Sensing load impedance

Thread Starter

coinmaster

Joined Dec 24, 2015
502
Hello, with my limited knowledge I'm trying to devise a way to sense the load impedance on the output of an audio amplifier without affecting the output signal.

For example, if a pair of headphones changes impedance based on frequency I want to create a servo to adjust the output accordingly.
The main problem I have right now is figuring out how to measure the impedance, maybe using some sort of current sensing or something but I wouldn't know how to go about that and not mess with the signal. Especially since current sensing methods that I see seem to rely on having dc voltage on the input. The resistor (headphones) will only have ac voltage across it.

Maybe I can measure the rms of the ac voltages and use that to determine the load impedance? The ac rms should change with changing impedance. The issue then is how do I build a servo that functions off of rms rather than direct voltage measurements.
Any ideas?



Mods edit:
Please don't do double post.
Another post, please refer to : Amplifying DC milivolts to volts.
 
Last edited:

AnalogKid

Joined Aug 1, 2013
10,987
At audio frequencies you can sense current with a very small series resistor or with a hall effect sensor. The hall sensor does not place a measurable impedance in series with the output, but there is some phase delay between the incident signal and the sensor output. A series resistor has no such phase delay, but it is a small resistance. You might be able to place the resistor inside the amp's overall feedback loop, reducing its effect on the system output impedance.

Separate from that, do you want to track the instantaneous load impedance, or it's RMS value? Or, more correctly, it's impedance based on the instantaneous or RMS value of the current through it? Is this for a test setup or something to be operating while playing music?

Finally, why?

ak
 

Thread Starter

coinmaster

Joined Dec 24, 2015
502
Separate from that, do you want to track the instantaneous load impedance, or it's RMS value? Or, more correctly, it's impedance based on the instantaneous or RMS value of the current through it? Is this for a test setup or something to be operating while playing music?

Finally, why?
I'm building an output stage for an amplifier and I'm tuning it so it works best with the impedance of my headphones (HD800s) which are listed as 300ohm impedance headphones.
The problem is, the impedance is not 300ohms, it changes with frequency, so I'm trying to use a servo that will feed a jfet to be used as a variable resistor for real-time adjustments so the amp is always optimized for whatever the changing impedance might be. In other words I want real-time instant tracking of load impedance so I can make instant adjustments to the output stage operation.
 

AnalogKid

Joined Aug 1, 2013
10,987
I understand what you are trying to do, but I think it will undo the design of the headphones. Headphones are designed to be driven by a constant voltage (zero-ohm) output stage, followed by some series resistance. For example, a home stereo might have the amp outputs connected directly to the speaker terminals, but connected to the headphone jack through 47 ohm resistors. Some headphone manufacturers specify the optimal source impedance, but most just assume it will be very low. This is the electrical environment the phones are electromechanically designed to work with. If you drive them with a variable-frequency response amplifier, my first opinion is that the audio experience will not improve. Whatever their frequency response, the phones assume an acoustically flat source.

ak
 

Thread Starter

coinmaster

Joined Dec 24, 2015
502
My output stage is a white cathode follower, the value of the plate resistor is determined by what the load impedance is.
A white cathode follower is a push pull set-up so both tubes need to have matching operating points for the best performance.
In order to keep them balanced the load impedance should not alter from whatever value you used to set the plate resistor.
So in other words what I'm trying to do is allow the WCF to achieve balanced operation with changing load impedances because if the load is not 300ohms the push-pull tubes operate differently causing distortion when the signal rejoins.
 

Thread Starter

coinmaster

Joined Dec 24, 2015
502
Would this be fine? Seems to show varying voltage in the sim based on a difference in load impedance.

current measure-1.gif
 
Last edited:

Thread Starter

coinmaster

Joined Dec 24, 2015
502
I would then need to create a servo that outputs high positive voltages based on the input from the current sense ic.

I have no idea how to create a servo that outputs positive voltage, my attempts at it have failed I simply do not know enough. Any suggestions?
 

