Transformers

[R.G.]

Yes, excitation current is the same as magnetizing current; using the Hammond data sheet was a great way to deal with both magnetizing current and primary inductance.

Primary inductance and magnetizing current are the two ways of looking at the question of whether you got enough turns on the primary for the core area, window area, and stack of core laminations. They're also some of the least-well-defined things. Core inductance and its dual magnetizing current vary a lot with things like how closely the laminations fit together when stacked into the core winding and the Es and Is fitted and jogged together. Careful hand stacking and gentle straightening and jogging together before cranking down the lamination bolts can vary these two things quite a bit, as not decreasing the air space between Es and Is to a minimum makes more of the magnetic path be air. Hammond has played it a little cagey in specifying the easily measurable magnetizing current rather than inductance. I back-calculated the primary inductance to hold the current down to the specified level. I believe that most of the Hammond 169VS units you could actually buy will have a lower magnetizing current than the spec.

A bit of the underlying magic/mythos is lurking here. For the same core and windings, a higher magnetizing current causes higher resistance losses in the primary wire, but lower excitation losses in the iron, as the air gaps cause lower flux density in the iron, and hence smaller BH loops. A loose core is also a little more resistant to DC offsets in the primary voltage. This is why toroids, with no discrete air gaps at all, are so sensitive to any DC at all causing the core magnetic field to walk over to the edge of saturation.

I'm planning to hop to a model of a Bassman replacement power transformer for the next couple of installments, introducing the difference between driving resistive loads and rectifier loads. Any special requests?

[martin manning]

I'm planning to hop to a model of a Bassman replacement power transformer for the next couple of installments, introducing the difference between driving resistive loads and rectifier loads.

This sounds very interesting; looking forward to it. Thanks!

Using only what has been discussed above, I knocked together a model for the above isolation transformer example and two others using information from Hammond data sheets. These models are just coupled inductors with series resistances. The inductances were determined by simple calculations as shown, with no tweaking. The results are impressively accurate at both open circuit and fully-loaded conditions, except for the 290GX/JCM800. That model does not match the claimed "typical" loaded condition, and note that the loaded voltage quoted is only 2% lower than the unloaded voltage on that winding. A max current of 213 mA is listed, but there is no voltage listed for that load.

Magnetizing current is the key starting point and, as mentioned above, it is easily measured. Knowing that, and the DC resistances, a model of any unknown transformer could be made and used to make estimates for the current capacity of its windings by assuming some percentage of voltage drop in a loaded state.

[Helmholtz]

Ultimately transformer load is limited by the internal hot spot wire temperature and the temperature class of the wire insulation.
While average wire temperature can be estimated from the DCR increase, this method won't give reliable information about the hot spot temperature.
We used to prepare (or had our suppliers prepare) evaluation transformer samples with embedded thermocouples.

Edit: My comment is not intended to derail R.G.s excellent transformer tutorial.

[pdf64]

... Using only what has been discussed above, I have knocked together a model for the above isolation transformer example and two others using information from Hammond data sheets. ...

Just to note that with the Hammond data, the primary voltage and resistance seem probably to be nominal (ie tolerances will be +/-x%), whereas the Iex is an absolute upper limit, so its average value must be lower.

[R.G.]

Actually, your comment is absolutely correct - the limitations on transformer power handling are hot spot temperatures burning out the insulation, not so much wire resistance, although that certainly contributes to temperature rise.
I have deliberately avoided mentioning how the max power handling on a transformer might be designed or evaluated. I've done this in the same interest as not involving a lot of math. The temperature rise is - as you well know! - a complex interaction of the winding, the insulation types and ratings, the amount of impregnation or lack of it, and the air flow over the outside surface, and any conduction cooling that might happen. Instead, I cheated by picking an example that already had the max loading defined well enough.
I've had some issues with writing up the Bassman example, not least because the actual DC current load maximum in the various Bassman models is ill defined. I was hoping to use the Bassman trannie as an example of the difference between the RMS current rating in the transformer for a resistive load versus a full wave rectified DC load with an average value. Confusingly (and accursedly, from my standpoint) the Hammond Bassman I picked has almost exactly equal DC currents and transformer RMS currents when used with a modeled 5U4 rectifier and a 35uF first capacitor.

[R.G.]

[...] except for the 290GX/JCM800. That model does not match the claimed "typical" loaded condition, and note that the loaded voltage quoted is only 2% lower than the unloaded voltage on that winding. A max current of 210 mA is listed, but there is no voltage listed for that load.

2% "regulation" (actually equals no load to full load sag) is getting into the realm of truly big transformers. I'd expect more.

