18w OT

[TIMBO]

Hi guys, I've got a Classic Tone 18w marshall OT that has a primary impedance of 10.8k@4 ohms and 9.2k@8 ohms. Am I able to sub the EL84s for 6V6gts. Thanks

[jazbo8]

Both should work, you may get a bit more output with the 9.2k/8 Ohm tap.

[HotBluePlates]

... a primary impedance of 10.8k@4 ohms and 9.2k@8 ohms. ...

... more output with the 9.2k/8 Ohm tap.

... could you explain how the 8 ohm tap will provide more output? ...

You're focusing on half of what Jazbo said; unfortunately, it's the wrong half. The important part is the change from 9.2kΩ to 10.8kΩ on the primary.

Look at the graph below. It shows the changes in power output (and distortion) for a single 6V6 when only the OT primary impedance is changed (the load impedance the tube actually sees). You will see there is a broad maximum between about 6-8kΩ, and power output falls at higher and lower load impedances.

True, this chart is for a single 6V6, but for class A push-pull, the load giving maximum power output is a value of plate-to-plate impedance equal to the total load impedance for a single tube (i.e., if 8kΩ gives max output for a single 6V6, then 8kΩ plate to plate gives max output for a push-pull pair of 6V6's at the same supply voltage).

As you work deeper into class AB, the rule changes. Lower load and higher voltage (along with larger bias voltage and less idle current) yields more output, but that's too many changing variables to draw a meaningful conclusion about a slight change in load impedance.

Yes, this chart is for a 6V6, not an EL84 (but I'm not clear from Timbo's original question which he plans on using). However, both are 12w tubes, so the power output curve with fixed supply voltage and changing load impedance should hold between both types. It also shows the general concept of a "ideal load for maximum output power" where higher and lower loads cause a drop in output power (and either more distortion or a different flavor of distortion).

It's a separate, and simpler, process to estimate a load giving maximum power if we know the supply voltage.

[TIMBO]

Thanks HBP, Your tube knowledge amazes me. The reason I asked is I have a full set of 18w transformers that I want to use and I just looking for am option on the power tubes as I have both  EL84s and 6V6s.Thanks 

[HotBluePlates]

... just looking for am option on the power tubes as I have both  EL84s and 6V6s.Thanks 

The only thing to be aware of on 6V6 vs EL84's is that EL84's require a little less bias voltage to reach the same idle current.

So if you build an amp with cathode bias the cathode resistor should be smaller for the EL84 version, or if you used fixed bias the bias voltage (or range of available voltage) should be less-negative for EL84's.

An easy example of this is how you might see a 250Ω cathode resistor for a pair of 6V6's, but something more like 100-150Ω would be typical for EL84's.

[TIMBO]

I am I the process of putting together one of the Trainwreck versions and i'll start with the EL84 with cathode bias (25uf/130r). The 65 Tupelo has a similar circuit but I'm not sure on the ELs being a bit brite and hashy may flick it over to 6V6s.Thanks 

[HotBluePlates]

... I still have problems with this discussion because the application as presented is outside the available design information.   ...

What application? I don't follow, as what I presented are general principles. Specific numbers change, but the notion of a maximum power output with lower power output and/or more distortion above/below the "ideal" range of load is common to all tubes. Really, it's common to transfer of power in general.

The output curve for primary resistance is plotted up to about 9K.  so the output  information if off scale.  

Yes, for our understanding it is unfortunate they didn't continue up to 10-14kΩ, so you could see the bell curve. But you must see there is the beginning of a downward trend just a bit below 8kΩ.

See the graph below, from page 566 of RDH4. Different tube type, different supply voltage and load specifics. However, you see the exact same overall trends.

If you stop thinking of tubes as a mystery, and instead pretend that in an instant in time they are resistors, then Ohm's Law and the equation for power will lead you to see that if you change supply voltage and/or allowed tube dissipation, the load for maximum power shifts higher/lower but overall you still arrive at the same conclusion...

The 6V6 data is stated at 250 volts, yet, the transformer will see significantly higher rms voltages by design. The mullard data I have on el84 is at higher voltages but one stated primary resistance.

