Correct me if I'm wrong, but isn't it the current through the voice-coil that makes the sound? So, why are we always measuring voltages on the output of amps? To makes things worse, most of the time a purely resistive dummy load is used for these measurements.
To see what happens current wise I used a 0.1 ohm 'current sense' resistor to measure the current. Three channels were captured and used for analysis: input voltage, output voltage and output current. A dual FFT analyzer was used for the analysis. Sampling rate was 192KHz, FFT size 768000, pseudo-random noise excitation, rectangular window.
Two amps were used for the measurements, my single ended tube amp and a NAD C320 solid state amp. The load was a Fostex208S in a Replikon BLH/BVR enclosure.
The graph shows the impulse response (voltage and current) for both amps. I'm definitely operating at the limit of what is possible with this setup, we're looking at individual samples 5 microseconds apart, but I think the result is still interesting.
My interpretation (open for discussion):
The back EMF from the driver is more pronounced in case of the tube amp and also reflected more in the current response. The graph for the NAD shows feedback in action, the amp is working like mad to get the voltage down.
To see what happens current wise I used a 0.1 ohm 'current sense' resistor to measure the current. Three channels were captured and used for analysis: input voltage, output voltage and output current. A dual FFT analyzer was used for the analysis. Sampling rate was 192KHz, FFT size 768000, pseudo-random noise excitation, rectangular window.
Two amps were used for the measurements, my single ended tube amp and a NAD C320 solid state amp. The load was a Fostex208S in a Replikon BLH/BVR enclosure.
The graph shows the impulse response (voltage and current) for both amps. I'm definitely operating at the limit of what is possible with this setup, we're looking at individual samples 5 microseconds apart, but I think the result is still interesting.
My interpretation (open for discussion):
The back EMF from the driver is more pronounced in case of the tube amp and also reflected more in the current response. The graph for the NAD shows feedback in action, the amp is working like mad to get the voltage down.
Attachments
The next one is perhaps even more interesting: input output cross-correlation. I must admit that I'm still not clear on how to interpret the results, but one thing is certain: Voltage and current each show a different picture. Looking at the voltage cross-correlation I would say that the NAD looks much better. For the current the picture is less clear, I think the tube amp has the advantage here (look at how the tail decays).
Attachments
Gerrit Boers said:
Hi,
A perfect voltage amplifier has zero output impedance and the
voltage level drives the speaker by producing whatever current
is necessary. Not so for a less than perfect amplifier (valve).
Of course its the current that moves the drivers but any look at
an impedance curve shows that the current is all over the place
for voltage drive of typical two way speakers.
The behaviour of the amplifiers current delivery into various loads
is best explored by various lower resistive loads, the complex loads
of speakers, e.g. Z= 6 ohm / phase angle = 45 is ~ equivalent to 4R
loading.
Measuring the current for a driver can tell you things, see :
http://klippel.de/pubs/Klippel papers/Loudspeaker Nonlinearities–Causes,Parameters,Symptoms_06.pdf
🙂/sreten.
This (quite controversial) paper by Bill Perkins (www.pearl-hifi.com) might be interesting:
http://www.pearl-hifi.com/06_Lit_Archive/07_Misc_Downloads/Power_Distortion.pdf
Over in the "Distortion Preception" thread this topic was briefly hit also (that current is the only thing that matters, and voltage drive may be flawed by design. Nonetheless the voltage interface has become the standard and is what current drivers are designed for).
- Klaus (big fan of higher impedance drive, espcially for full-range drivers)
http://www.pearl-hifi.com/06_Lit_Archive/07_Misc_Downloads/Power_Distortion.pdf
Over in the "Distortion Preception" thread this topic was briefly hit also (that current is the only thing that matters, and voltage drive may be flawed by design. Nonetheless the voltage interface has become the standard and is what current drivers are designed for).
- Klaus (big fan of higher impedance drive, espcially for full-range drivers)
I'm familiar with the work by Klippel, it's one of the things that prompted me to start this investigation.