[martin manning]

... Using only what has been discussed above, I have knocked together a model for the above isolation transformer example and two others using information from Hammond data sheets. ...

Just to note that with the Hammond data, the primary voltage and resistance seem probably to be nominal (ie tolerances will be +/-x%), whereas the Iex is an absolute upper limit, so its average value must be lower.

Yes, RG said as much above. Since the secondary inductances are calculated to match the no-load voltages, it won't make any significant difference in the results shown if say a 10% lower Iex is used.

[martin manning]

[...] except for the 290GX/JCM800. That model does not match the claimed "typical" loaded condition, and note that the loaded voltage quoted is only 2% lower than the unloaded voltage on that winding. A max current of 210 mA is listed, but there is no voltage listed for that load.

2% "regulation" (actually equals no load to full load sag) is getting into the realm of truly big transformers. I'd expect more.

The winding resistance seems high in that one too. If I increase the load to the stated 213 mA Max, the loaded voltage drops to 0.921 x the no-load voltage, so 7.9% regulation. The other two are 4.6% and 4.9%.

PS I've added a line to the results for each of the simple transformer models above to show the Iex at nominal input voltage. It is essentially the same as the data sheet value using just the simple equation shown to calculate Lpri.

[LOUDthud]

I noticed that the Hammond isolation transformer at the start of this thread is only specified at 60Hz. What limitations would it need to be observed if it was operated at 50Hz ?

When a guitar amplifier is producing a rail to rail square wave into a resistive dummy load, the power transformer is typically being operated beyond it's VA rating. How far beyond it's rating is safe ? Isn't the VA consumption somewhat lower when the amp is operated into a speaker load ?

[R.G.]

I noticed that the Hammond isolation transformer at the start of this thread is only specified at 60Hz. What limitations would it need to be observed if it was operated at 50Hz ?

The core mA Aagnetizes up to a level that's some constant times the volt-time integral on the primary winding. Using it at 50Hz means that the primary goes positive, then negative, by 60/50 as much time. In an ideal world where we're trying to take it easy on the transformer, we would reduce the primary voltage by 50/60 and this would produce exactly the same flux density in the core as with the specified voltage at 60Hz.
Of course, this is not an ideal world. What would likely happen is that it would just be hooked up to the same (115Vac as I remember) voltage at 50Hz. The magnetizing current would likely increase more than the 60/50 ratio of frequencies, as the primary winding turns and core area are probably designed to limit core flux so the maximum magnetizing current spec is made good. Put more simply, magnetizing current would increase more than the change in frequencies would suggest.
This is the result of economics. To get the best cost of materials, the transformer designer is given raises and praise for designing the smallest, lightest, cheapest cost of materials design he/she/the/they/em/ed/etc can which does not result in too many warranty claims. So for power transformers, the idea is to work the iron to the biggest extent possible.
A 50/60 Hz transformer is really designed for 50Hz, and so it works fine at 60 and has more voltage margin and probably smaller magnetizing current.

When a guitar amplifier is producing a rail to rail square wave into a resistive dummy load, the power transformer is typically being operated beyond it's VA rating. How far beyond it's rating is safe ? Isn't the VA consumption somewhat lower when the amp is operated into a speaker load ?

Yikes!! Can of worms alert!!!
Significant size/weight OTs have enough mass that they will have thermal time constants in the one-or-more-hours range for a hot spot to fully develop. Music typically has a crest factor - peak to average power - of 10db or more, so the long term average of the musical set is what determines the VA rating. True, playing pure square waves into an OT is tough duty, but it has to go on long enough for the thermal time constant to let the hot spots get hot. So I don't know that just a rail to rail square wave necessarily relates to it being over its VA limit. And there really isn't any such thing as an instantaneous VA rating. It's a time-averaging thing.

Certainly the volt-time integral for square wave drive on a transformer is more demanding than a sine wave of the same peak value. So an OT designer is back up against the wall of using the least possible iron, but with the complication that &designer has to now guess about the lowest frequencies and voltages, as well as music crest factor. And what if the amp's owner decides to play bass through it for a few sets?
With that as background, transformers are rated in Volt-Amps (VA) precisely because they are often used on non-resistive loads. VA is just that, VA. What really changes is that a resistor is non-varying in impedance with frequency and speakers are primarily inductive above their low frequency resonance. This is true even when the resistive air loading is taken into account. Perhaps more importantly, a square wave is a combination of a fundamental sine wave and many harmonics. The inductive impedance rises with frequency so the current in the partials is steadily decreasing with frequency. No surprise here - a series inductor is a low pass filter. So the actual amps in "Volt-amps" is a moderately demanding computation.