Yep, you found out why I didn't post an EL84 data sheet example: they chose to stick with a given load impedance and show changes in other variables.

So to see where I'm coming from, let's do a thought experiment:
- 6V6 and EL84 are 12w devices, so ideal load at 250v is the same for both (bias and drive voltages will be different).
- If the tubes are operating class A with our optimum 250v load, then if supply voltage is increased the load impedance must increase to hold down tube current.
- Now let's say we want class AB operation. Supply voltage has gone up, but we lower the load impedance to allow bigger peak currents for more output power.
- We still have to hold down idle current and keep the tube from overheating, so we increase bias voltage (compare bias voltages between hot class AB and lean class AB amps).
- From low-volt class A to high-volt class A, load impedance got bigger. From high-volt class A to high-volt class AB, load impedance got smaller. The net result might be the same between low-volt class A and high-volt class AB. Compare the 250v 6V6 (~8kΩ or a bit less) to a Princeton Reverb (high-volt class AB, using 8kΩ primary).
- Compare Princeton Reverb to Deluxe Reverb; both are high-volt class AB, but for more power the Deluxe Reverb uses a 6.6kΩ OT with basically-same supply voltage.

... we are in undocumented Territory, yet we know the valves work in this territory.    

It seems there is a disconnect between data sheets and guitar amps, yet it's not voodoo. It's challenging to get a handle on because few books exist that roll it all up in a way that covers all the bases.

If I was smart, I'd make a note of all the common stumbling-blocks and write a book that covered them. The challenge is how to avoid covering the simple things (that all the other books cover), while still explaining deeper subjects that assume you know all the basic info.

[jojokeo]

It seems there is a disconnect between data sheets and guitar amps, yet it's not voodoo. It's challenging to get a handle on because few books exist that roll it all up in a way that covers all the bases.

If I was smart, I'd make a note of all the common stumbling-blocks and write a book that covered them. The challenge is how to avoid covering the simple things (that all the other books cover), while still explaining deeper subjects that assume you know all the basic info.

Why not, and why wait? This type of info confusion & questions come up often. Combine it with proper PT requirements and you got something that everyone would want. Other's books makes these simple to understand things way way over complicated for most and doesn't even directly answer some of the most basic of questions. Same with your explanation of triode tubes where internal plate resistance matters over tube gain and how it relates to bandwidth...take many of the most-asked questions posed here on Doug's site and you've got everything you need for subject matter for you.

Present the info properly and "they" will come. Provide the basics with focus w/out too much technicality and many books will sell. Then you could come out with book 2 called "The math behind the madness" or "Questions you didn't know to ask...until now". ;-)

[HotBluePlates]

HBP, you keep using that magic work impedance,  The values of any impedance are frequency dependent, whether its an inductor, or a capacitor. ...

You're fixated on reactance (X), which varies with frequency and is a component of impedance (Z), along with resistance (R).

If you come to the realization that the primary impedance of a transformer is due (in production transformers) to the transformer's turns ratio between windings and the load attached to the secondary... well, the heavens will open and lights beam down upon you.

I recommend reviewing basic properties of a transformer, specifically voltage and current step up/step down ratios, and what those imply about primary and secondary impedance.

I think you have the digital copy of RDH4 (which can be found in the Library of Information, if you don't); see pages 199-201 on Ideal Transformers (you can look at "Practical Transformers" but you should know the ideal case first; practical transformers are simply limited in how high/low they go and still behave like an ideal transformer).


Very strictly, you are not wrong to question reactance. The inductive reactance must be great enough that the reflected primary impedance is smaller, at the lowest frequency of interest.

Let's say you have a transformer that has a primary inductance of 8H at a nominal current. Will it pass your guitar's 80Hz E without attenuation?

X_L = 2*pi*f*L = 6.28 * 80Hz * 8H = 4019Ω

If the transformer is supposed to reflect 4kΩ plate to plate, the ~4kΩ reactance is in parallel and reduces the total load. Bass response will be reduced.

