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

[Mike Sulzer]

One important factor is the load the attenuator places on the amp: there is agreement that it should be like that of a guitar speaker, and we will discuss that more later, but first let's look at the impedance of a couple of speakers. First is one of my favorite speakers, the no longer produced Eminence Red Fang, an Alnico speaker. The most frequently shown plot is the magnitude of the impedance versus frequency. I have made some measurements with the speaker in a Line 6 open back cabinet, shown in the first attachment.




The measurement technique is very much like what I use for guitar pickups: A resistor its placed between one speaker terminal and ground, and the other terminal its driven with a voltage. This voltage to ground is measured as is the voltage across the resistor. The instrument is an inexpensive recording interface attached to a computer. The waveform used is a Golay complementary sequence. Cross spectral analysis is performed.

The most noticeable feature is the low frequency resonance; the magnitude I measured is lower than the one published by Eminence. This is probably a function of damping introduced by the open back cabinet. The next most noticeable feature is the rise with frequency. One could look at this and might be confident that this is a result only of the inductance of gate speaker coil.

However, there is more to it than that. The real (the apparent resistance as a function of frequency) and imaginary (the apparent inductance as a function of frequency) parts are shown in the next attachment.



Now we see that the rise in impedance is a result of both inductance and increasing resistance. The latter is probably a result of eddy current losses in the metal parts.

The next question is what are all those little (maybe not so important) features? Are they the result of the measurements technique, including the cabinet, or are they a function of the driver? So here are the results for another speaker, a Celestion GP12-80. I wanted to measure a Celestial Blue, but it turns out that it did not survive Maria).



The overall impedance is very similar; some of the small bumps are similar, some are not. Thus we have a mixture of measurement and speaker effects.

When I get to the next post, it will be about why we need to load the amp and what we need to do th get a scaled copy of the voltage across the load.

[olddawg]

I use a speaker attenuator every time I gig. I use a Weber with the fancy speaker motor and stuff. I have the footswitch option so I can turn it off and on at will. It’s great for bringing lead volumes up with the amp saturated. It has some treble cut/boost switches on it I rarely use. It does depend on the cab. AND I’m not squashing the output.. just taking the “edge off”. Honestly though, as I mentioned in another post, for my last 20 watt build I simply added a $12 100 watt L-Pad from all parts on the back. I really don’t hear a huge difference in tone when used in that manor. Everyone thinks it’s magic in the little studio where I use it.

[Mike Sulzer]

I use a speaker attenuator every time I gig. I use a Weber with the fancy speaker motor and stuff. I have the footswitch option so I can turn it off and on at will. It’s great for bringing lead volumes up with the amp saturated. It has some treble cut/boost switches on it I rarely use. It does depend on the cab. AND I’m not squashing the output.. just taking the “edge off”. Honestly though, as I mentioned in another post, for my last 20 watt build I simply added a $12 100 watt L-Pad from all parts on the back. I really don’t hear a huge difference in tone when used in that manor. Everyone thinks it’s magic in the little studio where I use it.

Yes, I remember that post, and I think that it is important to find out how simple something can be and still work well, just as it is important to gain a better understanding of how things work by exploring more complicated designs.

[Mike Sulzer]

A "thought experiment" that can be tested

Suppose we haver a perfect device that can sample the voltage across a guitar speaker without affecting it and reproduce it exactly into any load at any power level. Connect an identical guitar speaker to the output. Would this second speaker sound just like the one one that is sampled? I think it must. Add a volume control to this device and you can scale the output while maintaining the sound as close as possible to the original given that the speaker is somewhat power sensitive as is human hearing.

Now replace the speaker that is being sampled with a silent device that has the same impedance, and therefore, the same voltage across it. Now "the device" and the speaker connected to it are a very good attenuator or booster. Off course, a good ss amp approaches this ideal device very closely. Given that electric guitar is not exactly audiophile, the ss amp might not have to be the very best. A light, inexpensive switching amp might be good enough.

Unlike many ideas, this one is double blind testable. You can can have someone else switch between the two, or not, and see what you can hear without knowing which is which ahead of time.

[Helmholtz]

First of all, thanks for starting this thread. I actually planned something similar.

Regarding speaker impedance, I think there are different effects to be considered.

First there is the "print-through" of speaker impedance on the frequency response of the amp-speaker system. This depends mainly on the source impedance of the tube amp. A higher inner impedance will produce more print-through.
The amp output impedance as seen by the speaker depends strongly on its NFB ratio, but also on tube type and operating conditions. Some measured values (@1kHz) relating to 8 Ohm outputs:

JTM 45: 2 Ohm
Super Reverb: 18 Ohm
AC 30: 70 Ohm
Tweed Deluxe: 80 Ohm
Marshall 18W: 90 Ohm

It seems that all the speaker needs to reproduce a corresponding frequency response would be a low impedance SS amp with a suitable resistor in series with the speaker. This also takes care of the looser damping of speaker ringing at tube amps.


But the above only holds for linear operation of the tube amp's output stage. When overdriven output impedance changes and I assume that the frequency dependance of the loadline caused by the varying speaker impedance plays a major role regarding quality of sound/distortion.

