Try testing a power amp at full power with a two way loudspeaker connected. Now do it at a frequency that is above the tweeter crossover frequency. How long can the tweeter take that?
You need a bursted signal, and an analyzer that can test the harmonic distortion of the burst. It is not a steady state measurement.
The price of the test equipment is out of reach of some DIY'ers. That is the "hard part".
If I still worked for the test and measurement company that I used to, I might have made this measurement by now.
I have not seen any literature that shows the results of such a test with a real loudspeaker.
I previously did measure phase and frequency response of single ended non fed back triode amplifiers when connected to a loudspeaker, but not at large signal levels.
It is nice to have a closed anechoic chamber for large signal testing.
Loudspeakers have the additional effect that their impedance can change at large signal
levels (due to several factors).
You need a bursted signal, and an analyzer that can test the harmonic distortion of the burst. It is not a steady state measurement.
The price of the test equipment is out of reach of some DIY'ers. That is the "hard part".
If I still worked for the test and measurement company that I used to, I might have made this measurement by now.
I have not seen any literature that shows the results of such a test with a real loudspeaker.
I previously did measure phase and frequency response of single ended non fed back triode amplifiers when connected to a loudspeaker, but not at large signal levels.
It is nice to have a closed anechoic chamber for large signal testing.
Loudspeakers have the additional effect that their impedance can change at large signal
levels (due to several factors).
When you think about it, a low output impedance is no more elegant a solution than say, backing up a lower PSRR with a stiff supply. The speaker swamps the amp. Both are brute force scenarios.
Problems are there to be solved. To call a well made SET amp poor engineering you need to mention size, price or efficiency.
Problems are there to be solved. To call a well made SET amp poor engineering you need to mention size, price or efficiency.
If the problem is variations in speaker impedance then the simplest most elegant solution is a low output impedance. Curiously, this is the solution which has been adopted by almost all audio designers ever since the issue was understood. I would not characterise simple elegance as being a "brute force" solution.
I'm not sure what connection you see between low PSRR (requiring a low ripple supply) and a stiff supply (i.e. low DC and subsonic output impedance). Low ripple and low output impedance are different PSU characteristic, although some circuit architectures deliver both.
Although SET is in itself a poor engineering choice (IMHO), having made that choice (for whatever reason) it is not necessary to continue to make other poor engineering choices yet some SET designers do (e.g. poorly designed PSUs). Not all do, of course, so some people can make the best of a weak architecture. I don't mention size, price or efficiency because people concerned about those will not choose SET.
I'm not sure what connection you see between low PSRR (requiring a low ripple supply) and a stiff supply (i.e. low DC and subsonic output impedance). Low ripple and low output impedance are different PSU characteristic, although some circuit architectures deliver both.
Although SET is in itself a poor engineering choice (IMHO), having made that choice (for whatever reason) it is not necessary to continue to make other poor engineering choices yet some SET designers do (e.g. poorly designed PSUs). Not all do, of course, so some people can make the best of a weak architecture. I don't mention size, price or efficiency because people concerned about those will not choose SET.
I noticed that S.E.T. are adequate with high efficiency alnico speakers and even sound better than their low impedance transistors counterparts paired with such speaker.
However a S.e.t. amplifier with typical modern speakers which are not alnico magnet and high efficiency results in a highly romanticized and poor bandwidth sound.
Any big bottle 815, 845 S.e.t. paired with cheap single cone or lampizator type speakers will sound incredible.However very poor sounding and annoying peaks and limitations vs. modern B&W or Harbeth speakers driven with let say Rotel or Creek.
However a S.e.t. amplifier with typical modern speakers which are not alnico magnet and high efficiency results in a highly romanticized and poor bandwidth sound.
Any big bottle 815, 845 S.e.t. paired with cheap single cone or lampizator type speakers will sound incredible.However very poor sounding and annoying peaks and limitations vs. modern B&W or Harbeth speakers driven with let say Rotel or Creek.
Last edited by a moderator:
I agree in that SET amps without global feedback do require a good speaker marriage. I use a triode strapped sweep tube that has an rp~450 at my operating point, I run it with a 3k load so I am not going for power but low distortion and decent damping. The only reason why I converted to such an amp is the fact I acquired a pair of Klipsch LA Scala's which I think sound marvelous together. Since the woofer doesn't go much below 40Hz I run a subwoofer for the low end that you can really only feel.