To compare, Hammond claims ~42H open circuit for its Bassman/Super Reverb replacement OT. That is an inductive reactance of >21kΩ at 80Hz. So the reflected 4kΩ in parallel with 21kΩ means the tubes see essentially 4kΩ.

A professionally-designed transformer always takes primary inductance into account, along with design-power-output and lowest frequency of interest. Design trade-offs are made to ensure the inductance is big enough to ensure full power at the lowest design frequency.

This is also why I keep telling people who want to use push-pull OTs (not designed for large unbalanced d.c.) that the OT will work in a SE application but might suffer reduced bass output: core saturation causes the inductance to drop severely, so bass will be rolled-off at a higher frequency than originally designed.

[EDIT: Went looking... We had this exact same conversation a couple months ago. I'll still post more proof on the new concept of power output curves; stay tuned!]

[HotBluePlates]

A few other data sheets show the same type of info Re: power output changes with load variation.

See first below the graph from the top of page 9 of the G.E. 6L6GC data sheet. The same general trends are shown for all graphed items as with the 6V6 (which was also from a G.E. data sheet). Too bad G.E. doesn't continue upward with the plate load impedance, but I suppose they saw decreasing power output and increasing distortion as a double-whammy for decreasing clean output power.

Next up is from page 10 of a KT66 data sheet; this tube is broadly similar to the 6L6, and some say one was created to sidestep patents on the other. The values of load continue higher than in the G.E. sheet, and show that above the broad "ideal range" of load, power drops off.

Page 8 of the GEC KT66 data sheet shows several similar plots (one is identical), but this time more of the graph is plotted on the side above the maximum power output point (mainly to highlight falling distortion for the UL connections given).

So you see, the specific tube type doesn't matter, the general trends tend to hold. Specifics on load impedance for maximum power and distortion numbers vary with tube type and operating voltages, but you should be able to see how we were able to make a judgement about power output based on typical, general trends and experience with a given tube & operating voltage.

[HotBluePlates]

... Timbo has a o/t which at 4 ohms has a primary impedance of 10.8k while at 8 ohms, has a primary impedance 9.2k.  Should we expect a different impedance at different outputs. ...

Agreed; I don't know how his transformer information is 10.8k:8Ω and 9.2k:4Ω. We have no info on the part provided, so I make all assumptions about power output based on the primary impedance.

You raised the point of reactance and my focus on reactance.  So I ask the question of how much of the reflected impedance is reactance ...  In  my book, 95% plus is reactance ...

This is the fundamental misunderstanding. You're not moving beyond the concept of a coil (inductor) to that of a transformer.

I don't grasp chemistry too well... Pure sodium has specific properties, among which is a pretty fun reaction with water. So sodium chloride has the same severe reaction, right? Or does the presence of, and interaction with, a chlorine atom change the properties of the total compound of sodium chloride?

The dominant property of a coil is its inductance, which increases with frequency. So any consideration of the coil has to be done with respect to a frequency (or frequency range).

A transformer adds (at least) an extra coil, and in addition to the inductance that each individual coil possesses, there is a mutual inductance from one coil to another.

This fact leads to the properties of an ideal transformer, wherein the volts-per-turn of one coil must equal the volts-per-turn of the other coil, that coupling happens from one coil to the other, and that impedance can be reflected from one coil to the other.

...  So I ask the question of how much of the reflected impedance is reactance, Looking at Hammond's data on a Marshal replacement transformer, the EDB1750Q, you will find this transformer has primary impedance of 7.4k, while the dc resistance of primary windings is about 160 ohms.  In  my book, 95% plus is reactance ...

You didn't include the open-circuit inductance for the 1750Q of 136.42H.

The claimed frequency response goes down to 50Hz for full power, so let's begin with inductive reactance at 50Hz. 2*pi*f*L = 42,857.6Ω

The 160Ω DCR is insignificant in the face of that amount of reactance, as a vector-sum totals 42,587.9Ω (99.4% is reactance).

The catch is that we don't operate the transformer open-circuit, but it isn't shorted either. And it isn't obvious how 42.5kΩ of impedance squares with the published primary impedance.

