I point out in #4331 of cfa-topology-audio-amplifiers-434.html that current drive only gives some of the advantages of motional feedback cos the amp doesn't "know" what the speaker is up to.
Hi all,
I have just noticed this thread and have not yet read all its messages. (I am not an expert in audio amplifiers or loudspeakers but find this topic interesting.)
In the 2nd part of his article: http://www.edn.com/design/consumer/...-of-current-drive-over-voltage-drive--Part-2-
the author says:
"In subwoofers, there is not a pressing need to migrate to current drive, as the principal flaws of voltage drive are forceful above the bass frequencies.
Aside from subwoofers, there is nothing that really supports the use of voltage drive...."
My question is:
Why not 'current-drive' for subwoofers as well? Why will a subwoofer NOT benefit from being driven by a current source (instead of a voltage source)?
(This question is because I think that, in comparison to woofers and tweeters, a subwoofer is the driver that needs most power - hence highest currents - from its own amplifier. Here, I am considering a system with an active crossover followed by separate amplifiers each driving its own dedicated speaker unit.)
Many thanks.
I have just noticed this thread and have not yet read all its messages. (I am not an expert in audio amplifiers or loudspeakers but find this topic interesting.)
In the 2nd part of his article: http://www.edn.com/design/consumer/...-of-current-drive-over-voltage-drive--Part-2-
the author says:
"In subwoofers, there is not a pressing need to migrate to current drive, as the principal flaws of voltage drive are forceful above the bass frequencies.
Aside from subwoofers, there is nothing that really supports the use of voltage drive...."
My question is:
Why not 'current-drive' for subwoofers as well? Why will a subwoofer NOT benefit from being driven by a current source (instead of a voltage source)?
(This question is because I think that, in comparison to woofers and tweeters, a subwoofer is the driver that needs most power - hence highest currents - from its own amplifier. Here, I am considering a system with an active crossover followed by separate amplifiers each driving its own dedicated speaker unit.)
Many thanks.
The link for 'the second part of the article' did somehow not appear in my previous message. Here is the link:
Loudspeaker operation: The superiority of current drive over voltage drive (Part 2) | EDN
Loudspeaker operation: The superiority of current drive over voltage drive (Part 2) | EDN
Richard, interesting that you see this as 'isolating the Vas from the back-EMF'.
I always looked at it this way: if the amp output tends to change in level due to the back EMF, any small deviation causes the feedback loop (due to the hopefully very high open loop gain) to counter-drive the input stage, Vas, driver, output stage to counter-act the deviation.
So, I would expect that I see somewhere in the amp, also in the Vas, a deviation from the input signal as it is counteracting the back EMF. Superficially you could think that the back EMF gets through the Vas through a badly-isolating output stage/driver, but actually it is the other way around: the back-EMF causes the Vas to counter it by steering the output stage in the opposite direction.
Not sure I am clear, though it is in my head 🙁
Jan
Last edited:
This effect is TOTALLY measured by the small & large level Zo of the complete amp.
It is naive to think of amp design as allowing 'EF3 or indeed anything else' to 'buffer the load from the VAS'. Feedback amps are systems and everything affects everything else.
The correct (ie most useful from a design point) view is the output stage provides GAIN (whether current or voltage). Various HF compensation methods just re-distribute this to better or worse effect.
____________
inventor, the author of the EDN article is very naive about how speakers work.
He graciously allows voltage drive for subs simply cos he dunno how to design good frequency response of subs with current drive. Don't think anyone else knows a good method either. But in fact, subs benefit a lot from current drive implemented properly.
____________
PS Its not back EMF that is evil (OK it is still sorta evil) but mostly the evil voltage caused by the EMF across the evil Z of the evil speaker that is alleviated (made less effective) by current drive.
PPS Evil = distorted
Last edited:
The classical T/S model is electrical on the left or input, mechanical in the middle and acoustic on the right or output. To get the load on the amplifier, you have to reflect the acoustic and mechanical back to the electrical side.
There was a Speaker Builder article that had SPICE models for various speaker-in-a-box types.
This is valid for small signals. The speaker's suspension changes with drive level and various other speaker parameters change with drive level.
In addition, there is a mechanism in the suspension that absorbs energy and changes with how much energy is absorbed. If you hit a speaker with a large signal the Fs changes some immediately and slowly changes over the course of a large fraction of a minute. If you remove the signal, the Fs recovers to the at-rest value.