The difference in output impedance between the two amps can be clearly seen in the amplitude response for the output voltage. The current response shows the impedance of the loudspeaker, with the inclusion of some nasty peaks.
It is my intention to take a closer look at distortion products in the current at several frequencies to see how the amps cope with the peaks in the impedance plot of the loudspeaker.
The difference in output impedance between the two amps can be clearly seen in the amplitude response for the output voltage. The current response shows the impedance of the loudspeaker, with the inclusion of some nasty peaks.
It is my intention to take a closer look at distortion products in the current at several frequencies to see how the amps cope with the peaks in the impedance plot of the loudspeaker.
Attachments
KSTR said:
Largely a consequence of a big turn -- some say a wrong turn -- that was made with the fairly concurrent introduction of small inefficient acoustic suspension systems and the transisitor with its cheap watts.
Fortunately systems that tend to prefer current drive are on the way back and in many cases redefining bang-for-the-buck and state-of-the-art.
dave
Re: Re: Current sense measurements on amp/loudspeaker
Valve amplifiers are certinly capable of being very good voltage amps... but why bother. You'd have to include nelson Pass' F1 & F2 into your less than perfect amplifier (only less than perfect because you are making the assumption that a perfect amp has to be a voltage amplifier). A more reasonable approach is to look at the amplifier/cable/speaker as a system (which it is). The ad hoc amplifier driving the ad hoc speaker is were perfection breaks down.
dave
sreten said:
Valve amplifiers are certinly capable of being very good voltage amps... but why bother. You'd have to include nelson Pass' F1 & F2 into your less than perfect amplifier (only less than perfect because you are making the assumption that a perfect amp has to be a voltage amplifier). A more reasonable approach is to look at the amplifier/cable/speaker as a system (which it is). The ad hoc amplifier driving the ad hoc speaker is were perfection breaks down.
dave
Re: Re: Re: Current sense measurements on amp/loudspeaker
Interesting, I will have to study this.
I like to look at the amp-speaker-room as one system.
Attached is the phase response for current and voltage. Notice that there are just four points where voltage and current are in phase.
KSTR said:
Interesting, I will have to study this.
planet10 said:
I like to look at the amp-speaker-room as one system.
Attached is the phase response for current and voltage. Notice that there are just four points where voltage and current are in phase.
Attachments
KSTR said:
actually by Kurt Steffensen (in the foreground)
(Bill edited a post(s) Kurt made on the JoeList)
dave
Hi Gerrit,
How good it is to see this investigation at the amplifier-loudspeaker interface.
You surmised >> isn't it the current through the voice-coil that makes the sound? <<
The current through the voice coil energises the voice coil, but as you have shown, the current flowing at any moment can be quite different to the applied voltage waveform.
If you have two identical output stages and connect one to a resistor and one to a loudspeaker, then have a sensing channel connected between the two outputs, you could observe any effect upon amplifier drive caused by the reactive loading alone.
Voltage/current differences observed at the amplifier output relate to amplifier output impedance characteristics and whether any form of feedback or error correction are applied; also to the nature of the loudspeaker load connected.
Re Post#1 traces.
I would really need to see circuit diagrams for both amplifiers to help me attempt to interpret what the amplifiers are doing, especially whether any form of feedback/correction is applied, whether the phase angle of any NFB damping is coherent throughout audio frequencies, also the nature of passive output impedance.
The SiTuIT control is clearly more 'relaxed' when compared to the NAD and shows different settlement characteristics which, although both are supersonic and above what we can hear via the LS, still cannot fail to impact upon reproduction because the deviations are in series with necessary on-going electrical corrections arising within the amplifier circuit.
A voltage alternation about the zero voltage line has much more significance with a SS NFB amplifier if this has a class-AB output stage due to current conduction repeatedly crossing over between output halves as the error voltage is 'actively' minimised.
