measuring a source's impedance over frequencies?

[castleofargh]

ok the story begins badly, I don't have a multimeter. ^_^
but anyway I would like to have more than 1 value, and instead get just like for IEMs, an impedance over frequencies. for IEMs I go at it with room EQ wizard and something like this

 
 
I suspect it could also work to measure a source, but sending a signal into the output of a device, I wonder how fast I'm going to blow something up from one of the stupid mistakes I'm very sure to end up making?
am I worrying to much?
 
I got a roundabout way to get one single value that gives me fine enough results,  but if there is a safe method(safe even from my mishandling it) to get results over a range of frequencies without having to do it one frequency at a time , I certainly would like to have it.

 

[SilverEars]

Your diagram looks strange.  Why is the probe parallel to the short?  You should be probing across the R1.  Also, the L or R signal should be to ground.  
 
You get the open voltage or without the load first, and then you connect the resistor and measure across it.  And then, you have an unknown which is the internal resistance.  
 
If the source's impedance is purely real or resistive, it should be constant.

 

[castleofargh]

?
I did the first configuration, and I put one channel of the IEM I want to measure, where the probe is on the diagram. like left channel to the resistor, and the ground ring of the IEM to the ground ring for the ADC. I just added to this circuit a "lazy" switch that bypasses the resistor for the calibration procedure in REW, then switch back to having the resistor in the circuit, plug the IEM, and REW is very happy to give me the impedance for the range of frequencies I have set to be measured.
 
http://www.roomeqwizard.com/help/help_en-GB/html/impedancemeasurement.html#top   here is the diagram they suggest, looks the same to me.
 
that stuff works fine for IEMs, my concern is if I wanted to measure a source over a range of frequencies instead of a transducer or a resistor? I've seen it done, but I don't know how, and most of all I don't know if it could be dangerous for some devices.

 

[SilverEars]

Ah, ok.  I looked at the diagram, and looks like the probe is the sound card left and right channels.  So basically, the L and R channel difference is the voltage across the Rsense, and you get the current.
 
If the soundcard is the probe, than you'd ground(the source's) the right soundcard input. Other end of Rsense(where the speaker signal goes) should be grounded.  And you make Voltage across Rsense measurements across the frequencies.  
 
Given that sound card L and R is the probe, you connect it across the R source output, you should get the open voltage(Vopen).
 
(Vopen - (Vleft-Vright))/Rsource=I=(Vleft-Vright)/Rsense
 
Rsource = ((Vopen - (Vleft-Vright))*Rsense)/(Vleft-Vright) 
 
You measure Vopen using the probe.  I think this would work.

 

[dprimary]

This might help you
 
http://www.artalabs.hr/AppNotes/AP1_MeasuringBox-Rev4Eng.pdf

 

[castleofargh]

this is also clearly aimed at transducers, those guys receiving a signal is what they're built for. can I really afford to use such a thing to send a signal into a DAP's output? I have some serious doubts about that being the proper/safe way(so much that I have never tried).  it's getting a source's impedance from 20hz to 20khz that I'm looking for. not to measure my IEMs/headphones, I'm very fine with what I already have for that purpose.
 
I thought about how to get it, pretty much just like getting 1 value with a test tone at Xhz and measure the output into 2 different loads. to do that over a wide range, I thought about playing an impulse response the same way into 2 different loads, but then what am I supposed to do with the result? is there a software(free?) that will give me the impedance at each frequency simply by looking at those recorded impulse? and I suspect I would also need a calibration to remove my sound input's potential variations. so it's not really something I feel I can pull out of a hat and create into excel.
 
I can't find where I've seen cellphones or stuff like with a pretty impedance graph from 20hz to 20khz, but it's been done, of that I'm sure. I just don't know how.

 

[MindsMirror]

I know Matlab can do everything that you want if you have the patience or desire to figure out how to do it. You could write a script to generate a frequency sweep, analyze the response, and create the pretty impedance graph.

 

[castleofargh]

  I know Matlab can do everything that you want if you have the patience or desire to figure out how to do it. You could write a script to generate a frequency sweep, analyze the response, and create the pretty impedance graph.

you're probably right, but I have zero experience with matlab. also it's not free, as the ultimate purpose would be to share the results, looking to crack it wouldn't be cool.
 
of course I'm asking this now, when I have already too many stuff I'm learning how to do. but I suspect I might have a chance importing the recorded impulse response in room eq wizard, maybe that thing can give me a frequency response from the impulse?(I have zero knowledge or practice about IR and still limited knowledge of REW).
if that works, then I could just export the FR with from the 2 loads and use that in excel or whatever that can give me a simple graph from numbers. I might try looking into that(funny how the ideas come from asking).
TBH I was curious to know how others did it, but it doesn't seem to be the activity of the month on headfi, or the kind of stuff people do in family to pass time on a week end.

 

 

[jcx]

I like SciLab for a free Matlab work-alike, but Octave is supposed to be more directly compatible and now is rumored to have a acceptable GUI - but many are now saying that Python's ecosystem is fine, more flexible than the dedicated analysis packages

personally I have used Audacity for play/capture and LTspice and SciLab for test waveform generation and analysis respectively

the various speaker analysis software should also do the right thing for frequency/phase analysis

 

[dprimary]

Oh output impedance I read it as you want to measure transducers.  This might help
 
http://www.sengpielaudio.com/calculator-InputOutputImpedance.htm
 
But if you don't have meter, I'm guessing you don't have scope either.
 
