[Steve Eddy]

Hi Steve,

Boy am I outta my league now.   

You said, "For any given tap, the amp will deliver a varying amount of power to the loudspeaker depending on the speaker's impedance curve, no differently than a typical solid state amp."

Okay, looks like was I right about the solid state amp's power output being modulated by the speaker's impedance curve.

Well sure. That's just basic Ohm's Law.

But my point was that a tube amp's power output is also modulated by the speaker's impedance curve. That's because even though a typical tube amp has a bit higher output impedance than a typical solid state amp, both amps are effectively voltage sources.

But.... what you're saying here implies that a loudspeaker driven by a tube amp, and that same loudspeaker driven by a solid state amp, will have the same frequency response - right?

No.

Since the typical tube amp has a bit higher output impedance, then its frequency response won't be quite as flat as that of a typical solid state amp.

But that doesn't mean that the tube amp is behaving as a constant power source.

se

[Duke]

Thanks for writing back, Steve.

If I understand your post correctly, tubes are voltage sources just as transistors are, but tube amps typically have a significantly higher output impedance which is what's producing the frequency-response-modifying effects I've observed.

Correct me if I'm wrong here, but my understanding is that the higher output impedance of a tube amp works in the opposite direction of the tube's voltage source characteristic, tending to increase (rather than decrease) the power output into a high impedance load.  So while a tube amp may not literally be a power source, under the right conditions it can mimic the characteristics of one to a certain extent - is this correct?

And when you say that a solid state amp has "flat frequency response", that's referring to voltage output, not power output - right?  So of course a tube amp that has something approximating constant-power output will not have a "flat frequency response" with respect to voltage - but it might with respect to power, at least over a given impedance range. 

The extreme of high output impedance's effect on frequency response would be what we see with a current source approximating amp, such as Nelson Pass has been building recently - which behaves essentially the exact opposite of a voltage source amp.  Maybe its output devices are not really current source devices, but the amp still approximates the characterstics of a current source because its output impedance is so high - but I don't know.

Duke

[Steve Eddy]

Thanks for writing back, Steve.

If I understand your post correctly, tubes are voltage sources just as transistors are, but tube amps typically have a much higher output impedance which is what's producing the frequency-response-modifying effects I've observed.

Yes. As I said, it's just basic Ohm's Law.

Correct me if I'm wrong here, but my understanding is that the higher output impedance of a tube amp works in the opposite direction of the tube's voltage source characteristic, tending to increase (rather than decrease) the power output into a high impedance load.

All else being equal, no.

If each amp is set to have an unloaded output voltage of 2.83 volts, then the power delivered by the tube amp into the loudspeaker will be less than that of the solid state amp.

So while a tube amp may not literally be a power source, under the right conditions it can mimic the characteristics of one to a certain extent - is this correct?

Well, if you've got an output impedance which was such that it was neither a very good voltage source nor a very good current source, I suppose you could say it would behave kinda sorta somewhat like a power source in that the power delivered to the speaker at 32 ohms versus 4 ohms wouldn't vary quite as much as it would compared to a voltage source. But I wouldn't call that a power source any more than I'd call it a current source.

And when you say that a solid state amp has "flat frequency response", that's referring to voltage output, not power output - right?

Yes.

So of course a tube amp that has something approximating constant-power output will not have a "flat frequency response" with respect to voltage - but it might with respect to power.

Well, you might say that it has a flatter response with respect to power than a solid state amplifier with a much lower output impedance. But by the same token, the loudspeaker won't have as flat an acoustic response as when driven by the solid state amp.

The extreme of high output impedance's effect on frequency response would be what we see with a current source approximating amp, such as Nelson Pass has been building recently - which behaves essentially the exact opposite of a voltage source amp.

Yes, a current source is essentially the inverse of a voltage source.

In any case, I'm wondering just what your point is in all this?

se

[Duke]

Once again I appreciate your taking the time, Steve.

Here is my point:  It's not just "tubes vs. solid state".  Speaker choice matters - most speakers stack the deck in favor of one type, but some stack the deck in favor of the other.  Because most speakers are designed for constant-voltage amplifiers, those that are especially "tube-friendly" will  be those that have an exceptionally smooth impedance curve.  A high average impedance by itself is not enough to to make a speaker "tube-friendly"; smoothness is even more important.