These responses would be quite different if the loudspeaker were multi-driver with crossover. Co-connected circuits generate separate but coincidental back-EMFs which simultaneously interact with respect to the amplifier's attempt to generate an undistorted copy of the input voltage waveform.
(Back later... I don't have time to study more posts at the moment.)
Cheers ........ Graham.
How good it is to see this investigation at the amplifier-loudspeaker interface.
You surmised >> isn't it the current through the voice-coil that makes the sound? <<
The current through the voice coil energises the voice coil, but as you have shown, the current flowing at any moment can be quite different to the applied voltage waveform.
If you have two identical output stages and connect one to a resistor and one to a loudspeaker, then have a sensing channel connected between the two outputs, you could observe any effect upon amplifier drive caused by the reactive loading alone.
Voltage/current differences observed at the amplifier output relate to amplifier output impedance characteristics and whether any form of feedback or error correction are applied; also to the nature of the loudspeaker load connected.
Re Post#1 traces.
I would really need to see circuit diagrams for both amplifiers to help me attempt to interpret what the amplifiers are doing, especially whether any form of feedback/correction is applied, whether the phase angle of any NFB damping is coherent throughout audio frequencies, also the nature of passive output impedance.
The SiTuIT control is clearly more 'relaxed' when compared to the NAD and shows different settlement characteristics which, although both are supersonic and above what we can hear via the LS, still cannot fail to impact upon reproduction because the deviations are in series with necessary on-going electrical corrections arising within the amplifier circuit.
A voltage alternation about the zero voltage line has much more significance with a SS NFB amplifier if this has a class-AB output stage due to current conduction repeatedly crossing over between output halves as the error voltage is 'actively' minimised.
These responses would be quite different if the loudspeaker were multi-driver with crossover. Co-connected circuits generate separate but coincidental back-EMFs which simultaneously interact with respect to the amplifier's attempt to generate an undistorted copy of the input voltage waveform.
(Back later... I don't have time to study more posts at the moment.)
Cheers ........ Graham.
Hi Gerrit
---Correct me if I'm wrong, but isn't it the current through the voice-coil that makes the sound---
Current i through the voice coil determines the force applied to the cone. (F = B.l.i) not the emitted sound pressure.
Drivers are voltage driven, which means that what determines the sound output is the voltage across the voice coil, according to the frequency response which has been established by an amplifier having an almost null output impedance and the standard voltage of 2.83 Vrms.
If, at a given frequency, you know the voltage across the voice coil impedance, watever is the preceding impedance or the inner impedance of the voice coil, you can know the sound output by looking the level at that frequency on the frequency response : just a matter of adding or substracting the dB value of the ratio between the voltage you measured and the 2.83 V reference.
---Correct me if I'm wrong, but isn't it the current through the voice-coil that makes the sound---
Current i through the voice coil determines the force applied to the cone. (F = B.l.i) not the emitted sound pressure.
Drivers are voltage driven, which means that what determines the sound output is the voltage across the voice coil, according to the frequency response which has been established by an amplifier having an almost null output impedance and the standard voltage of 2.83 Vrms.
If, at a given frequency, you know the voltage across the voice coil impedance, watever is the preceding impedance or the inner impedance of the voice coil, you can know the sound output by looking the level at that frequency on the frequency response : just a matter of adding or substracting the dB value of the ratio between the voltage you measured and the 2.83 V reference.
Interesting suggestion, I can do that. I just need some more cabling, my audio interface has plenty of channels and the analyzer software can also handle lot of channels simultaneously.
As far as I know the NAD is a class A/B amp with lots of feedback, my tube amp has no feedback. I attached the schematic, it is as simple a can be. The amplification is 0.5, hence the name: Single Tube Impedance Transformer.
Nice term, relaxed. Although the tube amp is a bit slower due to the bandwidth limitations (~50Kz) it looks like it needs fewer oscillations to get to zero. The NAD is faster but it oscillates more.
In 'Testing Loudspeakers' it says:
The efficiency calculation assumes that the driver impedance is resistive and equal to Re.