[HotBluePlates]

... Hammond's data on a Marshal replacement transformer, the EDB1750Q, you will find this transformer has primary impedance of 7.4k ...

The primary impedance on its own is 42.5kΩ as I've shown. But Hammond claims 7.371kΩ, so what's up?

They also say the turns ratio is 21.46:1 (this is for 16Ω, as you'll see). I've already linked the discussion with the basic mathematical derivation, so I won't repeat that here. Instead, I'll point out how Hammond's data sheet reiterates the fundamental properties of a transformer.

You mentioned, "What was not explained in this discussion was the voltage outputted at the 8 ohm leads is about 1.414 times higher than that found at 4 ohm leads."

Notice Hammond has a pair of number beside the indicated secondary impedances:
4Ω -> 14.1v, 3.5A = 49.35w
8Ω -> 20v, 2.5A = 50w
16Ω -> 28.3v, 1.8A = 50.94w

For the same power output, if the secondary impedance is lower, the voltage output is lower and the current is higher. Really, their chart should have 50w for each, but the differences are all rounding error (4Ω output voltage is really 20 * √2, or 14.142... v, 16Ω voltage is really 28.284... v).

Look at that other thread I linked earlier; I showed that impedance ratio = turns ratio²; turns ratio = voltage ratio.

- The voltage of the 4Ω tap for 50w is 14.14v, and the voltage of the 16Ω tap for 50w is 28.28v.
- Voltage ratio = 28.28v / 14.14v = 2:1
- Impedance ratio = 16Ω / 4Ω = 4:1 = (2:1)²
- To have our "constant volts-per-turn" rule hold, the 16Ω tap must have twice the number of tuns as the 4Ω tap.

So how does the voltage of the 8Ω tap wind up 1.414 times higher than the voltage of the 4Ω tap?
- 1.414 roughly equals √2.
- 4Ω has 14.14v for 50w; 8Ω tap has 14.14v * √2 = 19.997v (call it 20v)
- Voltage ratio = √2:1
- Impedance ratio = (turns ratio)² = (√2:1)² = 2:1

So the rules of voltage and impedance ratios hold whether you're talking about among section of a winding or from one winding to another. The basis is equal volts-per-turn.

Turns ratio is given on the data sheet as 21.46:1, but this is for the whole primary to the whole secondary. I already showed impedance ratio is turns ratio squared, so (21.46:1)² = 460.5316:1  (primary to secondary). 460.5316 * 16Ω = 7368.5056Ω, which just means Hammond rounded 21.463632:1 for the sheet.

The data sheet primary impedance was found without reference to any frequency, or with any info of the transformer's inductance, but based solely on knowledge of the turns ratio and fundamental transformer properties. Therefore, the transformer's primary impedance is based solely on the impedance attached to the secondary, and reflected via the turns ratio to give an apparent primary impedance.

So what about that 42.5kΩ we calculated earlier? If that's the reactance due to the open-circuit inductance (end-to-end) of the primary, and the properties of a transformer places a reflected impedance of 7371Ω on the primary (end-to-end), what happens if these values are effectively parallel impedances?

7371Ω || 42.5kΩ = (7371Ω*42587Ω)/(7371Ω+42587Ω) = 6283Ω

What about at the other end of the spectrum, 12kHz? Reactance there is over 10MΩ. 10MΩ || 7371Ω = ~7365Ω

Which is basically no effect. At the treble end, other transformer imperfections limit frequency response.

... Your discussion regarding the output at 80Hz only tells part of the story.    There is no discussion the effects of the speaker, where most available speakers have resonance frequency peak in this area of operation.  ...

The transformer will still have measurably reduced response at the upper and lower end if you use a power resistor as the load; while a speaker's varying impedance has an impact, it's not the sole cause for everything.

What about a discussion of hearing loss? That has the most impact on the loss of bass, right??? Throw as many complications in as you want, but you just make it an impossible problem to comprehend.

When you reach a point of understanding this stuff, you will find that simplifying assumptions get you the 80-90% solution. After you have that, you can refine with the complications that separate ideal components from real components and get the rest of the way there.