There are situations in which an amplifier that "behaves kinda sorta somewhat like a power source" would be useful.  One example would be with a fullrange electrostat whose impedance curve is a perhaps around 32 ohms (or more) in the bass region and around 4 ohms (or less) in the high treble.   

Also, there might be situations where an amplifier that behaves kinda sorta somewhat like a current source would be useful.  An example would be a high-efficiency speaker with an impedance curve that is very smooth everywhere except in the bass region, where a current source amp's increased power into the bass impedance peaks could be taken advantage of (via appropriate tuning) to extend the bass considerably deeper than it would otherwise have gone. 

Rather than just say "some speakers work better with solid state and some work better with toobs", I hope our discussion throws a little "why" on the subject.

Duke

Second edit - I see that I was modifying while you were typing.  Sorry 'bout that!

[Steve Eddy]

Once again I appreciate your taking the time, Steve.

No problem!

I guess my point would be this:  There are situations in which an amplifier that "behaves kinda sorta somewhat like a power source" would be useful.  One example would be with a fullrange electrostat whose impedance curve is a perhaps around 32 ohms in the bass region and around 4 ohms (or less) in the high treble.

Can't say I'm terribly knowledgeable when it comes to electrostats, so don't mind my asking why that would necessarily be useful? 

Also, there might be situations where an amplifier that behaves kinda sorta somewhat like a current source would be useful.  An example would be a high-efficiency speaker with an impedance curve that is very smooth everywhere except in the bass region, where a current source amp's increased power into the impedance peaks could be taken advantage of to extend the bass considerably deeper than it would otherwise have gone.

Isn't that what we used to have tone controls for?

Seriously though, why not just use better drivers and design the system properly?

se

[Freo-1]

Here is an article that lays out some deltas between Tubes and SS

http://www.spectrum.ieee.org/archive/1640

Both can sound terrific or terrible. IMHO, I generally prefer a well designed tube amp setup over solid state (vocals and instruments sound more lifelike, especially with Electrostatics).  It truly depends on the relationship of a given speaker system to the amplifier.

[Duke]

Steve,

The kinda-sorta-somewhat "power source" characteristic would be useful with an electrostat because the high impedance at low frequencies cheats a voltage-source-amp out of much of its rated power.   In the example given - 32 ohms in the bass region - a 200 watt amp will clip at only 50 watts into that 32 ohm load, so we've just lost 6 dB of headroom.   Also many electrostats tend to sound "lean" with solid state amps, but more "full-bodied" or "well-balanced" with tube amps having a fairly high output impedance.

I designed the high-efficiency speaker used as the other example, and the parts quality is actually not too bad.  Dunno if this will mean anything, but the woofer is a Pioneer TAD TL-1102 alnico magnet prosound woofer, list price $630 each.  The system can be tuned to 35 Hz or so and it has bass down into the mid-30's with a solid state amp.  With a high-output-impedance OTL amp (Atma-Sphere S-30, output impedance approximately 7 ohms) I get an extra 6 dB(!) into the bass impedance peaks compared with a solid state amp because of the OTL's kinda-sorta-somewhat current source characteristic (can you tell I like your wording??).  So I tune the speaker down around 27Hz, and get bass extension into the upper 20's.   

How feasible would it be to build a solid state amp with widely variable output impedance?

Duke

[Bob Reynolds]

Having thus made my disclaimer, next I'll concede that insofar as the voltage amplitude vs frequency output of a solid state amp is independent of the load, it's fair to say that the amp's frequency response - referenced to voltage! - is also independent of the load.  However, its power vs frequency response is definitely load-dependent! 

Yes, no debate there!

And if it is power (wattage) that drives a loudspeaker, then the frequency response of the voltage-source amplifier + loudspeaker combination is indeed sensitive to the loudspeaker's impedance curve, because that impedance curve is modulating the amplifier's wattage output, and it is wattage into the speaker at a given frequency which determines how loud the speaker plays at that frequency.

That's where I have a problem. Why do we think it's power that drives a loudspeaker? Isn't it the current flowing through the voice coil that creates the magnetic field that causes the driver to move? The loudspeaker inherently dissipates different amounts of power based on its impedance. I'm not following why power is a more meaningful metric than voltage.