My ears smile everyday.
My ears smile everyday.
What is a typical Zout for a tube SE amp?
I just built my first SE amp, which was fairly unusual in that I drove the transformer from a mosfet source follower operating at load, supply, and idle current conditions typical for a 300B amp. I got a 0.6 Ohm Zout and that probably sits pretty close to a lower bound for Zout of an SE amp. That was with an Edcor 25W 5k transformer.
Some back-of-the-envelope calculations reveal that you may be able to get Zout down to ~1.4 Ohms with a low-rp DHT with that same transformer. Is that the neighborhood where SE amps tend to sit WRT Zout (1.4 Ohm and up)?
My experience with the limited number of speakers that I have owned is that they sounded pretty good when driven by a 1.4 Ohm source, or even higher. I wish I owned the equipment to quantify the difference this higher Zout makes over a lower Zout in the distortions (linear and non-linear) radiated into the air. Are there any papers/articles out there that explore this that anyone can recommend?
At what point is Zout "good enough?"
I just built my first SE amp, which was fairly unusual in that I drove the transformer from a mosfet source follower operating at load, supply, and idle current conditions typical for a 300B amp. I got a 0.6 Ohm Zout and that probably sits pretty close to a lower bound for Zout of an SE amp. That was with an Edcor 25W 5k transformer.
Some back-of-the-envelope calculations reveal that you may be able to get Zout down to ~1.4 Ohms with a low-rp DHT with that same transformer. Is that the neighborhood where SE amps tend to sit WRT Zout (1.4 Ohm and up)?
My experience with the limited number of speakers that I have owned is that they sounded pretty good when driven by a 1.4 Ohm source, or even higher. I wish I owned the equipment to quantify the difference this higher Zout makes over a lower Zout in the distortions (linear and non-linear) radiated into the air. Are there any papers/articles out there that explore this that anyone can recommend?
At what point is Zout "good enough?"
Brute force: by keeping the output impedance an order of magnitude lower than the speaker, signal Voltage dropped across the amp becomes relatively insignificant. In the first order either is acceptable, as comparing to conjugating the load when a higher Zo is present.
In the second order, not fixing the issue at the source (the speaker), residual interactions may be susceptible to the non-linearity of the output impedance, (plus the magnitude of Zo becomes a design constraint). It is residual interactions such as this that drive the choices some people make, as witnessed I suppose by those who build class A push-pull amps without feedback, and who still 'overbuild'.
Curiously, well, there are a number of reasons.
That supply Voltage variations can be reduced by lowering the supply impedance just as output Voltage variations can be reduced by lowering Zo. Unrelated but furthermore I'd submit that supply variations with a single ended stage are of less concern due to the linear nature of the relation (when the supply source impedance isn't complicated by resonance).
Quote:
"You are talking about speaker measurements, not amp."
I was not talking about speaker measurements.
I was not talking about amplifier measurements.
But I was talking about System measurements.
The system was an amplifier and a loudspeaker together.
There was no microphone used, this was not an acoustic measurement.
I used a Vector Network analyzer to measure the difference
of the signal into the amplifier versus the signal output of the amplifier.
(Strictly electrical input to output)
The measurement was from 10 Hz to beyond 20 kHz.
For the first measurement, an 8 Ohm load resistor was used.
For the second measurement, the resistor was removed and a 2 way loudspeaker was connected as the load.
The frequency response and phase of the two measurements were compared.
The amplifier output stage was a single ended triode, non feedback, and had a damping factor of about 3 to 4.
The loudspeaker was a relatively easy load.
Although there were not a lot of large differences, it was easy to see the difference between the resistive and loudspeaker loads as they reflected back on the amplifier output.
Not only is the loudspeaker a complex impedance at many frequencies,
the amp output transformer has primary inductance, distributed capacitance, and
leakage reactance.
These transformer characteristics affect the triode driving it, and the output at the load resistor.
But the transformer characteristics interact even more when driving the varying complex load of a loudspeaker.
"You are talking about speaker measurements, not amp."