Thread Starter

coinmaster

Joined Dec 24, 2015
502
Actually, the voltage output from the current sense ic is a sine wave so I would have to use the rms voltage to adjust the jfet. I would have to use the rms of the AC signal to adjust the jfet between 75v-78vish.
Unfortunately the rms difference between 300 ohms impedance and 500 ohms impedance is a small fraction of a volt on the output of the ic.
I'm stumped on how to do this.
 
Last edited:

Thread Starter

coinmaster

Joined Dec 24, 2015
502
So I guess there is an IC that converts rms to DC, the LT1966, shown here using a current sense transformer to measure AC current https://www.uam.es/personal_pas/pat...nseCircuitsColectionAN105LinearTechnology.pdf

Is there some downside to putting a current sense transformer on the output of an audio amplifier? The windings would be inductive right?
Actually, what's the point of using the transformer to begin with, why not measure the AC directly? Maybe it assumes there is DC on the line?
 
Last edited:

Thread Starter

coinmaster

Joined Dec 24, 2015
502
Here's what I made, I'm going to assume the current sense transformer was just to separate dc or something.
LTC1966_TA09 (1)-1.gif

So the change from 300 ohms to 500 ohms impedance is about 2 mv.
So I need to change a difference of 2mv into a difference of 75-78ish volts.

I think I may start a more specific thread for that since I seem to be talking to myself here :p
 
Last edited:

sailorjoe

Joined Jun 4, 2013
364
Coinmaster, it looks like all threads lead to here.
Do you want to change the voltage across the FET from 75 to 78 Volts. I just want to be clear. And is it also true that you want to do that because of the changing impedance of your headphones at different frequencies. So in a complex audio signal, there will be many frequencies simultaneously, so how does the compensation circuit help balance the impedance under those conditions? Also, thinking about the whole audio band, is the changing impedance of the headphones linear across the band, or flat in the low to mid range, and then tapering in the high range? I ask because a compensation circuit would need to take that into account.
 

Thread Starter

coinmaster

Joined Dec 24, 2015
502
You bring up good points, there's also the fact that I just realized a music track has dynamic volume levels, so even if I could use an IC to sense current or RMS it would not matter if the voltage amplitude does not stay the same, unless I'm missing something.

I have another project I've been saving for sometime down the line that might help here in some way.
http://www.tubecad.com/2007/11/blog0126.htm
First article on the page, a buffer like that would allow the amp to know what's coming.

Do you want to change the voltage across the FET from 75 to 78 Volts
That is not an exact value but it is close, it would have to be adjustable anyway in order to compensate for volume attenuation on the music track and the amplifier assuming I'm still going to be using a current or rms sensor.

And is it also true that you want to do that because of the changing impedance of your headphones at different frequencies.
Yeah, my output stage is a white cathode follower http://www.tubecad.com/october99/page4.html, which is a push pull topology. The top and bottom triodes split and rejoin the signal canceling out distortion and allowing increased current capabilities and lower impedance.
The problem is, the top and bottom triodes must remain at the same operating points in order to sound its best.
If they are at different operating points then the signal does not rejoin properly and it introduces distortion.

The plate resistor value of the input triode is determined by the intended load impedance value, which will then allow the triode to be balanced with the bottom triode in my case a 440 resistor would be the value for a 300ohm load impedance.

What I am attempting to do is balance the tubes even though the headphones are changing impedance values via an adjustable plate resistor i.e the jfet.

So in a complex audio signal, there will be many frequencies simultaneously, so how does the compensation circuit help balance the impedance under those conditions?
I honestly have no idea, a deviation from 300ohm impedance on different frequencies must alter the input triodes current/voltage to some degree but I have no idea what the correlation is between a change in impedance on multiple simultaneous frequencies and the change in tube operating point, maybe the answer is obvious and I'm too ignorant to figure it out.

Also, thinking about the whole audio band, is the changing impedance of the headphones linear across the band, or flat in the low to mid range, and then tapering in the high range? I ask because a compensation circuit would need to take that into account.
This is what I have on that http://www.innerfidelity.com/images/SennheiserHD800.pdf
 
Last edited:

sailorjoe

Joined Jun 4, 2013
364
Well, you have a complicated situation with those headphones, based on the graphs. In the low to mid range, the frequency response is pretty flat, but the impedance changes from 350 to 650 ohms, and the phase is fairly linear. I think that means that adjusting your amplifier to compensate for impedance will severely unflatten the frequency response. An engineer might take the overall average of the impedance and set the triodes for that value. Another option would be a series resistor of 5 to 10 times the average headphone impedance in the output. Of course that would lose about 5 to 10 times your amplification, but maybe you already have more than enough. The idea is that the changes in impedance are sort of swamped by the large series resistor.

