Is it really "better" to use the whole secondary of your OT?

[martin manning]

I thought I'd try measuring the frequency response at each tap using a Pacific OT I have here. I'm feeding 5V into the primary, and connecting the appropriate resistive load to each of the 4, 8, and 16 ohm taps. The bandwidth is indeed wider on the 16 ohm, so this transformer is better (more Hi-Fi) when using the whole secondary. The other taps are still fine for guitar when you consider that speakers usually drop off above 5 kHz, and maybe that high frequency response isn't something you want anyway. The roll-off is a little sooner on the 8 than the 4, which is not what I expected.

Note that the response in dB is different in each case because the voltage ratio is different for each tap. With a little math you can work back to a nominal primary impedance, which comes out to be 4569, 5044. and 4841 on the 16, 8, and 4 respectively. That might affect the sound a bit too.

Looks like different combinations of speaker cab and OT tap is yet another variable to play with.

[pdf64]

What source (supply) impedance are you testing it with? eg 50 or 600ohms off the sig gen, something roughly equivalent to the primary impedance, eg 4k7, or something roughly equivalent to the likely anode impedance, eg 22k?
Hammond tend to use a source impedance equal to the primary impedance.

[martin manning]

The generator has two options, 50 ohm and "High-Z," but it is stated that the generator actually has a fixed 50 ohm output. I used High-Z. I can't find any more details on what High-Z mode is, but in that mode the amplitude drops by half with a 47 ohm load, confirming a 50 ohm output. With a 4k7 load, the amplitude does not drop. In any event, the frequency response analysis uses a ratio Vout/Vin, where both the input and the output are measured.

[wpaulvogel]

It depends on how the transformer is wound. Most of the output transformers made for guitar are wound with multiple taps that are basically like a long wire with a tap at the halfway point for the 4 ohm output, the 8 ohm output is next and it’s tap point is the length of the 4 ohm times the square root of 2 and the 16 ohm output is the full length. It’s not quite this simple because of the interleaving makes breaks in the length and these sections are paced between the primary sections. A Marshall Drake 50 watt output transformer has 3 primary coils and 2 secondary coils. By using only the 4 ohm output only half of the secondary winding is reacting to the induction of current. Because the interleaving pattern, it’s in both interleaves but only half the turns. Some patterns are different and the shorter secondaries aren’t using all the interleaving.
Some output transformers have the secondary separated and use a series connection for the 16 ohm output and parallel the two sections to achieve the 4 ohm output. In this scenario, the whole secondary is carrying current and therefore it should have equal performance for the most part. The 8 ohm output would suffer slightly because it’s .707 of the total secondary but I’m not aware of a strategy to use the total secondary with this impedance option.
A Marshall Dagnall C1998 100 watt transformer has a parallel 16 ohm winding that’s a completely different section and it establishes another interleave that if left out would definitely effect performance.
This is the problem with multi tap output transformers. The option of several impedance matches with a simple rotary switch costs performance for convenience and is the reason I have speaker cabinet arrangements that I can connect to achieve all three different impedance levels depending on what needs to be matched. It’s my opinion that using the entire secondary is the best option for optimal performance no matter how it must be done. Any other use would seem similar to having a racing vehicle that the throttle wasn’t adjusted properly and never achieved full opening.

[martin manning]

It depends on how the transformer is wound...

I was careful to say this transformer above for the reasons you state. It'd be interesting to try this measurement on a RS or Hammond with complex secondary connections for different impedances.

[wpaulvogel]

It depends on how the transformer is wound...

I was careful to say this transformer above for the reasons you state. It'd be interesting to try this measurement on a RS or Hammond with complex secondary connections for different impedances.

I would like to see the results also because I think they would perform flawlessly. The Schumacher transformers Fender used with single impedance are interleaved and work great.

