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

[Helmholtz]

The issue with electrolytics is their (often unknown) RMS current limit. (The actual current in the tank circuit will be even higher than the current supplied by the amp around the resonant frequency. A 4x12 Marshall cabinet has its impedance resonance at around 120 Hz, well within the guitar frequency range. Of course, prolonged operation at the resonant frequency is not very probable.)

Expect several amperes depending on amp power. In the most important mid frequency range the cap current load is around 2..2.5 A for 50W/8Ohm operation. Not really steady state, though, with typical guitar playing.

A solution could be paralleling AC motor caps for high power.

For lower power, paralleling of non-polar electrolytics will do. I actually needed to wire an additional 1 Ohm in series with the 160µ non-polar cap in my PB100 to get the right damping/Q. So ESR and Q don't seem too critical.

[Mike Sulzer]

The issue with electrolytics is their (often unknown) RMS current limit. (The actual current in the tank circuit will be even higher than the current supplied by the amp around the resonant frequency. A 4x12 Marshall cabinet has its impedance resonance at around 120 Hz, well within the guitar frequency range. Of course, prolonged operation at the resonant frequency is not very probable.)

Expect several amperes depending on amp power. In the most important mid frequency range the cap current load is around 2..2.5 A for 50W/8Ohm operation. Not really steady state, though, with typical guitar playing.

A solution could be paralleling AC motor caps for high power.

For lower power, paralleling of non-polar electrolytics will do. I actually needed to wire an additional 1 Ohm in series with the 160µ non-polar cap in my PB100 to get the right damping/Q. So ESR and Q don't seem too critical.

The problem with non-polars in tweeter crossover networks is that the charge on the Cs is only barely enough to handle the previous loud passage, not the current one that is louder. That is, the dc voltage across the two Cs is only as large as it charged to in the previous loud passage. The problem in a parallel resonant circuit is worse since you do not get enough charging until you have enough energy in the resonant range. I am considering using a voltage doubler operating from the amp out to charge the junction point between the two Cs through a resistor. Significant charging would occur at the beginning and then only a tiny current would be drawn to keep the Cs charged.

[Helmholtz]

The problem with non-polars in tweeter crossover networks is that the charge on the Cs is only barely enough to handle the previous loud passage, not the current one that is louder. That is, the dc voltage across the two Cs is only as large as it charged to in the previous loud passage. The problem in a parallel resonant circuit is worse since you do not get enough charging until you have enough energy in the resonant range. I am considering using a voltage doubler operating from the amp out to charge the junction point between the two Cs through a resistor. Significant charging would occur at the beginning and then only a tiny current would be drawn to keep the Cs charged.

I am not familiar with this effect. Sounds like a C-dependance on charge condition, i.e. a C-nonlinearity by a storage effect. How would you measure it?

[Mike Sulzer]

I am not familiar with this effect. Sounds like a C-dependance on charge condition, i.e. a C-nonlinearity by a storage effect. How would you measure it?

This is how a non-polarized capacitor works. Suppose you connect an ac voltage source (with an R in series) to an uncharged non-polarized capacitor. For either polarity of the voltage, one of the two series capacitors has very high leakage current, which partially charges the other. When the polarity changes, the other capacitor has high leakage current and it partially charges the first. After some number of cycles, both capacitors are nearly fully charged. Then the NPC behaves like a regular C as long as the voltage across it does not exceed the peak voltage of the signal that charged it. To work properly for higher voltages across it, additional charging must occur. During "charged" operation, it has a value of C/2, where C is the capacitance of each of each of the two. During "charging" it behaves more like a capacitor of value C with very high ESR.

So our resonant circuit cannot function at the expected frequency and Q until the NPC has charged. If you test this with a sine wave, the charging occurs before you notice any problem. When the circuit is idle, the charge eventually leaks, and you have to recharge.

[Helmholtz]

Thanks for the explanation. Will think about it.

But I prefer foil caps for my > 100W load anyway because of the current issue.

[Mike Sulzer]

Thanks for the explanation. Will think about it.

But I prefer foil caps for my > 100W load anyway because of the current issue.

Foil is better, especially if you can find a good deal. I will use the low power option for now.

[J M Fahey]

Another marketing triumph. It's like the Toyta MR2 was a disaster in France as "merde"sounds like MR2 and means "sh^t".

Mitsubishi had a similar flop in Argentina ... and I guess in the rest of Latin America.