I am not entirely sure how to put that large signal behavior into a SPICE model.
transitivity works
is the core of engineering, science and technology – we can test with “artificial” signals, loads – and with care, knowledge of the system behavior often predict the effects when connected to “real” loads
we do know a lot about “interfacing” electrical amps to loads – no need to fantasize about “unknown interactions” at audio frequencies - the theory, sims are very good, well tested against practice after 80+ years of negative feedback amplifier development, deployment
It is possible but pointless to make “more detailed” loudspeaker models - we can test the amp with much more extreme signals injected into the output with a Vsource on the other end of the load or a parallel Isource
if you try to model the speaker nonlinearity you mostly get harmonics - you can do a much more severe - and more easily interpreted load injection test with relatively prime sine, multitones, looking for IMD sum and difference between forward amplified signal and the "load injected" signal
large relatively prime input vs output test tones can drag the output I,V over any region of the I/V plane – easy to overbound worst case loudspeaker impedance and nonlinearity I,V trace
this can be done in sim or in the real world as a "tug-of-war" with a another big amp on the other end of your test load and another signal generator
at audio frequency there are no surprises until you clip/saturate/starve your output stage and then feedback can windup, recovery can become complicated
but for ordinary "mostly linear" operation within tha amps capabilities - the "interface distortion"/IMD is very close to, often less for "load injected" output stage current swing vs the the distortion seen with a "forward" normally amplified signal that give the same output Q currents into the same nominal load Z
there is no reason to believe current output audio amplification would be any different in this "infterface distortion" sense - its just ouptut compliance V that is changed by the load variations with a current output amp instead of the output current in a Vout amp
is the core of engineering, science and technology – we can test with “artificial” signals, loads – and with care, knowledge of the system behavior often predict the effects when connected to “real” loads
we do know a lot about “interfacing” electrical amps to loads – no need to fantasize about “unknown interactions” at audio frequencies - the theory, sims are very good, well tested against practice after 80+ years of negative feedback amplifier development, deployment
It is possible but pointless to make “more detailed” loudspeaker models - we can test the amp with much more extreme signals injected into the output with a Vsource on the other end of the load or a parallel Isource
if you try to model the speaker nonlinearity you mostly get harmonics - you can do a much more severe - and more easily interpreted load injection test with relatively prime sine, multitones, looking for IMD sum and difference between forward amplified signal and the "load injected" signal
large relatively prime input vs output test tones can drag the output I,V over any region of the I/V plane – easy to overbound worst case loudspeaker impedance and nonlinearity I,V trace
this can be done in sim or in the real world as a "tug-of-war" with a another big amp on the other end of your test load and another signal generator
at audio frequency there are no surprises until you clip/saturate/starve your output stage and then feedback can windup, recovery can become complicated
but for ordinary "mostly linear" operation within tha amps capabilities - the "interface distortion"/IMD is very close to, often less for "load injected" output stage current swing vs the the distortion seen with a "forward" normally amplified signal that give the same output Q currents into the same nominal load Z
there is no reason to believe current output audio amplification would be any different in this "infterface distortion" sense - its just ouptut compliance V that is changed by the load variations with a current output amp instead of the output current in a Vout amp
Last edited:
ADMITTANCE of a Loudspeaker as a function of frequency
Hi everyone,
I have no experience in audio amplifier or loudspeaker measurements. Nevertheless, I am curious about the quantity I mentioned in the title above.
Has anybody (anywhere in the world up till now) directly measured the admittance of a loudspeaker at different frequencies (in the audio frequency range) and obtained a plot/graph showing how the loudspeaker's admittance varies with frequency?
Let me explain what I mean by the phrase 'directly'.
I presume that the impedance measuring instruments used for measuring loudspeaker impedance operate in the following way (Indirect method):
-------------------
Indirect method:
A sinusoidal voltage signal of a certain frequency and certain amplitude (produced by a constant-amplitude variable-frequency voltage-source) is applied to the loudspeaker. The resulting loudspeaker current is then measured at that frequency as an RMS value (since the resulting current may not be purely sinusoidal). This measurement is repeated at different frequencies. The magnitude of loudspeaker's impedance at each frequency is then calculated as the ratio of the applied voltage to the measured current (i.e. calculating Z= V / I).
Once the "Impedance versus Frequency" graph is obtained in this way, the "Admittance versus Frequency" graph can be easily constructed by inverting the loudspeaker's impedance value at each frequency (i.e. calculating Y = 1 / Z).
-------------------
When I earlier said 'directly', I did not mean this Indirect method that relies on calculating Y = 1 / Z. What I mean is the following Direct method.
-------------------
Direct method:
This is almost the same procedure as above but (instead of a constant-amplitude variable-frequency sinusoidal voltage source) it uses a constant-amplitude variable-frequency sinusoidal current source to drive the loudspeaker.
At each frequency, for a given magnitude of driving current I (RMS), the resulting loudspeaker voltage V is measured (RMS). The loudspeaker admittance at that frequency is then calculated directly by dividing the applied current by the measured voltage (since Y = I / V).
-------------------
Note: In both methods, the sources used (the Voltage-source in Indirect method and the Current-source in Direct method) are ideal sources (or as close to ideal as possible).
So, the question is: Has anybody obtained a loudspeaker "admittance versus frequency" graph by using the Direct method above.
Thanks for your attention.
P.S. I presume that the "Admittance vs. Frequency" graph resulting from the Direct method will differ from that obtained via the Indirect method.