And this is why reactive dummy loads make sense.

[Chuck H]

And what of kickback voltages from an actual speaker that affect the reaction/response of the amp. Without an actual moving transducer that wants to recover to a fixed position this property is lost. How important it is could be brought into question. But it may be relative to the thread topic.

[Malcolm Irving]

...
Now we see that the rise in impedance is a result of both inductance and increasing resistance. The latter is probably a result of eddy current losses in the metal parts.
...

Could the increasing resistance also be partly due to skin effect and proximity effect?

[Chuck H]

Now replace the speaker that is being sampled with a silent device that has the same impedance, and therefore, the same voltage across it. Now "the device" and the speaker connected to it are a very good attenuator or booster. Off course, a good ss amp approaches this ideal device very closely. Given that electric guitar is not exactly audiophile, the ss amp might not have to be the very best. A light, inexpensive switching amp might be good enough.

Unlike many ideas, this one is double blind testable. You can can have someone else switch between the two, or not, and see what you can hear without knowing which is which ahead of time.

I think this is the basis for the Ho, or Ultimate attenuator. Except that design uses a simple load resistor rather than an active load. IIRC a 32 ohm resistor is used for all amps regardless of impedance setting. The premise being that although a speaker may be rated at a lower impedance, it's actual impedance across the spectrum is much higher when averaged.

[Mike Sulzer]

And thank you for your reply. This is the next topic! I agree that the impedance with the overdriven output stage changes from linear operation, and since the major purpose of this type of attenuator is to allow the generation of output stage distortion at reduced sound levels, we should get the load right. But I would like to add that the transition between two tubes and one tube conducting, that is, when the A in AB moves to B, there also is an output impedance change.

The output impedance is an instantaneous quantity that depends on output voltage and current; that is, energy storage matters.

First of all, thanks for starting this thread. I actually planned something similar.

Regarding speaker impedance, I think there are different effects to be considered.

First there is the "print-through" of speaker impedance on the frequency response of the amp-speaker system. This depends mainly on the source impedance of the tube amp. A higher inner impedance will produce more print-through.
The amp output impedance as seen by the speaker depends strongly on its NFB ratio, but also on tube type and operating conditions. Some measured values (@1kHz) relating to 8 Ohm outputs:

JTM 45: 2 Ohm
Super Reverb: 18 Ohm
AC 30: 70 Ohm
Tweed Deluxe: 80 Ohm
Marshall 18W: 90 Ohm

It seems that all the speaker needs to reproduce a corresponding frequency response would be a low impedance SS amp with a suitable resistor in series with the speaker. This also takes care of the looser damping of speaker ringing at tube amps.


But the above only holds for linear operation of the tube amp's output stage. When overdriven output impedance changes and I assume that the frequency dependance of the loadline caused by the varying speaker impedance plays a major role regarding quality of sound/distortion.

And this is why reactive dummy loads make sense.

[Mike Sulzer]

And what of kickback voltages from an actual speaker that affect the reaction/response of the amp. Without an actual moving transducer that wants to recover to a fixed position this property is lost. How important it is could be brought into question. But it may be relative to the thread topic.

I think that is included in the impedance, in the imaginary or energy storage part, to the extent that the speaker behaves as a linear device.

[Mike Sulzer]

Could the increasing resistance also be partly due to skin effect and proximity effect?

I had not thought of that; does anyone know the wire size used in a speaker coil?

[Mike Sulzer]

I think this is the basis for the Ho, or Ultimate attenuator. Except that design uses a simple load resistor rather than an active load. IIRC a 32 ohm resistor is used for all amps regardless of impedance setting. The premise being that although a speaker may be rated at a lower impedance, it's actual impedance across the spectrum is much higher when averaged.

http://www.ultimateattenuator.com

This says it uses a reactive load, meaning, I think, an inductor. I have no idea why it does not need an impedance switch; that does not seem right.

[Helmholtz]

I had not thought of that; does anyone know the wire size used in a speaker coil?

I guess JMF would know.
I can do something else: I have a recone kit for a Celestion blue and can measure the series AC resistance at 1kHz and 10kHz of the voice coil alone.
According to Zollner the increased loss resistance as well as the decreasing apparent inductance at higher frequencies are caused by eddy currents in the conductive ferromagnetic parts.

Edit: Just found a value of 0.22mm diameter in the literature (Zollner).

[Chuck H]

Ah. That might be a new (ish) thing. The reactive load. I know I read somewhere that the Ultimate was a copy of the Ho and that they both used a plain ol' 32 ohm resistor. But I can't fact check it because I didn't document the info.

[Mike Sulzer]

I guess JMF would know.
I can do something else: I have a recone kit for a Celestion blue and can measure the series AC resistance at 1kHz and 10kHz of the voice coil alone.
According to Zollner the increased loss resistance as well as the decreasing apparent inductance at higher frequencies are caused by eddy currents in the conductive ferromagnetic parts.

Edit: Just found a value of 0.22mm diameter in the literature (Zollner).

The skin depth is about .45 mm for copper at 20 KHz. So it does not explain how we get from 6.5 ohms at 3 KHz to about 30 at 20KHz. Must be mostly eddy current losses.