Attenuators between a tube amp and the guitar speaker: some measurements and theory

[Gregg]

I don't think reactive loads are bad per se otherwise a lot more people would have complained over the years which is not the case. Also bigger companies wouldn't have jumped on that train either. It's more like what mhuss said earlier:

I've often wondered if the bad rep attenuators get for blowing up output stages isn't due to the fact that most people wouldn't run their guitar amps on "11" if they had to deal with the full volume. Except for maybe arenas, it is sort of self-limiting. With an attenuator, one is much more likely to run the amp way hotter then normal. And particularly with the tubes being made today...

[Helmholtz]

Fully dimed through an attenuator, I have seen the screens actually sag & then melt/ short out.

Pentode screens are delicate (more so than beam tretrode screens) and heat up much faster than the plates. Screen dissipation increases strongly when the plates are driven into saturation. Screen dissipation is further increased with a more horizontal loadline, i.e. high load impedance. This is the reason why I warn against doubling speaker impedance.

If power tubes actually die early when using a power attenuator, there are only 2 possible explanations. Either the tubes would have died just as well when using speakers at prolonged periods of tube saturation. Or the attenuator impedance does not match the speaker impedance over the whole frequency range.

I have seen and measured the frequency responses of several "reactive" power loads/attenuators that showed way too high impedances around the resonance and above 1kHz. The reason was that they didn't provide appropriate damping for the inductors used. This could be an explanation for screen failures.

[Mike Sulzer]

Yes, as shown in the first post in this discussion, the magnitude of the impedance of a speaker could appear to imply that it is nearly perfectly inductive at higher frequencies. But a measurement of the full impedance shows that this is an illusion because there is a large resistive component that also is also rising in magnitude with frequency. Maybe some people designing attenuators do not understand the difference.

For the other possible explanation, I am not convinced that guitar amps are as fragile as that implies. The screen resistors are there for a reason. In a guitar amp, tube life might be nasty, brutish and short, but they do reach adulthood. My ears tell me that much of the hard rock recorded over the years used output stage distortion, often very heavy. Of course, this was encouraged by the lack of a master volume control on how many popular Fender models?

I do not think they were replacing the tubes every couple of hours.

Pentode screens are delicate (more so than beam tretrode screens) and heat up much faster than the plates. Screen dissipation increases strongly when the plates are driven into saturation. Screen dissipation is further increased with a more horizontal loadline, i.e. high load impedance. This is the reason why I warn against doubling speaker impedance.

If power tubes actually die early when using a power attenuator, there are only 2 possible explanations. Either the tubes would have died just as well when using speakers at prolonged periods of tube saturation. Or the attenuator impedance does not match the speaker impedance over the whole frequency range.

I have seen and measured the frequency responses of several "reactive" power loads/attenuators that showed way too high impedances around the resonance and above 1kHz. The reason was that they didn't provide appropriate damping for the inductors used. This could be an explanation for screen failures.

[g1]

I don't think anyone meant anything as extreme as tube replacement 'every couple of hours'.
How would you be with 'significantly more often than without the attenuator'?

[Helmholtz]

EDIT: I forgot to say that the restive divider used in the Russian circuit can be replaced with a transformer. This does a better job of copying a reduced voltage from the load to the speaker. What you want is a "bridging" transformer, one that is considerably higher than 8 ohms looking into it.

That's the principle of my Marshall PB 100. I uses a speaker emulating circuit connected to a multi-tapped autotransformer for power attenuation.

[Helmholtz]

Could someone explain how to calculate the voltage and current distribution in the modeled circuit? I'm not sure how to specify power handling requirements where the network has reactive/storage devices and resistance/dissipation components.

Here are some simulations of current/voltage distribution(s) for an emulation of a 8 Ohm 4x12 Marshall cab. Input voltage 37Vrms (171W apparent power), corresponding to a clipping 100W head, kind of worst case:

1) At bass resonance (116Hz)
SEres.pdf
2) At 400Hz
SE400.pdf
3) At 5kHZ
SE5k.pdf

Resistor power is easily calculated from the voltage across them.

The chosen frequencies do not necessarily represent worst case for specific components. E.g., the 12.5mH inductor current increases below the bass resonance to around 5A.
In real guitar signals, power is distributed over a frequency band with a probability maximum at middle frequencies, so some de-rating of component loads at the frequency extremes is feasible. But distortion increases high frequency energy content.

Current distribution will be different with different component values.

[SoulFetish]

Here are some simulations of current/voltage distribution(s) for an emulation of a 8 Ohm 4x12 Marshall cab. Input voltage 37Vrms (171W apparent power), corresponding to a clipping 100W head, kind of worst case:

1) At bass resonance (116Hz)
[ATTACH]51700[/ATTACH]
2) At 400Hz
[ATTACH]51701[/ATTACH]
3) At 5kHZ
[ATTACH]51702[/ATTACH]

Resistor power is easily calculated from the voltage across them.

The chosen frequencies do not necessarily represent worst case for specific components. E.g., the 12.5mH inductor current increases below the bass resonance to around 5A.
In real guitar signals, power is distributed over a frequency band with a probability maximum at middle frequencies, so some de-rating of component loads at the frequency extremes is feasible. But distortion increases high frequency energy content.

Current distribution will be different with different component values.

That's kick-ass, man. That was super helpful to see some visual examples like that.
Made it clearer to think of it in terms of Kirchhoff's laws.
I will say, that at first glace I was kind of surprised that the inductance presents that much of a difference at 400Hz vs 116Hz.

[Mike Sulzer]

What is the effect of a resistive load versus a speaker load on the amplifier waveform?

