OT... Source impedance?

[Chuck H]

I'm working on a new attenuator with independant bass, mid and treble controls. To choose the cap and inductor values I need to "average" a guitar amp OT's source impedance to figure the frequency knee targets. Would it simply be the DCR on the secondary or should I assume 0 as an ideal constant current source? Are ther other considerations?

With 4/8/16 being the load, of course.

Thanks

Chuck

[pdf64]

I think that in that (excellent) book download of teemuk's solid state guitar amp design, the output impedance of a TR is measured at a bit over 1 ohm.
I'll see if I can find the pdf and page.

[R.G.]

I'm working on a new attenuator with independant bass, mid and treble controls. To choose the cap and inductor values I need to "average" a guitar amp OT's source impedance to figure the frequency knee targets. Would it simply be the DCR on the secondary or should I assume 0 as an ideal constant current source? Are ther other considerations?

With 4/8/16 being the load, of course.

Transformers as such don't really have impedances - they have ratios.

It helps a lot to get comfortable with the standard transformer model. Here's the first good illustration I ran across.

It is handy to think in terms of a series of approximations. To a first approximation, the transformer is ideal, not resistances, inductances or capacitances, and for a turns ratio of Np:Ns, has Vs = Vp* Ns/Np, Is = Np/Ns, and Zs (the load impedance on the secondary external to the transformer) is presented at the primary looking like Zp = (Np^2/Ns^2)*Zs ( "^2" means 'to the second power').

To a second approximation, the idea transformer is the same, but the wiring DC resistances are in series with the primary or secondary, or better yet, referred to the primary entirely as Rp' = Rp + (Zp/Zs)*Rs.

To a third approximation, you'd add to that the leakage inductances, Lp and Ls. Beyond that, you're designing signal transformers, of which output transformers are a special case, and you need to take into account distributed and primary-to-secondary capacitances.

For your purposes in designing an attenuator, you probably don't need to get beyond a second approximation. It won't make nearly as much difference as the speaker itself or tubes themselves make.

[Chuck H]

Thank you R.G. I still don't understand how to find the OT's secondary source impedance. I really don't have those kind of parameters on any guitar amp OT's anyhow. I've seen the "ideal transformer" model explained in a book I own but I never paid much attention since I never had numbers to plug in for the parameters. And then there's the math

To say an OT has an 8 ohm secondary (or 4 or 16) to my understanding indicates the load that should be presented to it, not the source impedance from the secondary itself. This is my confusion. On models I've done using a 0 ohm source and an 8 ohm load changing the source impedance to even a few ohms shifts all my frequency knees dramatically. So I'm trying to find a reasonable source impedance for my model.

Chuck

[octal]

Well, if my understanding is correct, the source impedance is going to vary depending on the damping factor of the amp, transformer ratio, etc. Is this attenuator going to be used with one particular amp? If so, I THINK you can measure the source impedance of the amp by measuring how much the output voltage drops under a load. For example, feed a sine wave generator in with an 8 ohm dummy load on the output, measure the voltage across the resistor. Add a higher value resistor in parallel (say a 16ohm) and measure the (now lower) output voltage. Imagine an internal series resistor connected in series with the output of the "perfect" amp and solve for its value using ohm's law.

This is all speculation on my part; I'm ready to be set straight by someone more knowledgeable!


Nathan

[pdf64]

Another factor would be whether there's global NFB on the power amp - the open loop output impedance will be reduced according to the feedback ratio. But when the amp reaches clipping that goes out of the window, leaving you with 2 different models!

[R.G.]

I still don't understand how to find the OT's secondary source impedance. I really don't have those kind of parameters on any guitar amp OT's anyhow. I've seen the "ideal transformer" model explained in a book I own but I never paid much attention since I never had numbers to plug in for the parameters. And then there's the math

Unfortunately, there isn't much way to make it simpler. If you want the apparent impedance of the amplifier including the output stage and OT, as seen by an attenuator looking back into the secondary output windings of the OT, you will see