I was not talking about speaker measurements.
I was not talking about amplifier measurements.
But I was talking about System measurements.
The system was an amplifier and a loudspeaker together.
There was no microphone used, this was not an acoustic measurement.
I used a Vector Network analyzer to measure the difference
of the signal into the amplifier versus the signal output of the amplifier.
(Strictly electrical input to output)
The measurement was from 10 Hz to beyond 20 kHz.
For the first measurement, an 8 Ohm load resistor was used.
For the second measurement, the resistor was removed and a 2 way loudspeaker was connected as the load.
The frequency response and phase of the two measurements were compared.
The amplifier output stage was a single ended triode, non feedback, and had a damping factor of about 3 to 4.
The loudspeaker was a relatively easy load.
Although there were not a lot of large differences, it was easy to see the difference between the resistive and loudspeaker loads as they reflected back on the amplifier output.
Not only is the loudspeaker a complex impedance at many frequencies,
the amp output transformer has primary inductance, distributed capacitance, and
leakage reactance.
These transformer characteristics affect the triode driving it, and the output at the load resistor.
But the transformer characteristics interact even more when driving the varying complex load of a loudspeaker.
Then why this?
So, what's the problem?
Have you tried 4 Ohm load resistor?
I did too, using Firefox.
However, just copy the link location (rh mouse button) and paste it into Chrome...
http://www.keith-snook.info/wireless-world-magazine/Wireless-World-1950/Bass Without Big Baffles.pdf
For a conventional speaker (i.e. designed for voltage drive) I have heard DF=20 given as the point where further improvement achieves little. For nominal 8 ohms this means Zout no greater than 0.4 ohms. For other speakers Zout should match whatever the speaker designer intended - which in some cases merely means matching his amplifier - which in many cases will be unknown to the public.SpreadSpectrum said:
We obviously differ in our understanding of 'brute force'. If you want to feed a constant voltage to a load with varying impedance then low source impedance is the simplest way to do it. Alternatives are:AllenB said:
1. monitor the load voltage and adjust the drive - but this simply reduces output impedance by using feedback, so is merely an implentation of 'brute force' (according to you).
2. model the load and adjust the drive - this is a form of feedforward or predistortion and is only ever used in electronics where feedback is unusable for some reason.
I don't know what you mean by 'conjugating the load' in this context.
Yes, output impedance should be low and linear. If sufficiently low, it doesn't matter too much if it is a bit nonlinear. Typically, SS amps go for very low but rather nonlinear; conventional valve amps go for low and fairly linear; SET amps go for a bit higher and possibly a bit more nonlinear.
Making it resitive. Eg at a minimum, one series RLC across the speaker (with 2 drivers working together at a crossover each -6dB for half total power). I also choose to high pass at line level thus avoiding operating at the speaker fundamental resonance, which helps.
Whether that helps would depend on exactly how the speaker is designed. You could make the impedance vary more away from your design frequency; wideband 'conjugation' is always difficult and often impossible. Be aware that putting something across the speaker to flatten the impedance curve means you are accepting that the speaker expects something like a voltage drive so why not just give it a voltage drive?
Since this is an acoustical impedance variation in the two crossing drivers creating the need for less power, the symmetry of the electrical impedance peak depends largely on the drivers being used well enough in band and beginning with the same base impedance. Even still, assymetry only takes a few more components and at the end of the day, the impedance phase falls to zero as the magnitude is flattened. 'Difficult' is subjective and I'd never say impossible. I agree that voltage drive is desirable.
Last edited:
DF96 look at alnico magnetisation vs other magnetics and you can see how critical the different magnets affect the performance of audio.
Also maybe the demagnetization curves explains why the bass frequency is sounding very good with low impedance amplifiers on alnico.
A low impedance high feedback amplifiers will have to fight that force and kill the bass harmonics and impact, a low power tube amp will let the cone finish the motion and will give more 2n harmonics but a nice sound.
Also maybe the demagnetization curves explains why the bass frequency is sounding very good with low impedance amplifiers on alnico.
A low impedance high feedback amplifiers will have to fight that force and kill the bass harmonics and impact, a low power tube amp will let the cone finish the motion and will give more 2n harmonics but a nice sound.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.