worth while to read what is in Radiotron, and also look up Partridge and his original work on designing output iron (similar issues to mod iron) - also there are quite a few neat articles on high power mod iron designs online and in the mags...

Just be cautious about grabbing higher VA power transformers and assuming that they will work as well as the little one in Stu's experiment.

                  _-_-WBear2GCR

[WA1GFZ]

If you cut a tape core in half you will have a mess but I have seen them many times come in two pieces. I've used cores like that to induce transients on a long cable. Audio transformers are an art.

[AB2EZ]

Bear

Hi!

Re-measured this morning using a stand-alone audio sine wave generator as input to the audio "power" amplifier that drives the ferrite transformer:

Test conditions:

Capacitance in series with the transformer: 2.5 uF

Audio signal source: A sinusoidal audio output signal from a GW GAG-810 audio frequency signal generator is used as input to a Samson Servo 120a audio power amplifier (rated 60 watts output into 8 ohms). The output of this audio frequency signal generator is displayed as the 1st trace on a dual-trace scope

Input-to-output chain: Audio output from GW GAG-810 audio frequency signal generator => input of Samson Servo 120a => output of Samson Servo 120a => 12 volt windings (in parallel) of 25VA toroidal transformer => 230 volt output winding (two 115 volt windings in series) of 25 VA toroidal transformer => modified Heising input of Johnson Ranger (Heising choke = 20H rated for 300 mA, coupling capacitor = 2.5 uF) => >90% modulated RF => REA off-air pickup/detector => 2nd input of dual-trace scope, in parallel with REA modulation monitor input.

Scope settings: Both inputs are AC-coupled. Scope is a 100MHz Tektronix TDS3012B. Traces are amplitude-overlayed when 1kHz audio is applied... by adjusting the fine scale setting of trace 2. Then, no further adjustments of the trace scale settings are made. The sweep rate is adjusted to accomodate (display) the audio frequency being used. Triggering is from trace 1. There is a time offset (delay) between trace 1 and trace 2 at observed low frequencies, corresponding to phase shift.The time offset at 1kHz is negligible.

Results:

From 25Hz to roughly 6500 Hz, the amplitudes of the traces continue to track within a few percent of variation. The audio frequency was adjusted smoothly, and there were no bumps in the tracking.

The output deviates from the input by less than 3dB (less than a 70% reduction in output amplitude) at 10kHz

There is about 30 degrees of offset between the input and the output at 50Hz, with the output sine wave leading the input sine wave.

I also applied a 125Hz square wave audio input signal... and I observed a droop decay time of 4 milliseconds (to the 1/e point) on the output waveform of the REA pickup/detector (trace 2). Note that the input waveform (trace 1) had a droop decay time of 12 milliseconds.


Best regards
Stu