Hi JPV,
In connection with reactive load, I believe that on occasion I have seen some evidence of increased THD with reactive loads in some amplifier reviews in Stereophile. Those reactive loads were the simulated loudspeaker load usually used to load the amplifier in the frequency response plots.
I'm wondering if anyone else here has seen that effect, either in measurement or simulation.
Cheers,
Bob
What would be interesting is to measure the crossover distortion in the pressure wave taking then into account the phase of the system. Of course it is impossible to separate this from the distortion generated by the speaker itself.
An interesting experiment is to simulate with spice a closed box and a vented box ( distortion free), connect them to the simulated amplifier with crossover distortion). In this way the influence of the phase shift on the level of crossover distortion in the pressure wave can be measured. These two systems have radically different phase shifts.
I am waiting for your book to start finally to experiment in depth with spice.
Leach has some good spice models for different speaker systems, it is rather straightforward but good to have on hand
JPV
didn't Cherry look at "interface distortion" - and conclude that there really wasn't anything surprising there?
basically with common power amp's distortion dominated by output Q behavior with current load - even if reactive loads draw more current at different V the distortion isn’t substantially different than what would be with resistive load drawing the same current at audio frequencies
so sizing output, driver, ect. for the peak reactive current should suffice - including allowing for the higher thermal load on the outputs driving reactive loads
basically with common power amp's distortion dominated by output Q behavior with current load - even if reactive loads draw more current at different V the distortion isn’t substantially different than what would be with resistive load drawing the same current at audio frequencies
so sizing output, driver, ect. for the peak reactive current should suffice - including allowing for the higher thermal load on the outputs driving reactive loads
After quick review of Cherry's paper it seems to me that he is addressing the interface intermodulation problem. This is different. Some have stated that the load beeing an emf generator, it is a kind of active circuit and could inject signals in the output that can intermodulate with the signal of the input on any internal nonlinearity.
Cherry has shown that the distortion generated by intermodulation of two signals injected at the input is alway greater than the intermodulation of the same signals one injected at the input and the other at the output.
The back emf source is not an independent thevenin source but a dependent one and is totally equivalent to an RLC circuit representing the load. Therefore there is not a special intermodulation phenomenon that would be worse than the one excited by two input signals.
But injecting a signal at the input and another at the output we could drive the amplifier in all the operating points ( Vce, Ic) that an arbitrary signal with any reactive load could reach and it is a valuable test method providing you go through all the realistic points
He mention ( en passant ) that these operating points outside the resistive locus could generate parasitic oscillations, triggering of protection circuits or accentuation of crossover distortion and this is my point.
If the accentuation is even more on the pressure loudspeaker output signal for phase reasons, then crossover can be more audible.
This is only hypothetical
JPV
Hi JPV,
Simulating an amplifier's distortion with simulated loudspeaker loads is a good idea. I show a simulation model of a simple closed-box woofer in my book, but I don't go as far as a vented one, which would be interesting.
Indeed, one could put together a simulation model of a complete three-way speaker system with a fairly complex crossover.
In the real (measurement) world, one can synthesize a loudspeaker load using a second (back-feed) amplifier and active filter techniques. I say a bit about this in the book.
There is also another approach that I mention in the book. It rests on the observation that the phase angle of a load just corresponds to the voltage and current excursions being out of phase. If a second amplifier is used to back-feed an amplifier under test through a "load" resistor, and the back-feed amplifier is fed with a slightly different frequency, a beat frequency will result that will cause the output voltage of the amplifier under test and the current to go through different phase relationships. In a sense, you can think of the back-feed amplifier as emulating the counter-EMF of a loudspeaker.
Looking at the THD residual in the output voltage of the amplifier under test as a function of time could be very revealing.
Cheers,
Bob
Hi,
I also orderd my book at amazon.uk and delivery is planned early dec. Reading JPV's question brought up things I been thinking about.
It's often said that measuring thd with resistive load gives the best results.
If a the load is inductive at higher frequenses, the loading is less where the feedback starts to decrese, and can give less distorsion. I saw this first in an JAES engineering report by Paul Klipsch (JAES july/aug 78 p547).
If this is true then it would be interesting to know if crossover distortion is less with such a load.
Then for crossover distortion with reactive load. If the crossover glitch is offseted would there be a masking effect? The only listening test I read is Geedes and Lee (http://www.gedlee.com/downloads/Distortion_AES_I.pdf) and I'm not sure if they covered that case. But they say that nonlinearities at lower signal-levels are more audible and if the glitch is moved does that change masking? Even if the size of the glich itself is the same.
/örjan
I also orderd my book at amazon.uk and delivery is planned early dec. Reading JPV's question brought up things I been thinking about.
It's often said that measuring thd with resistive load gives the best results.