[bepone]

old known story in this thread. interleaving doesnt metter for this, and ofcourse it is more expensive to wound "better" transfomer which no one for guitar use is. they are all wild. and it is ok like this, this creates the sound so no worries.

bigger problem is ringing on the square wave when not used at full secondary, but this also doesnt matter for guitar use at all...all is good

[bepone]

for the sound crucial are core material, wild winding technique, certain interlayer insulation materials and insulation varnish ,all special requirements and are not often found in guitar transformers because they increase price even several times !

[pdf64]

Using the full secondary is not simple if, as is the case with many vintage UK valve guitar amps, there’s a 100V output available (on my Sound City MkIV+ it works out to an intended load of 200ohm). Similarly for 70V outputs.
I suspect those windings may be poorly coupled anyway.

Poorly coupled / non interleaved windings might cause stability issues when they’re included in global feedback loops.

[R.G.]

It's a crap shoot, in the sense that the frequency response of the OT depends on the first-approximation lumped leakage inductance/wire resistance/self-capacitance of the winding sub-sections shading to the second approximation distributed RLC of each section relative to all the other section, and on down to the position of every turn relative to every other turn and to the iron and whether the iron is grounded or not.
It gets complicated fast. Well, OK, it's not random, it's very complex, and generalizations from one transformer may not hold for another that's differently wound. For some background, read the articles on the "Williamson" OT and the unity-coupled MacIntosh ones when they were introduced.
In large generalization, sure, using the entire secondary would be better, if the OT was wound with an interleaving (i.e. interleaved at all!) and whether that interleaving was carefully spread with not only secondary sections between balanced primary sections but also balanced sub sections of secondary between balanced sub sections of primary, so that the coupling between primary and secondary was good and the coupling from half-primary to half primary and half-secondary to half secondary was good. The more that mutual and cross coupling is good, the less it ought to vary with using partial secondaries.
It's tempting to make a generalization about using all the secondary. Sure, probably not a bad idea. But it may not be a momentous result, as the other posts note. In non-class-A push-pull setups, each different set of secondary sections will have different coupling to the primary half-sections, so the distortion product mix will be different. This is the problem the unity-coupled MacIntosh design was trying to fix.

[Stevem]

Maybe we can all atleast agree that if your running a output stage hard that is better to make use of the heatsink effect provided by a longer length of wire?

[R.G.]

Maybe we can all atleast agree that if your running a output stage hard that is better to make use of the heatsink effect provided by a longer length of wire?

Yes, kind of. It's not so much heatsink effect as it is spreading out the generated wire resistance heating over a bigger volume to eliminate hot spots. The real heat sinking is done by the copper wire heat transferring through the impregnating material between wires and then to the next wire, and so on out to the outer wire layers and iron. Voids inside the windings set up uneven heating. But it's a small effect, as inactive copper will still conduct heat, it just doesn't generate its own.
The start of a winding design assumes a given power dissipation per unit volume inside wire coil volume and core volume, then proceeds to muck around with wire gauge, layers, current density and core flux density to make that come true-ish, as nearly as possible.
Actually, the big issue with transformer design is usually the temperature rating of the wire insulation and how hot you want to let it get. Iron and copper don't mind heat up to the curie point of the iron - if the insulation will take it, the transformer will be fine. The insulation is what gives up first.

But (I can hear you say) what does that do to the frequency response/tone? Not a thing. That's still all about what the inductance/capacitance of the placement of each wire relative to the others happens to be.

[trobbins]

If the amp uses global feedback, and one has the test equipment to determine frequency response and phase shift, then it may be worthwhile checking terminal connection options as internal stray capacitances can have quite an influence on higher frequency resonances - not typically a concern for guitar amps with low levels of feedback, but could be noticeable if speaker loading or presonance type feedback led to reduced stability margins. One example can be transformers where secondary side windings can be connected in series or parallel for different speaker impedances, or where connecting one end of a winding to ground (versus the other end) can show up a difference (eg. due to the capacitance effect of the unused portion of winding).