.... This approach was taken because, as we have seen, the actual driver impedance varies widely with frequency, and calculation of the input power at any one frequency is problematical.
So, because calculating the real thing is problematical we make an assumption which we know is a gross simplification. We do this because all modern solid state amplifiers are voltage amps and because it is simple, not because it is the right thing to do.
Attachments
Hi Gerrit.
Another intereting plot is to have the resistor current as input to an 'X' axis and the difference input (between resistor and loudspeaker outputs) to 'Y' axis as if feeding and oscilloscope screen. Will your software do this ?
Reactive loudspeaker induced changes will cause what would be a straight line diagonal trace to loop out along the 'Y' axis wrt 'X'.
However any crossover induced errors will be seen as additional spikes along the 'Y' axis these become shifted away from the zero resistor current axis as the loudspealer currents become phase shifted.
Thus when 'X'-'Y' comparing the resistor-loudspeaker diffence outputs from a class-AB amplifier two sets of spikes might be revealed; one from resistor loading close to the zero current voltage axis, and then another phase shifted and displaced in both amplitude and time by the loudspeaker loading.
Clearly a tube amp would not generate such spikes.
Re the tube amp trace in Post#1. There is some slight ringing. Maybe the output transformer could do with having a Zobel fitted ? Pure guess maybe 2200pF/1000V in series with 470 ohms.
If the heater is DC maybe one 220uF to each side of the heater, or if AC to a heater winding centre tap. Each electrolytic should also be parallel bypassed by say a quality 1uF as well because these components are in series with both input grid and output anode circuits.
Will read the other posts shortly.
Cheers ......... Graham.
Another intereting plot is to have the resistor current as input to an 'X' axis and the difference input (between resistor and loudspeaker outputs) to 'Y' axis as if feeding and oscilloscope screen. Will your software do this ?
Reactive loudspeaker induced changes will cause what would be a straight line diagonal trace to loop out along the 'Y' axis wrt 'X'.
However any crossover induced errors will be seen as additional spikes along the 'Y' axis these become shifted away from the zero resistor current axis as the loudspealer currents become phase shifted.
Thus when 'X'-'Y' comparing the resistor-loudspeaker diffence outputs from a class-AB amplifier two sets of spikes might be revealed; one from resistor loading close to the zero current voltage axis, and then another phase shifted and displaced in both amplitude and time by the loudspeaker loading.
Clearly a tube amp would not generate such spikes.
Re the tube amp trace in Post#1. There is some slight ringing. Maybe the output transformer could do with having a Zobel fitted ? Pure guess maybe 2200pF/1000V in series with 470 ohms.
If the heater is DC maybe one 220uF to each side of the heater, or if AC to a heater winding centre tap. Each electrolytic should also be parallel bypassed by say a quality 1uF as well because these components are in series with both input grid and output anode circuits.
Will read the other posts shortly.
Cheers ......... Graham.
Exactly my point of view in these matters. The fact that most modern drivers are designed to give flat response with a voltage source and may have problems with current drive does not mean that there cannot be another, probably better way, ranging from matched impedance drive (power drive) to true current drive.Gerrit Boers said:
Personally I also have the feeling (call it a hypothesis) that for any given small frequency range of a driver there will be an optimum drive impedance where the driver behaves best wrt to breakup/distortion, just because of the right amount of electrical damping from infinite to zero (which means Zout can also go a bit negative, up to the Re level). This would be worth a closer investigation and would lead to a specific amp for a given driver/speaker (yes, the true system approach) with varing Zout vs frequency and properly tailored frequency response. I do know that there is something going on as I have examinened with guitar speakers (which operate mostly in breakup regions) that their directivity changes a bit with varied drive impedance, which can only be due to an alteration of cone modal patterns arising from different damping of the VC.
- Klaus
Hi Gerrit,
Driving loudpeakers with high impedance may sometimes be an advantage and reduce distorsion (see Hawkford or Greiner & Sims's AES papers) however it leaves them with only the mechanical damping of the main resonance.