I think that the difference in resistor loads and real speaker loads is probably explained by the increasing frequencies of the harmonics that make up a square wave drive. And the heating difference in the transformer is mostly the resulting phase shifted/filtered result of the speaker's complex impedance. And I also think that designers of guitar OTs probably shave the iron and turns as much as they can just like the power transformer guys do, and don't necessarily sweat VA ratings. They probably build prototypes and run them in test jigs to see how hot they get how soon. Have you seen guitar OTs with real VA ratings? I can't remember any - only "good for X watts speaker output when run from Y plate to plate impedance. To be fair, I would probably give guitar OT designers a pass. They have an even less well-defined set of requirements than power transformer designers.

[LOUDthud]

When a guitar amplifier is producing a rail to rail square wave into a resistive dummy load, the power transformer is typically being operated beyond it's VA rating. How far beyond it's rating is safe ? Isn't the VA consumption somewhat lower when the amp is operated into a speaker load ?

I was thinking about a Power Transformer. A 100W class AB power amp draws about 140W from it's B+ when putting out a 100W sine wave IIRC. When the same amp is putting out a rail to rail square wave, it's drawing about 200W from B+. If the power supply can safely supply this 200W with a low to medium temperature raise, the power transformer is going to run cool and last (almost) forever.

In step the bean counters. Do you know of any examples of how small a transformer could be and survive in the real world or any that were too small and suffered from a high failure rate. Fender amps even back in the tweed era seem to have a low failure rate.

[R.G.]

I was thinking about a Power Transformer.

Oh, sure, demand that I read ALL the words and get them right.
Curses! I knew that speed reading course would get me into trouble one day.

A 100W class AB power amp draws about 140W from it's B+ when putting out a 100W sine wave IIRC. When the same amp is putting out a rail to rail square wave, it's drawing about 200W from B+. If the power supply can safely supply this 200W with a low to medium temperature raise, the power transformer is going to run cool and last (almost) forever.

OK, I'm with you now. Yes, sine wave output power is smaller than square wave output power. For the same peak, a square wave has 1.414(ish) times the RMS voltage. Into a resistor, this will deliver twice the RMS power (as the current is 1.414 times as big too). An amp rated for 100W RMS continuous with a full rail sine output will need 200W with a rail to rail square wave into a resistor. So it's even worse. A square wave into a real speaker will use less real power as the load is not converting as much of the drive waveform into in-phase current, so less heat comes out.
I'm trying to recall the theoretical efficiency of an AB amplifier. I know for solid state outputs, the theoretical is about 79% in power used versus power out; the output stage itself dissipates power higher than the output load, 30% more if I remember correctly and tube output stages have similar efficiency numbers. Yeah, the PT sees a big increase in delivered power.

This gets back to the question - what did the designer design for? Continuous duty, what the tube databooks and ARRL handbooks called Continuous Commercial Service? Probably not, as this would get the designer fired for spending too much on transformer parts. A smart designer would gather historical data from earlier model amps with similar power ratings and do some lab work on measuring actual VA and temperature rise for simulated test conditions. I have not seen guitar > power transformers < rated for VA either.

In step the bean counters. Do you know of any examples of how small a transformer could be and survive in the real world or any that were too small and suffered from a high failure rate. Fender amps even back in the tweed era seem to have a low failure rate.

It's an easy answer but not a satisfying one. For scrapless EI laminations with a core stack of between 1:2 and 2:1, VA capability is about proportional to the finished transformer's weight/mass, given that scrapless EI forces area products to be similar if the winding window is wound to about 80-90% full and the insulation classes are the same. Our component qualification lab did the correct testing a lot: take one, load it to what you think is the desired VA out and do a temperature rise per time test to determine the thermal time constant and probable target temperature. This matters because the same design, same stack, number of turns, impregnation, and so on will have two very different VA ratings if one is done is Class A/105C insulation materials and the other is Class B/130 or even Class H/180. Changing the insulation class has a profound effect on how hot you can run the hot spot, and therefore on what the VA rating would be. Non-identical insulation classes means non identical VA ratings even if everything else is the same.
So the simplest thing to do is to instrument an amp and measure the required loading, maybe even run temperature testing to find the thermal time constant and eventual temperature rise. If that temperature rise is right at or over the insulation rating, the transformer will have shorter life and more percentage field failures.

In Fender's early years, the vast majority of transformers were wound with Class A/105C insulation. The last three or four decades have seen a rise in availability and lower cost for Class B/130 insulation materials. This makes it very hard to say that smaller is worse. It's certain that smaller is hotter, but that might be OK with better insulation. Bottom line is probably that I'm striking out on your real question; no hard examples.