Back to the headphones, there is clearly a drop off in frequency response in the high range. Do you want to try to fix that, or let it ride?

Last question today. How will you measure the headphone response over the frequency range?
 

Thread Starter

coinmaster

Joined Dec 24, 2015
502
Well, you have a complicated situation with those headphones, based on the graphs. In the low to mid range, the frequency response is pretty flat, but the impedance changes from 350 to 650 ohms, and the phase is fairly linear. I think that means that adjusting your amplifier to compensate for impedance will severely unflatten the frequency response
Hmmm, By changing the value of the plate resistor I wouldn't be altering the impedance or phase of the signal on the headphones, just the operating point of the tube.
For example at 100hz the headphones peak at around 650 ohms, lets say the headphones were always a flat 650 ohms across the spectrum, I would then have to change the resistor value so both tubes operate at the same current/voltage point. So by adjusting the plate resistor value all I'm really doing is syncing the tubes operating points rather than affecting phase and impedance, if the tubes operating points are mismatched that means the the split signal is rejoining at different amplitudes which will alter the frequency response. So in other words instead of working against the graph I would be working toward it, unless I'm missing something?

Another option would be a series resistor of 5 to 10 times the average headphone impedance in the output. Of course that would lose about 5 to 10 times your amplification, but maybe you already have more than enough. The idea is that the changes in impedance are sort of swamped by the large series resistor.
The series resistor sounds like an interesting prospect. I do indeed have plenty of gain to lose, I'm not sure about x10 gain but quite a lot in anycase. Worst case scenario I can add more gain to the amplifier.

Back to the headphones, there is clearly a drop off in frequency response in the high range. Do you want to try to fix that, or let it ride?
Flatter frequency responses are always welcome so the answer is yes but I don't think it would be an immediate priority unless you know a way to fix it without deviating too much from the current goal. I'm interested to hear your idea though if you have one.

Last question today. How will you measure the headphone response over the frequency range?
I'm sure there's a fancy piece of equipment I would have to buy. I'll have to look into that.
Is there any reason I would have to measure headphone response? I would think that all I would have to do is measure the voltage and current of the WCF and see how their operating points change over the frequency range with the HD800s plugged in and then adjust them until the operating points stay the same over the entire range.
 
Last edited:

ScottWang

Joined Aug 23, 2012
7,397
Please reduce the resolution as 800x600 before you upload, and converting the file format to gif, png file format is only for vertical and horizontal line can be reducing the file size, the files you attached, I converted to gif from 1Mb to less than 100Kb.
 

sailorjoe

Joined Jun 4, 2013
364
Hmmm, By changing the value of the plate resistor I wouldn't be altering the impedance or phase of the signal on the headphones, just the operating point of the tube.
For example at 100hz the headphones peak at around 650 ohms, lets say the headphones were always a flat 650 ohms across the spectrum, I would then have to change the resistor value so both tubes operate at the same current/voltage point. So by adjusting the plate resistor value all I'm really doing is syncing the tubes operating points rather than affecting phase and impedance, if the tubes operating points are mismatched that means the the split signal is rejoining at different amplitudes which will alter the frequency response. So in other words instead of working against the graph I would be working toward it, unless I'm missing something?
If we assume that the test signal that went into the headphones had a fixed impedance, and the frequency response is flat, then modifying the impedance should affect the frequency response.
I get that you want to optimize the operating point of the tubes, and that would make sense if your signal wasn't all over the whole band essentially all the time. Probably the best you could is analyze the instantaneous frequencies of any signal, and adjust the impedance to some nominal value based on the largest few frequency components at any given time. That's a very complex problem to solve.
The series resistor sounds like an interesting prospect. I do indeed have plenty of gain to lose, I'm not sure about x10 gain but quite a lot in anycase. Worst case scenario I can add more gain to the amplifier.
At this point, I think the least disruptive option is the series resistor. It's value depends on the ideal, or perhaps the maximum, impedance of the amplifier. If all you can get is 3x the headphone average, then so be it. Since it's likely not possible to chase the headphone impedance, the best is to minimize its variation from the view of the tubes.
Flatter frequency responses are always welcome so the answer is yes but I don't think it would be an immediate priority unless you know a way to fix it without deviating too much from the current goal. I'm interested to hear your idea though if you have one.
The typical solution is to adjust the preamplifier to compensate for the headphone losses. Barring that, the signal can be conditioned at the input to the amplifier. Most of those circuits are active transistor or op amps, but passive circuits are also available. Depends on whether you prefer capacitors or inductors. If neither, then I don't know of any solutions.