I am not aware of data about an optimal driving impedance to smooth cone break up.
Nominal sensitivity at a specified frequency (SPL expressed in dB for a 2.83 V input - no Watt please - and at 1 m) is a much more useful data than efficiency. The frequency response is nothing else than the sensitivity response for every frequency. There is no over-simplification at all using it.
Using an amplifier with a not insignificant output impedance (i.e. valves, transformers, passive filters) you can isolate the transfert function in the frequency domain of the amplifier-driver interface by loading it with the intended loudspeaker.
The overall output pressure at any frequency is obtained by the addition in dB of the interface transfert function and the initial frequency response of the driver when voltage driven.
Driving loudpeakers with high impedance may sometimes be an advantage and reduce distorsion (see Hawkford or Greiner & Sims's AES papers) however it leaves them with only the mechanical damping of the main resonance.
I am not aware of data about an optimal driving impedance to smooth cone break up.
Nominal sensitivity at a specified frequency (SPL expressed in dB for a 2.83 V input - no Watt please - and at 1 m) is a much more useful data than efficiency. The frequency response is nothing else than the sensitivity response for every frequency. There is no over-simplification at all using it.
Using an amplifier with a not insignificant output impedance (i.e. valves, transformers, passive filters) you can isolate the transfert function in the frequency domain of the amplifier-driver interface by loading it with the intended loudspeaker.
The overall output pressure at any frequency is obtained by the addition in dB of the interface transfert function and the initial frequency response of the driver when voltage driven.
But we can clearly see some (not all) of the breakup modes in the impedance plot. Which IMO opens the way to have some influence on it electrically. As I said, this will need further experiments to see how much influence is possible.Pan said:
- Klaus
Graham Maynard said:
Hi Graham,
The software will not do this directly, but maybe I can route the signal through my digital audio workstation and use an M/S matrix to get the sum and difference for two channels. I would need to make sure that the DAW does not mess things up.
As for the ringing, I did not put any capacitor across the input transformer because I could not see any effect while measuring with the Clio system (48 KHz SR). The new measurements indicate that I have to look into this again.
The heating supply is DC, CRC filtered followed by a voltage controlled current source (VCCS). The capacitors are Black Gate NH non-polarized types in a 'Super E cap configuration', definitely not standard electrolytic caps.
This is why I think you should use SET amps with low Q drivers.
With respect to voltage and current:
Suppose we have a super-conducting voice-coil. What would happen?
Re will approach zero ohm, does this mean that the Qes also approaches zero? Would this also imply that the efficiency approaches infinity?
I'm not very good at mathematical limits, maybe somebody can help me out here.
I will take lose look at what happens to the distortion (both voltage and current) at, and around, the impedance peaks. I will also need to do some detailed acoustic measurements.
The peak at about 70Hz has at least one positive effect, the dissipation in the voice-coil is severely reduced at this point.
Hi Gerrit
---Suppose we have a super-conducting voice-coil. What would happen?
Re will approach zero ohm, does this mean that the Qes also approaches zero? Would this also imply that the efficiency approaches infinity?---
A circuit simulating a voice coil with almost no resistive component exists : it uses an amplifier with a negative output resistance.
The Qes is nearly zero and the loudpseaker then acts like a differentiator, with avery high damping.
There is a patent and an AES paper by Erik Stahl which shows that the mechanical factors of a driver can be very well controlled by electrical means.
---Suppose we have a super-conducting voice-coil. What would happen?
Re will approach zero ohm, does this mean that the Qes also approaches zero? Would this also imply that the efficiency approaches infinity?---
A circuit simulating a voice coil with almost no resistive component exists : it uses an amplifier with a negative output resistance.
The Qes is nearly zero and the loudpseaker then acts like a differentiator, with avery high damping.
There is a patent and an AES paper by Erik Stahl which shows that the mechanical factors of a driver can be very well controlled by electrical means.
- Status
- Not open for further replies.