Why people keep talking about using pentodes in triode mode for power amp?

[SY]

Reducing source impedance s not just for the speakers- it also reduces distortion from the output transformer.

 

[lcsaszar]

There is another way to get around the boomy sound with highish impedance power amps. Amps are made to give flat response into a flat resistive load. Speakers are made to give a desired (usually flat) response when driven by a voltage (low impedance signal). But if you add a parallel RLC network, to an overall speaker's terminals, designed to make it present a flat resistive overall impedance to the amp, you get the same benefit as if the amp had a low output impedance -- the amp plays a flat voltage response and boominess and midrange/tweeter weirdnesses go away.

It's not as hard to do as it might seem, and certainly easier to do than making amp mods to lower output impedance. You just need a response curve of the speaker's impedance (using something like WT3 is easiest), and load the data into a speaker circuit simulator program (such my Xsim app). Then add components simulated across the given impedance to pull the higher impedance parts of the curve down to match the lowest parts... basically make lossy series-LC resonators to pull on the speaker's parallel resonators. Even just playing around without much circuit intuition, it's not very hard to see how it works and to get there. When you have a network that does what you want, get the parts and strap them at the amp's outputs or at the speakers' inputs. Then it doesn't matter what the amplifier's output impedance is, it just plays into its designed resistive load and is happy.

I did that with my horn speakers when used with a NFB SET amp that had things sounding very boomy and kind of strange through the midrange. A handful of components on each speaker system brought the impedance that originally varied between 4.5 ohms and as high as 30 ohms down to a 4 ohm curve with only about an ohm of variation (you can only bring impedances in places down, not up, though). Sound tightened right up with bass as solid as it was with a solid state amp, but with the SET HD2 detail enhancement.

If anyone out there wants to try this, get or borrow a woofer tester (WT2, WT3, DATS, or other) and measure your speaker's impedance into a "ZMA" file and email it to me. I'll be happy to toss together a network design for you to build up (not much work, it would just be a few Rs, Ls and Cs most likely). I'm getting convinced that it makes more sense to get speakers to have a flat impedance than to demand amplifiers be impedance immune. Why let speaker designers be sloppy and toss the grief onto amplifier designers? Though both being well behaved might be a good way to go, too!

I hope some of this made sense. I'm a bit sleep deprived tonight.

It is not possible to equalize a parallel LCR to being pure R.

 

[bwaslo]

There is some confusion here. Most speakers are designed to be fed from a very low impedance, well below an ohm. Your parallel RLC network would need to present this sort of impedance to properly compensate for a high amp output impedance. This would flatten the speaker frequency response, as you say, but also severely load the amplifier. Power goes down, distortion goes up, and LF extension takes a battering too. Not the way to correct a problem.

It might work as a partial correction, provided you can put up with the downside. Much better to fix the amp, or buy more suitable speakers.

Being fed "from a low impedance" is equivalent to being fed from a voltage that is not affected by the speaker impedance. Adding a network to make the speaker+network impedance to appear overall resistive does the same thing.

It does not "severely load the amplifier" -- if the speaker without network dips to say, 4 ohms, over some range it is effectively a 4 ohm speaker and on a tube amp you'd want to drive it from the 4 ohm taps, anyway. Bringing the impedance down to 4 ohms over the rest of the audio band via a network still leaves the load at about 4 ohms, but without the big jumps at resonances. Also makes the load line act more like a line and less like an oval. Sure, it's less efficient at resonances but that's not useful in most cases, it's overall efficiency -- and PROPER RESPONSE -- that matters.

 

[bwaslo]

It is not possible to equalize a parallel LCR to being pure R.

It's possible to flatten it close enough. Done it, have you tried?

 

[TheGimp]

So, run a frequency response plot of the amp and speaker combined. Then work out the network in parallel with the plot to linearise it?

 

[bwaslo]

So, run a frequency response plot of the amp and speaker combined. Then work out the network in parallel with the plot to linearise it?

Not quite, that is equalization, which would mean the amp and speaker would have to remain paired to use that network. For impedance flattening a speaker, get an impedance plot of the speaker, then work out the network to shunt the speaker with by a circuit simulation. That makes a speaker that is more immune to amplifier output impedance (damping) and cable resistance, for whatever amp you want to use it with.

 

[DF96]

Being fed "from a low impedance" is equivalent to being fed from a voltage that is not affected by the speaker impedance.

You are confusing two different effects: the effect of the speaker on the amplifier output voltage (which your method will correct), and the effect of the amplifier output impedance on the speaker bass resonance and hence sound output level (which your method will only make a small difference to).

 

[bwaslo]

Um. If the voltage at the speaker terminals stays the same (over frequency), how could that affect the speaker damping? It could only affect by the speaker putting current back to the amp... which, coming from an impedance would affect the voltage. You are confusing EMF effect as if it were somehow different than impedance effect.

Or look at it this way: take an amp with zero ohms output impedance. Connect to speaker, sweep over the audio frequency range. Measure the voltage sweep across the speaker terminals. Then use the same amp with, say, 8 ohms but flat output impedance (a very bad case). Modify the impedance load seen by the amp with a shunting network so it becomes 8 ohms flat resistance (ideally). Turn the amp up 6.02dB (to make up for the series loss of the amp), and repeat the measurement. Same voltage results across the speaker terminals, so speaker behaves exactly the same as before. The amp (or amp with network) can only put a voltage across a speaker system and support the resulting current draw. No other magic happens. And the same response happens whether you call the network part of the amp, or part of the speaker.

 

[DF96]

In one case the speaker, acting as a voltage source with finite output impedance, sees a load of zero ohms (the amplifier output impedance). In the other case the speaker, acting as a voltage source with finite output impedance, sees a load which varies from around 8 ohms down to around 4 ohms (amp in parallel with compensation network). Do you regard these two quite different circuits as being identical? Your mistake is to only consider the effect on the speaker of the signal coming from the amp. You also need to consider the effect on the speaker of the signal coming from the speaker. The circuit has two sources, not one.

To look at it another way, you are saying that an amp with zero output impedance is equivalent (at least for that speaker) to an amp with an output impedance which varies from very high down to around 8 ohms.

 

[Merlinb]

Do you regard these two quite different circuits as being identical?

I think he is assuming the speaker can be corrected so that the load appears to be a constant resistance, in which case the voltage across the speaker really would remain constant with frequency regardless of the output resistance of the amp (provided it is also constant). But as has already been mentioned, you can't correct a loudspeaker to appear as a constant resistance.

 

[lcsaszar]

The voice coil inductance can be compensated by a series RC connected parallel to the speaker terminals. R is the DC resistance of the voice coil, and the time constant of RC and L/R should be the same. But I am not sure how the low-frequency impedance peak could be compensated. And a vented box has multiple peaks.

 

[bwaslo]

Compensate with lossy series resonant RLC circuit, configured in shunt across the speaker terminals. Requires pretty high inductances to do that, though (but they can be made of thin wire, since resistance is being added anyway).

More practical (and what I did most recently) is to make one low-Q RLC (but with the resistor across the capacitor) to compensate both, also tamed the impedance in the midrange. Of course that doesn't perfectly compensate, but brings the impedance peaks down from the 20+ ohm range to about 6 ohms, which made all the difference in the bass when using a SET amplifier.