Since P = I^2 * R and R varies, then I must also vary accordingly to keep P constant.  Or, since P = E^2 / R and R varies, then E must also vary accordingly to keep P constant. How do these constant power amps vary either I or E  appropriately for any R (loudspeaker)?

I think it's justifiable to focus on wattage (not voltage) into the loudspeaker because wattage is the most precise and direct description of the electrical energy going into the speaker that is then converted (or transduced) into acoustic energy - and that acoustic energy can also be expressed in watts. 

Wattage is the amount of power (energy consumed per unit time) dissipated by the loudspeaker and since power is a function of voltage, current and impedance I don't know why its any better than voltage.

I think I could back up my position by making frequency response measurements of a speaker driven by a voltage-source-approximating amplifier and a power-source-approximating amplifier, assuming the latter maintains that characteristic over the impedance fluctuation range of the loudspeaker.

Yes, I agree the graphs will look different, but I'm not sure that one is any more meaningful than another.

Let me know if you see any obvious (or subtle) mistakes that I'm making. 

Thanks,

Duke

Will do. Thanks for contributing to this thread. I've been curious about this topic for a while.

[Bob Reynolds]

Yes, a typical tube amp is rated at approximately the same power on the 2 ohm tap as on the 4 ohm tap as on the 8 ohm tap, etc. But that's not the same as delivering constant power regardless of the speaker's impedance curve.

For any given tap, the amp will deliver a varying amount of power to the loudspeaker depending on the speaker's impedance curve, no differently than a typical solid state amp.

se

Steve, please explain why tube amps use output transformers and why they typically have a few output taps? Intuitively, I've always assumed that the transformer was there to "match" the impedance of the amp to the loudspeaker. So, you'd use the 4 ohm tap when driving a "4 ohm" speaker. But, I really do not understand the benefits of matching the impedance.

[Bob Reynolds]

If I understand your post correctly, tubes are voltage sources just as transistors are, but tube amps typically have a significantly higher output impedance which is what's producing the frequency-response-modifying effects I've observed.

Isn't it the case that all audio amps are voltage amplifiers? The gain of an amp is determined by the ratio of its input and output voltages.

Since the output impedance of an amp is modeled by a resistor in series with the loudspeaker, the higher that resistor the greater the voltage drop across it. As the value of that resistor approaches the magnitude of the impedance of the speaker, it becomes a significant determinant of the entire circuit. The output impedance of a typical solid state amp is so small relative to the impedance of the speaker that its contribution to the behavior of the entire circuit is negligible.

[Bob Reynolds]

If each amp is set to have an unloaded output voltage of 2.83 volts, then the power delivered by the tube amp into the loudspeaker will be less than that of the solid state amp.

Steve, please explain why this is so? I think I know why, but an explanation would be useful.

Well, you might say that it has a flatter response with respect to power than a solid state amplifier with a much lower output impedance. But by the same token, the loudspeaker won't have as flat an acoustic response as when driven by the solid state amp.

That's an excellent point, could you expand on that?

[Steve Eddy]

Here is an article that lays out some deltas between Tubes and SS

Thanks. I'm familiar with that article.

Bottom line is that in terms of what people like and what they prefer, which in my opinion is really the only thing that counts at the end of the day, there aren't so many absolutes. Ultimately, there will always be those who prefer tubes and those who prefer solid state.

se

[Steve Eddy]

The kinda-sorta-somewhat "power source" characteristic would be useful with an electrostat because the high impedance at low frequencies cheats a voltage-source-amp out of much of its rated power.   In the example given - 32 ohms in the bass region - a 200 watt amp will clip at only 50 watts into that 32 ohm load, so we've just lost 6 dB of headroom.

Have we? Who says the speaker necessarily needs more than 50 watts at the low end where the impedance is 32 ohms in order to have an appropriate acoustical frequency response?

Also many electrostats tend to sound "lean" with solid state amps, but more "full-bodied" or "well-balanced" with tube amps having a fairly high output impedance.

Ok. But is that necessarily due to the amplifier's higher output impedance?