(1) the transformed plate impedances; pentode and power beam tube (i.e. 6L6 type) plates themselves are pretty decent constant current sources. This is in series with the internal equivalent impedances of the transformer. For your purposes, if I were doing this I'd refer everything to the secondary. A 4400 to 8 ohm OT has a voltage ratio of 1:24 (secondary voltage is 1/24th of the primary swing) and a current ratio of 24:1 (secondary current swing is 24x the primary swing). You get this by taking the square root of 4400/8 = 23.452...; I rounded this to 24. So the source impedance contribution of the tubes is the impedance of a 6L6 (or EL34, or...) times 8/4400 (about 1/550). But then you have to know the plate impedance of a 6L6.
(2) to that, you add the series resistances and leakage inductances of the OT itself. A 6L6 OT might have a primary wiring resistance of 300 ohms (about!) and the 8 ohm secondary might be 0.1 ohm. So the transferred 300 ohms is 300*(8/4400)= 0.545, and you add the (guessed at) 0.1 ohm to that to get 0.645... ohms resistive. This is in series with the transformed plate impedance of the 6L6s, which is ... well, high.

The Sylvania tube handbook lists the plate resistance of a 6L6GC as 25 to 35K in a single ended application, and notes "approx", and refuses to list one in push-pull Class A or AB. Great. Thanks, Sylvania. But we can forge ahead. If we assume the rp is 33K, and transform it, get we get rp(sec) = 33K* (8/4400) = 60. That is, the transformed plate resistance looks like about 60 ohms to the secondary at the output terminals.

This is so high you can neglect the 0.645 ohms as trivial. Leakage inductances of a few millihenries get treated the same as the wiring resistance, but they are frequency dependent. They get up in the range of the 6L6 plates in the highest octave of audio (10k to 20kHz.)
(3) to this, you add in parallel the transformed (8/4400) primary inductance (10H and up) and eddy current losses (resistors, but frequency dependent ones). There's nothing for it but to measure these.

This is all headed toward saying this: I know *how* to measure and then calculate the properties, including the OT's inner stuff, but the answer is different for each OT, and it is swamped by the series impedance of the output tubes and the parallel impedances of any speakers. So you have to have a unique answer for each OT, or at least each model of OT. They're different in the things which you'd measure, and then what you measure may not have the biggest effect.

Or, more accurately, you'd note that the OT itself can be ignored in the face of the 50-70 ohms the tube plates look like. I'd do that if I were doing what you are doing, and just bang in 60 ohms.

To say an OT has an 8 ohm secondary (or 4 or 16) to my understanding indicates the load that should be presented to it, not the source impedance from the secondary itself.

That is correct. Or, put another way, if you put 8 (or 4 or 16) on those terminals, they will load the tube plates with the nominally correct about-4K that a pair of power beam tubes need to give best power.

This is my confusion. On models I've done using a 0 ohm source and an 8 ohm load changing the source impedance to even a few ohms shifts all my frequency knees dramatically. So I'm trying to find a reasonable source impedance for my model.

Put in 60 ohms. What you actually see is mostly the transformed plates. This fits with the idea that the transformer itself should not have a huge effect. It is designed not to be the most prominent thing there. Good ones mostly achieve this.

[Chuck H]

+++

Thanks

I do see the picture now. I am surprised at 60ish ohms though. That seems really inefficient into, say, an 8 ohm speaker load. I will plug that in and see how it goes.

I did find an article that discussed current drive as opposed to voltage drive transducers. There was a brief discussion about source impedance as it relates to damping. It was stated indicated most valve amp output transformer secondaries of old had a source impedance of between half and twice the speaker load. So this would be in contrast to your figure, but they didn't back their number with any process either.

Chuck

[octal]

Going back to what I mentioned earlier, why not actually measure it? Yes, it will vary from amp to amp, but if it's a question of "60 ohms" (R.G's answer) to "half to twice the speaker load" (2 to 32 ohms, depending on a speaker load ranging from 4-16 ohms) that's quite a difference. I measured the damping factor on a couple of amps quite a while ago by injecting a drive signal into the secondary while the amp was on... I might be able to find my results somewhere in my notebook, but the damping factor was pretty appallingly bad in my no NFB designs.

Nathan

[R.G.]

I did find an article that discussed current drive as opposed to voltage drive transducers. There was a brief discussion about source impedance as it relates to damping. It was stated indicated most valve amp output transformer secondaries of old had a source impedance of between half and twice the speaker load. So this would be in contrast to your figure, but they didn't back their number with any process either.