Well, now that all the easy answeres have been given --- the THd is greatly lowered with just simple motional feedback.... at the low freqs/bass. And, it is the low freq/bass where the distortion is highest in most systems. [We all know motional feedback cant see the cone break-up etc etc which also affects distortion and response.] And, if the bass resonant freq shifts around, under dynamic conditions, it doesnt matter as it isnt freq dependant -- at least in my implimentation.
Isolation, as I call it, has been demonstrated by at least 2 people here at DYIAudio using SIM ... a generator applied to the output (no input signal) can be seen at the Vas input. In other sim's it shows increased distortion. The greater the OPS isolation from Vas the lower the distortion (again shown in SIM, recently). The 3EF does a better job of keeping the signal at the output from getting back to the Vas (again shown in SIM).
Depending on the circuit topology, there is diffinitly lowered distortion via OPS isolation (my choice of words). And, if the Zo is not zero, this can allow back emf from causing distortion. The very low Zo at bass freq of high GNFB circuits does a better job than higher Zo circuits. However, low Zo from the amp does little to reduce speaker distortion at fundamental resonance.... but motional-feedback does reduce driver distortion... acoustic distortion from the speaker/driver.
So, again, we can expect in Some Amps this speaker back emf could create distortion in some amp designs.
Who has an electrical model of a real speaker that is not aimed at T-S tests/parameters but also includes back emf generation? (yes, the amount would be variable and to be useful, a range of back emf would be determined and used or measured for one specific speaker/system).
THx-RNMarsh
Isolation, as I call it, has been demonstrated by at least 2 people here at DYIAudio using SIM ... a generator applied to the output (no input signal) can be seen at the Vas input. In other sim's it shows increased distortion. The greater the OPS isolation from Vas the lower the distortion (again shown in SIM, recently). The 3EF does a better job of keeping the signal at the output from getting back to the Vas (again shown in SIM).
Depending on the circuit topology, there is diffinitly lowered distortion via OPS isolation (my choice of words). And, if the Zo is not zero, this can allow back emf from causing distortion. The very low Zo at bass freq of high GNFB circuits does a better job than higher Zo circuits. However, low Zo from the amp does little to reduce speaker distortion at fundamental resonance.... but motional-feedback does reduce driver distortion... acoustic distortion from the speaker/driver.
So, again, we can expect in Some Amps this speaker back emf could create distortion in some amp designs.
Who has an electrical model of a real speaker that is not aimed at T-S tests/parameters but also includes back emf generation? (yes, the amount would be variable and to be useful, a range of back emf would be determined and used or measured for one specific speaker/system).
THx-RNMarsh
Last edited:
I don't know about that, but the 3 systems i have used (1 manual, 2 computer based) used current source amps (either directly or with a BIG resisitor in series with a voltage source)
dave
The Woofer Tester 2 is a pure/ideal 3mA current source. It is the latest and greatest incarnation of our original Woofer Tester from the mid-1990s.
Infinity is something like 330 ohms.
I am working on a 4A/250W implementation of this for large signal characterization using the Woofer Tester Pro. The amp is going to be somewhat spendy.
You can see this at woofertester.com
Yes, Norton source for impedance testing is universal. It's easier to measure voltage than current. You'd get the same answer either way if your amp and wiring had near-zero output impedance and you could measure current accurately without appreciably altering the source impedance.
Admittance and impedance are by definition reciprocals of each other, so there's no such thing as a measurement of one without the other.
Admittance and impedance are by definition reciprocals of each other, so there's no such thing as a measurement of one without the other.
I would think that this would be simple to add to a SPICE model used for T/S parameter extraction. The voltage e (at the input of the BL gyrator) is velocity times BL. IIRC, that velocity is the current in the mechanical loop.
You could measure this relationship if you had a mechanical pusher for the speaker cone similar to DUMAX but AC instead of a DC pusher that DUMAX used.
If you measure the speaker in the 'traditional' way, you have a frequency dependent ralationship between voltage and impedance.
It is then just a matter to present this info in a different format to show a curve of admittance with frequency. Both are based on the same parameters.
Edit - SY beat me to it. Again.
jan
That's the "guess" that we're all waiting to see tested . . . with off-the-shelf currently available drivers. Simple measurements showing actual benefit is what's missing in all this discussion . . .
Does "current drive" (single driver, no crossover) give "lower acoustic output distortion"?
Z = 1/G by definition. "Theory" is irrelevant. If you measure one, you measure the other.
No, it's a definition. You may get two different numbers with two different measurement methods, since there's an infinite number of incorrect ways to take a measurement. But if one or the other is correct, it gives both reciprocal quantities, whether linear, nonlinear, time dependent, gravity dependent, temperature dependent, or anything else. That's what we mean by a "definition."
BTW, in my earlier post, I should have used Y instead of G; that was sloppy terminology.
BTW, in my earlier post, I should have used Y instead of G; that was sloppy terminology.