I designed the high-efficiency speaker used as the other example, and the parts quality is actually not too bad.  Dunno if this will mean anything, but the woofer is a Pioneer TAD TL-1102 alnico magnet prosound woofer, list price $630 each.  The system can be tuned to 35 Hz or so and it has bass down into the mid-30's with a solid state amp.  With a high-output-impedance OTL amp (Atma-Sphere S-30, output impedance approximately 7 ohms) I get an extra 6 dB(!) into the bass impedance peaks compared with a solid state amp because of the OTL's kinda-sorta-somewhat current source characteristic (can you tell I like your wording??).  So I tune the speaker down around 27Hz, and get bass extension into the upper 20's.

Sounds like you're talking about simple damping here. If you drive a typical dynamic loudspeaker with a higher impedance source, you effectively increase the driver's Q which will have an effect on low frequency response.

How feasible would it be to build a solid state amp with widely variable output impedance?

Sure. Just get a high current switch and some power resistors to switch in series with the amplifier's output.

se

[Steve Eddy]

Steve, please explain why tube amps use output transformers and why they typically have a few output taps? Intuitively, I've always assumed that the transformer was there to "match" the impedance of the amp to the loudspeaker. So, you'd use the 4 ohm tap when driving a "4 ohm" speaker. But, I really do not understand the benefits of matching the impedance.

Sure.

One reason output transformers are used is because generally a tube can't supply sufficient current into a low impedance load to get any appreciable power. That's largely due to the fact that tubes generally run at very high voltages; a couple hundred volts and more.

For example, to deliver 1 watt of power into 8 ohms, you need to deliver about 350mA of current. Let's say the tube's running at 300 volts. 350mA times 300 volts is over 100 watts.

But let's say the load is 2,000 ohms instead of 8 ohms. 1 watt of power into 2,000 ohms is only about 22mA and at 300 volts, you're only looking at about 6 watts.

So, how can we take that 1 watt we've got into the 2,000 ohm load and transfer it to our 8 ohm speaker load?

We can use a transformer.

To get that 1 watt into 2,000 ohms, the tube is swinging 44 volts. And in order to get that down to 2.83 volts for our 8 ohm load, we can use a transformer with a step-down ratio of about 16 to 1.

And while transformers will step voltages up or down as a function of their turns ratio, they reflect impedances by the square of their turns ratio. So for a transformer with a 16:1 voltage/turns ratio, impedances get reflected by a factor of 16 squared or about 256.

And if we multiply 8 ohms by 256, we get about... 2,000 ohms.

In a typical tube amp, say a single-ended triode amp, the tube's output (the plate in this case) is connected to one end of the output transformers primary and the other end of the transformer's primary is connected to the power supply (or B+). So the load the tube sees is really the loudspeaker's impedance as reflected to the transformer's primary, or in this example, 2,000 ohms.

The reason for the multiple taps on the transformer is because a given tube works most linearly into a given load. So in order for the tube to see the roughly same load whether you're using a 4, 8, or 16 ohm speaker etc., multiple taps are wound with different ratios.

For example, let's say the speaker load is 4 ohms instead of 8. To get the same 2,000 ohms reflected to the transformer's primary, we need a step-down ratio not of 16 to 1 as in the 8 ohm case, but about 22 to 1 which means the 4 ohm loudspeaker load is reflected back to the primary by a factor of 22 squared or about 500.

Of course this means that the 44 volts the tube is swinging into the 2,000 ohm primary load gets stepped down by a factor of 22, so instead of 2.83 volts across the loudspeaker, it's just 2 volts. And 2 volts into 4 ohms is... one watt. Same as for the 8 ohm tap with an 8 ohm speaker.

And that's why tube amps are typically rated for the same power into 4, 8, 16 ohms, etc.

This help?

se

[Steve Eddy]

If each amp is set to have an unloaded output voltage of 2.83 volts, then the power delivered by the tube amp into the loudspeaker will be less than that of the solid state amp.

Steve, please explain why this is so? I think I know why, but an explanation would be useful.

It's simply from the loss due to the higher output impedance.

Take an ideal voltage source set for 2.83 volts. Put it across an 8 ohm load and you've got 1 watt into that 8 ohms.