It is inefficient. Tubes get HOT. Notice that I hand-waved you some. The data book never said what Rp was in a push-pull amplifier ( I did at least mention that... 8-) )and so my quickie analysis is something of a guess. For damping purposes, the speaker sees a lower driving impedance than that because the power amp uses negative feedback and that lowers the closed-loop output impedance. The tubes still handle the same voltage, current, and power, but feedback "precompensates" for some of the sag that would otherwise happen, and makes the output look like it's a lower impedance than it is. That's one more complication to your calculations.

For you to take away from this, the OT's impedance itself is the least of your worries. The tubes, the speaker, negative feedback, these all add hugely to the problem. Particularly when the speaker load is a big effect, and you're trying to fake the speaker load.

[R.G.]

Going back to what I mentioned earlier, why not actually measure it? Yes, it will vary from amp to amp, but if it's a question of "60 ohms" (R.G's answer) to "half to twice the speaker load" (2 to 32 ohms, depending on a speaker load ranging from 4-16 ohms) that's quite a difference. I measured the damping factor on a couple of amps quite a while ago by injecting a drive signal into the secondary while the amp was on... I might be able to find my results somewhere in my notebook, but the damping factor was pretty appallingly bad in my no NFB designs.

It for sure is a big difference. And yes, damping is appallingly bad in no-FB designs. And these things vary all over the map. That's one reason to go active, and break the dependence on the finagling of the OT, feedback and speaker loading. But that's just me. 8-)

[Chuck H]

Why not just measure it??? I may, but I do prefere to build things that can be used with gear other than my own (though I have made exceptions). So I'm aiming for for a happy medium or possibly the least bad choice for any amp.

It turns out that on my model changing from a 0 source impedance to, say 10 ohms, makes a huge difference in the knee frequencies. But the difference between 10 and 100 is more subtle. So I think I'm good from here.

Perhaps the reason the article I read on current drive transducers sited a lower impedance than your figure R.G. is that high end tube home audio gear of old typically used comparably high amounts of NFB and also often ran pentodes in triode or UL operation.

I also agree that simply running the amp into a load and building a mosfet amp with EQ to re-amp the signal would be the most efficient and cost effective approach. I've wondered why the "Ultimate Attenuator" never did this since it's a re-amp device. I've used the Ultimate Attenuator and liked it very much. My reasons for going passive are simply that I know some players don't want the power going into their speakers to come from anything but a tube AND I won't have to plug my attenuator in to make it work. Perhaps silly reasons for my approach, but it's OK with me if it's OK with you.

Thanks a lot guys.

Chuck

[R.G.]

My reasons for going passive are simply that I know some players don't want the power going into their speakers to come from anything but a tube AND I won't have to plug my attenuator in to make it work. Perhaps silly reasons for my approach, but it's OK with me if it's OK with you.

I have no gripe with your going passive for whatever reason. I just wanted to be sure it had come up somewhere in your thinking, even if it got discarded. Depending on what you had in mind, it might be easier, more accurate, whatever.

But not having/wanting a power supply for it is a powerful argument for going passive. Go for it.

[J M Fahey]

Although it *can* be calculated, no doubt about it, not only the math gets complex, but, much much worse, you do not have the parameters to plug into those equations.
I think you *must* measure it.
Not under ideal conditions but under real ones.
When I switched from tubes to SS in '72 (long story), i found that there were a zillion differences in sound.
*One* of the differences I found was that *my* 2x6L6 amp had an internal impedance of around 4 ohms (with NFB) and 16 to 20 without, when driving a 4 ohm load.
In fact I independently invented what was later known as "Valvestate" some 15 years earlier, out of sheer necessity.
Fact is , all my SS amps have damping "1".
And how did I find that?: by driving the amp and lifting the load, measuring when loaded and unloaded or going to 8 or 16 ohms.
The math becomes very simple. You know both voltage values and the external load; it's very easy to calculate the internal impedance, for your *real* amp, tubes, transformer.
Try that.

[R.G.]

Rod Elliot is busy re-documenting a lot of basic electronics at his web site, sound.westhost.com. Here's his take on variable daming in amplifiers: Variable Impedance Amplifiers.

I've never been through valvestate in detail, but I suspect it's something similar. Adding a softclip limiter in front of the amp (see Thomas Organ Vox amps, or Elliott's own "soft clip" article) is a useful thing to think about. Using opamp techniques, you can mimic much of the external effects of tube power amps. It probably takes an infinite number of refinements to get them all. 8-)