It's taken me more than 10 years to stop making things harder than they are; once I stopped that, electronics became easy to understand.

[PRR]

> I don't know how his transformer information is 10.8k:8 and 9.2k:4. We have no info on the part

If he'd bothered to give us a link, it might be this:  http://www.classictone.net/40-18037.pdf

The odd impedances are odd. However it is all +/-20%, so it is "all the same" on stage.

[HotBluePlates]

Hmmm... the numbers look suspect, but it's not the first time Magnetic Components' sheets didn't add up (IMO).

EDIT: Got it. The winder couldn't figure out how to start with a certain number of turns for the 16Ω tap, have that number /√2 be a whole number of turns for the 8Ω tap, and the 8Ω number of turns /√2 be a whole number. Looks like he kept rounding up a little for each of the taps, and came out with odd numbers for the primary impedance as a result.

[TIMBO]

Thanks PRR, you beat me too it.

[jazbo8]

Man, I went away for awhile and missed a bunch of good discussion here, HPB has already covered all the grounds as usual - the detour to reactance was a bit geeky ... yes, we know the actual load lines are elliptical not straight as they are normally shown... anyhow, assuming purely resistive load, below is the output power & distortion vs. load for the push-pull EL84 stage. It can be seen that the maximum power is realized ~8k, and drops off a bit ~10k (although the distortion does increase quite a bit). Slightly OT, isn't running the EL84s at 250V kinda conservative for 18W type of amp?

[HotBluePlates]

... I believe the primary and secondary impedances are determined at 1kHz, ...

No, it is not.

It is determined by turns ratio and the value of the load attached to the secondary.

... determined at 1kHz, and also relevant to these impedances is the voltage feeding the transformer. 
Looking at the data sheet there is some critical information.  This is the 1kHz reference, and ...

Hammond has specified response as 50Hz - 12kHz, +/- 1dB.

dB is not an absolute unit, but a ratio. A reference level must be specified for it to be relevant. The reference level, in this case, is whatever the output is at 1kHz.

... Looking at the data sheet there is some critical information.  This is the ... reference, and 424v.  ...

The turns ratio was given as 21.46:1. You mention 424v, which was given as the (primary) voltage for 50w output.

424v / 21.46 = 19.75v  (primary voltage / turns ratio)
424v / 20v = 21.2:1     (primary voltage / sec. voltage)
20v * 21.46 = 429.2v   (sec. voltage * turns ratio)

So it looks like Hammond was providing the needed primary voltage to arrive at 20v output on the 8Ω tap, to get the 50w output specified for the transformer. But there's some rounding error; no matter, as the primary voltage is only ~1.2% low of what it should be according the the turns ratio.

...  While an article by Meeno van der Veen "Measuring Output - Transformer Performance", does not fully address this issue, he presents how the secondary impedance varies as function of voltage, knowing the frequency.  ...

No fair not linking the reference.

He shows the secondary inductance changes (as well as primary inductance) as a result of the applied voltage (and the current which results) either not overcoming magnetostriction, or saturating the core.

To use your reference, what is the primary impedance at 1kHz for the transformer shown in Fig 7 of that article? (Hint: the inductance at the high point of that curve is a little under 800H.

Notice also that in figure 7, the x-axis is voltage, not frequency.

... The data shows inductance of 136 H, at 240v, 50 Hz.  The voltage and the frequency can be found at some wall plugs, so this should be a straight forward measurement. ...

This is an output transformer... so how does wall voltage figure into the discussion?

You tell me: what's the impedance of 136H at 50Hz? How does that impedance compare to the claimed primary impedance? Where else does 50Hz appear on the data sheet?

[PRR]

> how does wall voltage figure into the discussion?

