Attenuator question

[jeff]

 I'm thinking of building an attenuator. Now as I understand it an attenuator reduces the power to the speaker so you can crank the amp and only get a fraction of the volume. Typically wired like A.

 If the speaker was exactaly 4 ohms, in this circuit, 1/4 of the power is going to each resistor and 1/4 to the speaker. I've heard that this type of attenuator isn't that good because it's impedance doesn't rise and fall as the speaker's does. Say, for example, at a certian frequency the speakers impedance is 50ohms. At that frequency(using A) you'd have 2ohms in series with 4 and 50 ohms in parallel for a total of 5.7ohms. Not much different than 4.

Can I wire the resistors in series instead and connect to the 16 ohm tap as in B? Where the speaker's responce is 4 ohms each resistor would still recieve 1/4 power and the speaker would recieve 1/4 power. I'd have a 16 ohm load going to a 16 ohm tap.  Where the speaker's impeadance goes to 50 ohms the total load would be 62ohms. Different than the speaker alone but now closer to the speakers impeadance that 5.7.

But now is power divided up differently between the 12 ohm resistance and the speaker in a bad way?
 
If it's OK I'll try both ways to compare the sound but I'm just just asking first in case there is a problem using B.

[PRR]

Looks OK.

[jeff]

 This did make me think of something I've overlooked.

 There is a difference between hooking up four 8 ohm speakers series/parallel VS. parallel. If you hook up the four speakers series/parallel and connect them to the 8 ohm tap the total cab will have the same impeadance as each single speaker. But if you connect them all in parallel and hook them up to the 2 ohm tap(like a bassman), the impedance of the cab isn't the same as one single speaker.

 If a single speaker rises to 50 ohm at a certian frequency, then series/parallel cab will also rise to 50 ohm but a paralleled cab will only rise to 12.5.

I not sure what this means.... but there is a difference. Never thought about that before.

[jjasilli]

Another option is an L-pad, sort of combining the two circuits, with a series and a shunt resistance.  The series resistor attenuates signal, and the shunt resistor maintains the impedance level.  This can done with a rheostat for a range of attenuation.  E.g.:  http://www.parts-express.com/pe/pshowdetl.cfm?&Partnumber=260-265&CFID=6559502&CFTOKEN=95428347

If hi frequencies drop-off too much, you could try a bypass cap across the rheostat.  (The cap could be switched, or operated through a pot.)  

Note that a 100W RMS rheostat will handle a 50W RMS amp, as the amp will put out 100W when overdriven.

Also ted weber has a page which computes the values and wattage ratings for a fixed resisitor L-pad, but I can't find it.  Here's another one:  http://www.bcae1.com/lpad.htm  and see:  http://www.regiscoyne.com/ampoline/

[jeff]

Basically "A" is an L-pad. It looks different because I used all 4 ohm resistors but imagine the two 4 ohm resisters in parallel as one 2 ohm resistor.
 
 I'm just wondering if the L-pad is used only because you can match your 4 ohm speaker to your 4 ohm amp. In other words is B better for keeping the load impeadance closer to the speakers impeadance, but is and L-pad used only so you don't have to have taps or worry about connecting it right at the cost of not acting like a speaker?

4 ohm amp to L-pad to 4 ohm speaker. No fuss no muss.

[jeff]

 OK I'm having a little bit of conflict here. On the one hand an L-pad won't sound like a speaker because it's impeadance isn't flying around like a speakers and on the other we use NFB beacuse the speakers impeadance is flying all around. Maybe it's more about division of power. When the speaker is acting like 4 ohms the power is split equally between the resistors and the speaker, each get 1/4 power. But when it's higher it's not. The resistors are straight resistance and the speaker is an impedance so the speaker isn't always getting 1/4 the power.

Please someone drop some math on me!

What happens when there is 5 watts coming from the amp. How much is the power split between speaker and resistor(s) in example A VS B if the speakers resistance is 50 ohms?

Getting back to "it don't act like a speaker because the impeadance isn't flying around" VS "use NFB because the impeadance is flying around", maybe a solution is this. Use B but take the NFB from the speaker not the tap. Then the amount of power the speaker sees controlls the NFB and that makes less of a difference of how the power is divided. If the speaker is getting 1/4 the power then the NFB is higher but when the speaker gets less power there is less NFB???(Adjusting the NFB resistor for the lower signal to the speaker)

One way the NFB signal is coming from the total output of the amp regardless to how much power the speaker sees. The other way the NFB  signal is depentant on how much power the speaker sees.

Does that make any sense???

Of course this would have to be build into the amp, but that's OK.

[HotBluePlates]

... If a single speaker rises to 50 ohm at a certian frequency, then series/parallel cab will also rise to 50 ohm but a paralleled cab will only rise to 12.5.

I not sure what this means.... but there is a difference. Never thought about that before.

You mentioned that before, and I had no comment because you were mixing resistors and speakers. This time, you appear to be talking about all speakers for your hypothetical case. As such, you are probably arriving at a false conclusion that connecting the speakers in parallel mitigates the variations in speaker impedance.