The top panel shows the waveform with resistive load, the bottom with the speaker load. (The voltage levels are very different because the sample ports on the two loads have very different levels of attenuation.) The resistive load is 7.8 ohms; the speaker load is that described earlier.

The amp uses EL84s, grid biased for about 70% dissipation limit. You can tell this is hard overdrive because you can see that one tube shuts off before the other turns on as a result of charging of the coupling capacitors. The asymmetry might be a result of differences in the drive, output tubes, or both; I have not looked into this.

The frequency is 300 Hz; thus the fundamental is at the minimum point in the speaker impedance load where the impedance is real. The bass resonance is not involved in this test; the differences are a result of the damped inductor that has its major effects above 300 Hz. The effect of the speaker load is to "round" the waveform, reducing the content of high harmonics while distorting the waveform on a time scale that introduces more lower harmonics.

[Helmholtz]

Thanks for this impressive piece of evidence, Mike!

[Mike Sulzer]

Thanks for this impressive piece of evidence, Mike!

Thanks. One goals of this work was to look at how a speaker affects a tube amp waveform without going deaf or getting murdered.

[SoulFetish]

What is the effect of a resistive load versus a speaker load on the amplifier waveform?

The top panel shows the waveform with resistive load, the bottom with the speaker load. (The voltage levels are very different because the sample ports on the two loads have very different levels of attenuation.) The resistive load is 7.8 ohms; the speaker load is that described earlier.

The amp uses EL84s, grid biased for about 70% dissipation limit. You can tell this is hard overdrive because you can see that one tube shuts off before the other turns on as a result of charging of the coupling capacitors. The asymmetry might be a result of differences in the drive, output tubes, or both; I have not looked into this.

The frequency is 300 Hz; thus the fundamental is at the minimum point in the speaker impedance load where the impedance is real. The bass resonance is not involved in this test; the differences are a result of the damped inductor that has its major effects above 300 Hz. The effect of the speaker load is to "round" the waveform, reducing the content of high harmonics while distorting the waveform on a time scale that introduces more lower harmonics.

[ATTACH=CONFIG]51755[/ATTACH]

Nice traces! Im wondering how the bias excursion caused in part by the capacitor coupling effects the output waveform under heavy drive vs a directly coupled drive? My amp is dc coupled EL84s, so I could adjust the bias to 70% (HT 320V) and compare waveforms once I construct the reactive load. Actually, I use a 1X12 eminence red fang i can scope. Ill just use ear protection. The resolution on my digital scope is total shit, so i may end up using the old 465 instead.
The resolution on your images looks great though. What are you using?

[Helmholtz]

Thanks. One goals of this work was to look at how a speaker affects a tube amp waveform without going deaf or getting murdered.

Two questions:

- In an earlier post you mentioned "high" voltage spikes at the power tube plates. Did these only show at the primaries?

- Does your amp employ global NFB?

[Mike Sulzer]

Nice traces! Im wondering how the bias excursion caused in part by the capacitor coupling effects the output waveform under heavy drive vs a directly coupled drive? My amp is dc coupled EL84s, so I could adjust the bias to 70% (HT 320V) and compare waveforms once I construct the reactive load. Actually, I use a 1X12 eminence red fang i can scope. Ill just use ear protection. The resolution on my digital scope is total shit, so i may end up using the old 465 instead.
The resolution on your images looks great though. What are you using?

Thanks. With dc drive you should see less rounding since you can get higher peak current, but you do not know for sure until you try it. I have some concern for the screens of the EL84s, but with 320 volts I hope you are OK.

For signal sampling I am using an Apogee Element 24. It feeds the scope from Electroacoustics Toolbox. When I measure impedance, I use the Element 24 and my own software.

[SoulFetish]

Thanks. With dc drive you should see less rounding since you can get higher peak current, but you do not know for sure until you try it. I have some concern for the screens of the EL84s, but with 320 volts I hope you are OK.

The screens are right around 300V, and have 1k/5W screen resistors. Plus there is shared cathode resistance adding degeneration for some compression effect. Im not worried, but im putting it through abusive testing to make sure.

[SoulFetish]

I'm resurrecting this thread and project because I think it's a good one, and I was hoping to find out if anyone has made any progress on building/completing/testing this and hear what kind of results they got. It's been on the back burner for me for a while, but I have some time and a little extra cash to start sourcing parts. One of the challenges has been finding the balance of cost and size efficiency to use film caps for the ≈390µF capacitance.
I found some really good deals today on 100-150µF film caps, with voltage ratings from 100-500V and was going back and forth over +/- $4 (man, I must be painfully cheap.) Anyway, I was working in my shop on another project and I found a stock of 50µF/250V electrocubes caps I got a while ago totally forgot about. There were like 12-15 of them in the box, so I decided to make a capacitor bank using 8 in parallel.


Best part is I get to hold onto that $4.
I was going to buy a pair of Jantzen 1031 0.47mH 18 AWG Air Core Inductor for Lh0 & Lh1 and wanted to double check that these are appropriate.
I know that Mike used the secondaries of a filament transformer for the large inductor (Lb) in the prototype. But if I was sourcing out a designated filter choke for the project, can you guys offer up some suggestions to get me in the right direction? I was thinking 4-5A rating (maybe 6?). The whole unit is probably going to have to be large anyway to accommodate the caps. Go big or go home, I say.
The last question for Mike, Helmholtz, Nickb (and others) is what kind of ideas where you thinking about for the secondary solid state drivers?

for those who may be interested, here are some great prices on film caps for the capacitance needed:
130µF-500V
110µF-450V
120µF-250V
100µF-450V
100µF-100V
100µF-63V