It doesn't; but it is super convenient that a semi-typical wave for a tube amp carrying bass is very similar to what comes out of the wall in the UK, Germany, China. Instead of building a 100 Watt super-good bass amp to test OTs, just plug-it-in. Ideally no current will flow. In real world there is current. We make a note of the current of an acceptable OT, and reject any that pull much more (much less is also suspicious). Secondary voltage should be right also. It's just a Production Test. Like hiring me to go "Whang!" on every Gibson guitar made. I can't tell if it has the magic soul, but I can tell when it just don't work (no sound, pegs pull out...).
______________

> certain number of turns for the 16

Yes, the turn-count didn't come out even. It never does. It is not untypical to find a 2:3:4 (4/9/16) ratio, four layers all the same.

Also note that "4/8/16" is a fairly new and US-centric thing. Well into the 1950s we used 3.2 ohms for short-run (record-players, cars) and 15 ohms for long-work (talking picture theaters). So the 3/7/15 series shown on that plan is not dart-toss.

Any speaker will vary +500%/-20% from its midbass trough impedance. The amplifier can not be too fussy. And the "original power amp", a triode, is VERY unfussy about impedance.

Plotting the power in load, assuming a fixed source resistance and varying load impedance, is instructive. There is a wide optimum.

Pentodes are not a resistance; or rather they are two resistances. Below and above the knee. Perhaps 1K at low plate voltage and 25K at high plate voltage. The "best" load is then split-the-diff, 5K. However the split is so wide that it isn't real fussy.

_____________________________

> Meeno van der Veen "Measuring Output - Transformer Performance", does not fully address this issue, he presents how the secondary impedance varies as function of voltage

How the UN-LOADED impedance varies.

Put the Nominal Load across the un-loaded impedance, it hardly varies at all.


> determined at 1kHz

Basic Concept: Audio is Wide-Band. It *has* to cover 300Hz-4KHz pretty-much "flat". In many cases, 100Hz-10KHz or 50Hz-15KHz.

That is a w-i-d-e range of frequency.

So any "large" variation with frequency MUST be avoided.

So we may "assume" that the 500Hz and 2KHz numbers are very similar to the 1KHz numbers.

And there is no real "significant" variation until we get to the extremes of the intended band width. If the iron is rated 100Hz-10KHz, we DO expect a difference at those points. Very often +/-30% (3dB) at those points, +/-10% (1dB) an octave inside those extremes.

10% is very small. 30% is not large. The Champ probably has 40% OT impedance drop at 82Hz bottom-note. (Also note that the tiny speaker in a genuine Champ can't do 82Hz well, so the OT is not the limiting fator.)

[sluckey]

What a detour! All Timbo ever wanted to know was "Am I able to sub the EL84s for 6V6gts."

[HotBluePlates]

Give me a chance to defend myself ...

Unfortunately, you see an attack where we are trying to help you where we see a misunderstanding.

Good luck with your investigation of tubes and transformers.

[TIMBO]

Hi guys, When it comes to math and theory that is in bedded in doing this stuff, I fell at times I ask way too much of you more learned guys. I am thankful for any info that is freely given and when I post stuff I hope that all that read it can benefit from what ever is put forward. Thanks   

[jazbo8]

I’ll be the first to admit that audio engineering and electrical engineering are not my field of expertise, Yet my back ground, involves evaluating total systems and the interactions of the various unit operations, and because of this, and my skepticism, I know I have inserted foot in mouth, and left foot prints if not a lot of sand and mud on most of my teeth.   As several people on this forum have pointed out, I have made errors in some of my postings. I am grateful for the patience the moderators have shown me.   I know from other postings I have made, that I have frustrated at least two of the moderators.  I apologize to both Sluckey and HBP.  To other posters, I also apologize for any confusion I have been the source of.  .

You have presented some useful information for us to digest, I think there are two related topics that were brought up, from a system point of view, of course, they do interact and should be considered in the overall design. What you dug into with speaker reactance and its knock-on effect on power transfer and frequency response, etc. is commonly treated separately (perhaps you do not agree with this approach, that's fine). Ditto for the output transformer specification and qualification, they are usually treated separately as well. When we look at power transfer from the output stage via the OPT to the speaker, we usually assume certain mid-band, nominal impedance, e.g., 4/8/16 Ohm. It's by no means perfect, just a shortcut to get us into the right ball park.