If a the load is inductive at higher frequenses, the loading is less where the feedback starts to decrese, and can give less distorsion. I saw this first in an JAES engineering report by Paul Klipsch (JAES july/aug 78 p547).
If this is true then it would be interesting to know if crossover distortion is less with such a load.
Then for crossover distortion with reactive load. If the crossover glitch is offseted would there be a masking effect? The only listening test I read is Geedes and Lee (http://www.gedlee.com/downloads/Distortion_AES_I.pdf) and I'm not sure if they covered that case. But they say that nonlinearities at lower signal-levels are more audible and if the glitch is moved does that change masking? Even if the size of the glich itself is the same.
/örjan
Copyright 2011??? Did I miss a few months somewhere?
Steve.
Hi Steve,
Your eagle-eyed observation made me curious, so I checked. It turns out that it is customary in the publishing industry to use the following year's date for anything published after the end of August - sort of like with the automobile industry.
Cheers,
Bob
Yes, Ed Cherry took a look at Interface Intermodulation Distortion (IIM). With the IIM test, a 7 kHz HF signal is driven forward in the amplifier while a 60 Hz signal was back-driven into the output of the amplifier under test by a second amplifier. This is similar to the SMPTE IM test where both signals are driven forward. This is not quite the same thing as concern about crossover distortion being exacerbated by driving a reactive load. There were no real surprizes with the IIM investigations apart from the heightened awareness that amplifiers needed to be able to deliver high current into reactive speaker loads under some conditions. The assertion that NFB was a villain in IIM creation was disabused.
Cheers,
Bob
On the same subject, there is an interesting paper in the AES from W. Sansen who is expert in NL distortion in electronic circuits.
In this paper, he calculates the distortion components for an amplifier with a second order voltage non linearity driving a loudspeaker with a second order
NL impedance and a current source, this for open loop and closed loop
It is shown and it is rather obvious that there is intermodulation distortion only if there is an independent current source in the load, a depended one will simply enhance the harmonic distortion coefficients. Intermodulation distortion will exist even if there is no impedance NL in the load in case of independent current source.
The back emf and the blocked impedance model of a loudspeaker can be replaced by a simple LRC circuit. The back emf is a dependent source and will not generate intermodulation.
An independed source could be created by external mechanical vibrations like acoustical refelections, pressure waves in the surrounding or heavy overload.
A test was done where the system was connected to a 100hz input signal and the output current in the load was analyzed with a selective voltmeter tuned at the mechanical resonance. If an external vibration is acting, there must be a trace of a current at the resonance of the system.
The experience was negative and the conclusion is that there is no independent current source so no interface intermodulation even under severe overload which is very interesting.
JPV
JPV
JPV,
Do you have a dat or something? The only AES paper by Willy I can find is:
Dynamic Range Limitations of Optical Couplers
Edit: found it:
Interface Intermodulation Distortion (IIM) in Power Amplifiers
This paper deals with intermodulation distortion (IIM) which is cased by loudspeaker-induced error signals. It is shown that the loudspeaker must fulfill certain conditions in order to produce IIM-distortion. Experiments with a real loudspeaker show that these conditions are not fulfilled: loudspeakers are unlikely to produce IIM-distortion.
Authors: Corveleyn, L.; Bossuyt, F.; Sansen, W.
AES Convention:71 (March 1982) Paper Number:1869
yes?
jan didden
Last edited:
Hi JPV,
I also did an AES conference paper on IIM, which is available on my website at CordellAudio.com - Home (and discussed a bit in my book). It shows actual IIM measurements (using Otala's IIM test) done on an amplifiers with high open-loop output impedance and low open-loop output impedance, otherwise identical. The one with high feedback and high open-loop output impedance did not exhibit higher IIM.
Low and linear closed-loop output impedance is what minimizes IIM, as long as the amplifier has adequate output current capability. From the point of view of establishing low and linear closed-loop output impedance, doing it with high open loop output impedance that is reduced by high NFB is just as good as doing it with low starting open-loop output impedance reduced to the same value by a smaller amount of negative feedback.
Cheers,
Bob
Hi Andrew,
Wow! To Scotland! Now we're cookin'.
When you say interviews, are you referring to the various threads here on DIYaudio that had my name on them? If so, they most certainly did. Answering questions and participating in discussions here on DIYaudio was very helpful in identifying topics that should be covered and in me refining my views on them. There are actually a couple of plots in the book which are plots that I actually created to explain some issues here on DIYaudio.
Thanks are due to all those who participated in discussions here.
Cheers,
Bob
What a conventional amplifier ( Vin x gain = Vout ) can not supress is distortion in the CURRENT that comes back from the loudspeaker. I found with the Klippel Analyser that a speaker with an Alnico magnet had much less distortion in the current then the same speaker with a ferrite magnet. In both cases the amplifier was connected with the speaker and the distortion of the current was measured with a current sensor. Both speakers showed very low distortion in the voltage.