Do the same but with say a 2 ohm resistor in series with the voltage source. Now you've effectively got 2.83 volts across 10 ohms. And instead of 354mA of current, you end up with 283mA of current. 283mA through 8 ohms is only 0.64 watts.

Well, you might say that it has a flatter response with respect to power than a solid state amplifier with a much lower output impedance. But by the same token, the loudspeaker won't have as flat an acoustic response as when driven by the solid state amp.

That's an excellent point, could you expand on that?

It's simply that if the amplifier's frequency response (in terms of voltage) isn't flat, then the output from the speaker will be altered compared to driving it from a lower output impedance.

Saying it wouldn't have as flat an acoustic response perhaps wasn't the best choice of words as the speakers response could be all over the road depending on the speaker. The point however was that no matter what its response, that response will be altered due to the higher output impedance.

se

[Duke]

Hi Bob,

I can see how your argument for current as being what drives a loudspeaker makes sense, but I don't think the measurement of current (or voltage) will tell you how loud a speaker will play as directly as will the measurement of wattage.

Let's say we have three speakers of 1% efficiency.  One is a 4-ohm speaker; one is an 8-ohm speaker; and one is a 16-ohm speaker.

How loud will each play with a 1-watt input?  92 dB; 92 dB; and 92 dB.

How loud will each play with a 2.83 volt input?  95 dB; 92 dB; and 89 dB.

How loud will each play with a .3535 amp input?  89 dB; 92 db; and 95 dB.

With voltage or current, we have to convert to watts to see how loud the speaker will play - which to me means that watts is the most accurate metric for describing the energy that drives the speaker.  Also as I mentioned before, a loudspeaker transduces electrical watts into acoustic watts.  We don't speak of acoustic volts or acoustic amps.

Duke

[Duke]

Hi Steve,

Let me see if I get what you're saying (pardon me if I'm slow to grasp the obvious, or on the other hand quick to miss the obvious): 

I think you're saying that the increased output at very low frequencies that I observed with the OTL amp is due to the roughly doubled effective woofer Qes arising from the amp's high output impedance, rather than due to the amplifier actually putting out more wattage into the impedance peaks (in what would be a kinda-sorta-somewhat current source-ish way). 

Am I understanding you correctly?  Kinda? 

If so, I have an idea for a test.

Thanks,

Duke

[Steve Eddy]

I can see how your argument for current as being what drives a loudspeaker makes sense, but I don't think the measurement of current (or voltage) will tell you how loud a speaker will play as directly as will the measurement of wattage.

It seems to me that you're trying to make all of this more complicated than it needs to be.

With voltage or current, we have to convert to watts to see how loud the speaker will play...

Why do we have to do that?

Since virtually all power amplifiers are more or less voltage sources, and since virtually all loudspeakers are designed to be driven by a voltage source, why not simply use a voltage reference (i.e. 2.83 volts) as we commonly do now?

Again, I just don't see the "problem" here that you're apparently trying to address.

se

[Steve Eddy]

Let me see if I get what you're saying (pardon me if I'm slow to grasp the obvious, or on the other hand quick to miss the obvious): 

I think you're saying that the increased output at very low frequencies that I observed with the OTL amp is due to the roughly doubled effective woofer Qes arising from the amp's high output impedance, rather than due to the amplifier actually putting out more wattage into the impedance peaks (in what would be a kinda-sorta-somewhat current source-ish way). 

Am I understanding you correctly?  Kinda?

Yes.

If so, I have an idea for a test.

A test for what exactly?

se

[Duke]

Steve,

What I'd like to test for is whether the OTL amp has anything resembling approximate power source or current source characteristics. 

I'd measure a single driver driven by the OTL amp at a fixed input voltage, and then measure two such drivers wired in series.   If the amp has a voltage-source-plus-series-resistance characteristic, the broadband SPL should be unchanged.  If for some reason the amp kinda sorta somewhat approximates a power source or even a current souce, that should be obvious from a significant increase in SPL with the two drivers wired in series.  I could test the parallel connection also, anticipating less than the voltage-source-predicted 6dB increase in SPL.

I could run the same test with a solid state amp as a control.

Assuming I keep all input levels the same and take care with microphone setup and use decent measuring equipment, does this sound like a reasonable test?

Duke