The specific value of impedance arrived at as a result of placing the speakers in series or in parallel doesn't really matter, because if all the elements are speakers, they are all varying their impedance in essentially the same way (they all rise/dip at roughly the same frequency).

If the series case results in a bigger total number for the impedance at a given spike (at the bass resonant freq for example), and has the highest nominal impedance (16 ohms vs 4 ohms in your example), it also has the highest driving voltage from the amp.

In any event, my point is that if you are looking at why a resistive load won't accurately mimic a speaker load in an attenuator, you're on the right track. If you're heading down a path to conclude that the amp responds differently on a 4 ohm tap with a 4 ohm speaker load vs. on the 16 ohm tap with a 16 ohm speaker load, you might be reaching false conclusions.

[HotBluePlates]

Follow-on to my statement:

The basic approach any old-school designer would take would be to simplify the complications, and assume the speaker behaves like a resistor. In most guitar-amp situations, that would be good enough, then the designer would build the amp at see what it does. The next step would be to tweak the amp if necessary, knowing how the simplifying assumptions deviate from reality.

A hi-fi designer might consider more of the complications up front, especially if heavy negative feedback is used. Feedback can make an amp unstable if you use too much and don't consider complications like parasitic capacitances/inductances in an OT. The hi-fi designer then needs to think up front about tricks to neutralize those parasitics, and provide a way to adjust on test to make the amp stable.

Check out page 3 from the 1959 Gibson catalog, which is implied to be Gibson's design lab. I see a number of Hewlett-Packard oscillators in the racks, and one item that is probably an HP 330B, which provides a test signal and measures the distortion of a device under test.

Anyway, the implication is that they design, build, then see how the amp does/doesn't perform as expected.

[jjasilli]

Basically "A" is an L-pad. It looks different because I used all 4 ohm resistors but imagine the two 4 ohm resisters in parallel as one 2 ohm resistor.  Good point.  But:

3/4 attenuation may be insufficient.  It will be noticeable, but maybe not enough?  Presumably, the attenuator will be used to get the tone of the fully cranked amp, but at significantly lower volume.  For an amp (or anything else) to sound half as loud to the human ear, the power needs to be cut by a factor of 10.  At 3/4 we're not quite there.  And, half as loud might still be too loud depending on your application.  L-pads can be designed to provide such larger power-volume reductions, while retaining good enough impedance matching.  

If you want reactance in the circuit for better speaker emulation, then check-out some of the attenuator DUMMY LOAD circuits such as at Aiken.com -- which might be adapted for an attenuator.  Or use Weber Mass drivers.  

A rheostat is more flexible than a fixed circuit.  Or it can be used to more easily breadboard an ultimate fixed circuit, with any necessary treble bypass, etc.  Portability vs. home/studio use is another design factor.

[HotBluePlates]

On the one hand an L-pad won't sound like a speaker because it's impeadance isn't flying around like a speakers and on the other we use NFB beacuse the speakers impeadance is flying all around. ...

One way the NFB signal is coming from the total output of the amp regardless to how much power the speaker sees. The other way the NFB  signal is depentant on how much power the speaker sees.

You have read something somewhere that's got you all spun around. 

You have a very good point that NFB is used to mitigate the effect of the speaker impedance varying.

If you are concerned that the amp response will change due to resistive loading, you're right; it will. If you want to mimic a speaker, skip to the answer jj suggested, and use the solutions others have figured out.

There is something being overlooked: the OT is a mirror. The OT transforms the speaker impedance into a larger impedance the tubes can work against. If there are variations in the speaker impedance, the important factor is how those variations affect the operation of the output tubes. You have also noted NFB is about offsetting the effects due to variation of load impedance.

If the amp used havy NFB (like some hi-fi amps), the output stage approaches immunity to variations in the load, and the effect of speaker impedance variation is unimportant (this case is an ideal that is generally only approached in tube amps). Most guitar amps with feedback use light NFB by comparison, and there's only moderate immunity to load variation; it's still enough to control speaker flap, plus provide some other effects. But many players seek to reduce NFB to gain more "tubey flavor". Depending on who you ask, the effects of speaker impedance variation may be a good thing or a bad thing.

We're not going to be able to design a world-beating attenuator here on the forum; there's too many variables involved. Pick yer poison, and choose a solution out there. Try and see what you think. Don't expect the amp to feel/perform exactly as it would with a speaker.

[PRR]

> then series/parallel cab will also rise to 50 ohm but a paralleled cab will only rise to 12.5.
> I not sure what this means....

8 rising to 50: 6.25 ratio
2 rising to 12.5: 6.25 ratio.

Same thing, except all scaled lower.

The concept you are grasping: if impedance _rise_ is important, don't shunt the speaker with a low fixed resistor. 8||8 is 4, 8||50 is 6.9.... only a 1.7 ratio, not 6.25 ratio. OTOH series 4+4+4+4 is 16, and 4+4+4+50 is 62... not the "ideal" 6.25 ratio, but better than the 1.7 of parallel.