The Nissan Pathfinder sold like hot cakes, so Mitsubishi wanted its share of the market, and issued the Mitsubishi "Pajero".

Most probably Engrish google translation of "Hay carrier" or similar, implying it was excellent for cattle raising Estancias ... only problem is that Pajero literally means "wanker" ....

[Mike Sulzer]

Mitsubishi had a similar flop in Argentina ... and I guess in the rest of Latin America.

The Nissan Pathfinder sold like hot cakes, so Mitsubishi wanted its share of the market, and issued the Mitsubishi "Pajero".

Most probably Engrish google translation of "Hay carrier" or similar, implying it was excellent for cattle raising Estancias ... only problem is that Pajero literally means "wanker" ....

And let's not forget the Chevrolet "no va".

[Leo_Gnardo]

And let's not forget the Chevrolet "no va".

Buick was wise enough to paste on an alternate name for their LaCrosse model sold in Canada. From Wiki: The name was borrowed from a Buick concept car shown in 2000 referencing the sport of lacrosse. ... The LaCrosse was originally sold as the Buick Allure in Canada, as "crosse" is a vulgar word that can mean either "rip-off", "scam", or "masturbation", depending on the context, in Quebec French slang. Pity those who bought LaCrosses in the USA and drove 'em home to the Francophone areas.

And back in the 70's, Campbells wasn't wise enough to not name their canned dinners "Big Joe's" line "Gros Jos" in Canada. Translates to, um, "vast trrracts of laaaand." If you get my drift. Well what's wrong with that? Puts a big smile on any hungry man's face. But, in a can?



No, not in a can, no way!

[loudthud]

The family name of the guy who started Toyota was Toyoda which means "abundant rice harvest".

IMHO what makes a simple resistive divider type attenuator sound bad is the low output impedance as seen by the speaker. What is needed is to simply attenuate by adding resistance in series with the speaker, then add a resistor across the output of the amp to ground to act as a load for the amp. An improvement is to make the total load act like a speaker, but this is of lesser importance. The Dumble attenuator (whatever he calls it) is an example.

[Helmholtz]

Sorry guys, please show some respect to those interested in the technical (and sonic) aspects of this thread and try to keep the jokes and puns to the soap box.
Sometimes I miss the moderator.

[Helmholtz]

This is how a non-polarized capacitor works. Suppose you connect an ac voltage source (with an R in series) to an uncharged non-polarized capacitor. For either polarity of the voltage, one of the two series capacitors has very high leakage current, which partially charges the other. When the polarity changes, the other capacitor has high leakage current and it partially charges the first. After some number of cycles, both capacitors are nearly fully charged. Then the NPC behaves like a regular C as long as the voltage across it does not exceed the peak voltage of the signal that charged it. To work properly for higher voltages across it, additional charging must occur. During "charged" operation, it has a value of C/2, where C is the capacitance of each of each of the two. During "charging" it behaves more like a capacitor of value C with very high ESR.

So our resonant circuit cannot function at the expected frequency and Q until the NPC has charged. If you test this with a sine wave, the charging occurs before you notice any problem. When the circuit is idle, the charge eventually leaks, and you have to recharge.

So a simplified model for an NPC would look like this?

NPEC.pdf

The resistors representing the (voltage depending) inverse leakage resistances.

[Mike Sulzer]

So a simplified model for an NPC would look like this?

[ATTACH]51196[/ATTACH]

The resistors representing the (voltage depending) inverse leakage resistances.

Yes. Method 2 in the text I included above adds external diodes to get the job done faster, at least under some conditions.

[Mike Sulzer]

I have built a prototype using stuff from the garage, simulating as necessary to get reasonable response before building.

Here is the simulated circuit also with a description of the components used in the prototype.
I

Here is the Python program used for the simulation.

cs.txt

Next post, look at the computed and measured responses to look for reasonable agreement.

[SoulFetish]

Any progress on a physical prototype?
The large capacitance is a challenge, though. Plus, isn't the need to be tethered to an external power supply for a proper NP electrolytic capacitor supply kind of a drag?
If you're willing to sacrifice some physical space, I think you can achieve this capacitance using much higher quality film capacitors)
For instance, doing a quick search, I you could build a bank of five 70µF/500V capacitors totaling an area of 152.5mm(L) X 57.5mm(W) X 50mm(H) for $20.95. Obviously, 500V is not needed and lower voltage/smaller caps are available, but I selected these were less expensive.