A thought just occurred to me. If you feed your headphones from the amplifier with a single frequency, can you see the distortion on the tube output as the headphone impedance varies along the band?
 
Last edited:

Thread Starter

coinmaster

Joined Dec 24, 2015
502
If we assume that the test signal that went into the headphones had a fixed impedance, and the frequency response is flat, then modifying the impedance should affect the frequency response.
Which is exactly why it won't affect the frequency response because the impedance isn't fixed therefore the tubes don't maintain synchronized operation with a fixed plate resistor.
I get that you want to optimize the operating point of the tubes, and that would make sense if your signal wasn't all over the whole band essentially all the time. Probably the best you could is analyze the instantaneous frequencies of any signal, and adjust the impedance to some nominal value based on the largest few frequency components at any given time. That's a very complex problem to solve.
Yeah that's the issue it seems, the tube operating point must shift based on an average of impedances or something.

The typical solution is to adjust the preamplifier to compensate for the headphone losses. Barring that, the signal can be conditioned at the input to the amplifier. Most of those circuits are active transistor or op amps, but passive circuits are also available. Depends on whether you prefer capacitors or inductors. If neither, then I don't know of any solutions.
I'm not quite sure what you mean by this, how exactly would I condition it? By preamp I think you mean the input stage or dac?

At this point, I think the least disruptive option is the series resistor. It's value depends on the ideal, or perhaps the maximum, impedance of the amplifier. If all you can get is 3x the headphone average, then so be it. Since it's likely not possible to chase the headphone impedance, the best is to minimize its variation from the view of the tubes.
The quoted output impedance of the amp is 600 ohms, that doesn't leave much headroom, if I put a resistor on the output wouldn't it "see" the sum of the impedances no matter how much resistance you throw at it?

In either case I think the solid state approach is worth looking into because, while more complex, it solves a common issue and tube tech has received little innovation since they became obsolete in the 60s even though they sound so good.

A thought just occurred to me. If you feed your headphones from the amplifier with a single frequency, can you see the distortion on the tube output as the headphone impedance varies along the band?
That's what I was counting on.
 
Last edited:

Thread Starter

coinmaster

Joined Dec 24, 2015
502
Maybe I can use some sort of mechanism that will sample the frequencies and their amplitude and at any one moment and then take the average of the frequencies and their amplitude to create a reference voltage for the jfet. That seems like a good solution does it not?
The tube can only have 1 set of voltage and current characteristics at a time so obviously if the load has variable impedance levels on different simultanious frequencies then the voltage and current of the tube must alter via an average of those impedances.
 

blocco a spirale

Joined Jun 18, 2008
1,546
Maybe I can use some sort of mechanism that will sample the frequencies and their amplitude and at any one moment and then take the average of the frequencies and their amplitude to create a reference voltage for the jfet. That seems like a good solution does it not?
The tube can only have 1 set of voltage and current characteristics at a time so obviously if the load has variable impedance levels on different simultanious frequencies then the voltage and current of the tube must alter via an average of those impedances.
Have you ever heard the saying; "the cure is worse than the disease"?
 

Thread Starter

coinmaster

Joined Dec 24, 2015
502
How is my cure worse than my disease? Balancing triode operation seems like a nice cure to me.
Unless you're implying that it's "too much work" in which care I don't care :p
 
Top