If you are looking for an RMS multimeter check the frequency bandwidth it can measure ones that go over 1k are still fairly expensive.

 

[SilverEars]

  this is also clearly aimed at transducers, those guys receiving a signal is what they're built for. can I really afford to use such a thing to send a signal into a DAP's output? I have some serious doubts about that being the proper/safe way(so much that I have never tried).  it's getting a source's impedance from 20hz to 20khz that I'm looking for. not to measure my IEMs/headphones, I'm very fine with what I already have for that purpose.
 
I thought about how to get it, pretty much just like getting 1 value with a test tone at Xhz and measure the output into 2 different loads. to do that over a wide range, I thought about playing an impulse response the same way into 2 different loads, but then what am I supposed to do with the result? is there a software(free?) that will give me the impedance at each frequency simply by looking at those recorded impulse? and I suspect I would also need a calibration to remove my sound input's potential variations. so it's not really something I feel I can pull out of a hat and create into excel.
 
I can't find where I've seen cellphones or stuff like with a pretty impedance graph from 20hz to 20khz, but it's been done, of that I'm sure. I just don't know how.

For DAPs, getting the tone input is the problem.  I guess you play a tone sweep, and have a program sample the sweep across the Rsense.  Do some excel work, and you get the impedance.  As far as how to know the specific frequency per sample is questionable.  If there is information about timing of the tone sweep playback file is available, that should help.   

 

[yuriv]

I recorded an old iPod Video from 2005 playing impulses, with and without a 16-ohm load on one channel. Then I loaded the .wav files into REW, exported the frequency response to .csv files, and then uploaded them to Google Sheets for the math and the charts.
 
I went from this: https://dl.dropboxusercontent.com/u/6735266/iPod%20Video%20-%201%20Hz%20Impulse%20response%20train%2C%20open%20circuit%20and%20with%2016-ohm%20load.zip
 
 
...to this in REW:
 

 

 
 
 
... to this in Google Sheets:
 

 
 
I'm not sure I trust these quick and dirty measurements a whole lot. The magnitude vs. frequency plot looks about right though. You could probably get an estimate for the size of the iPod Video's blocking capacitor that's not too far off the mark. The tools are free, and it didn't take that much time. If there's a package out there that can do this automatically, it'd be good to know about it.

 

[castleofargh]

  Oh output impedance I read it as you want to measure transducers.  This might help
 
http://www.sengpielaudio.com/calculator-InputOutputImpedance.htm
 
But if you don't have meter, I'm guessing you don't have scope either.
 
If you are looking for an RMS multimeter check the frequency bandwidth it can measure ones that go over 1k are still fairly expensive.

but that puts me back to measuring 1 frequency at a time. I have a weirdo method for that already that's not accurate, but gets me in the ballpark of the proper value. I was looking for some way to get the audio range in one go. I guess I didn't do the best job at explaining myself ^_^.

 

[yuriv]

 
Here's what I got when I let Room EQ Wizard (REW) measure the output impdedance using a MacBook Pro's headphone out and line in:
 

iPod Video output impedance using REW, 100-ohm R_sense, and MacBook Pro stock audio.

Compare that with the one I got earlier using the "safe" method of measuring and analyzing impulses:

 
It's pretty close. For the magnitude, the one that was computed with recorded impulses got numbers that are closer to the results from the voltages measured with my scope's built-in meter. The phase response, though, gets weird at the frequency extremes, especially at the high end. Maybe we can get better results by recording at 96 k.

100 ohms for R_sense might be too high for other things you want to measure with REW, especially if the headphone output that's supplying the test signal has limited voltage. When I tried R_sense = 100 ohms with the headphone outputs of a Startech ICUSBAUDIO2D USB audio interface, REW says that the input level is too low. Make R_sense high enough compared to the output impedance of the device under test (DUT),  and the voltage at the DUT's terminals will be too low for REW to measure reliably.

So I tried it with R_sense = 33 ohms:

 
When done by hand with a multimeter, here's what I got for 1 kHz:
140 mV rms (open-circuit voltage, -12 dBFS)
122 mV rms (16.8-ohm load measured, -12 dBFS)

http://www.wolframalpha.com/input/?i=140*16.8%2F%28R%2B16.8%29+%3D+122

REW would probably have a closer result to this if I had just left the test lead resistance at 0.000 ohms. For sources that have even lower output impedance, you would have to use lower and lower values for R_sense, and as a result, the device that's playing the test signal would have to supply more and more current. That's not going to be practical for sources with near zero-ohm output impedance. Besides, if you were worried about destroying the DUT in the first place, you wouldn't want to send lots of current through it. You want to keep R_sense high.

What you might be able to do is to put a resistor, let's call it R₁, between R_sense and the DUT. Have REW measure R₁ and the DUT in series. Then have REW measure R₁ alone. Now subtract the second result from the first using REW's trace arithmetic. Here it is in action with 100- and 50-ohm resistors for R_sense and R₁, and a FiiO E6:

 
The green line near zero is the difference. Here it is in more detail:

 
Of course this won't be terribly accurate because the voltage going to the E6 is very small. Noise and crosstalk could start affecting the accuracy and precision of the measurement. The real magnitude response would probably be a flat line. The measured phase at high frequencies is probably bogus. But it should give you an idea of how low the impedance could be.

You might just decide that you don't want to go through all of this trouble. Portables nowadays don't use coupling capacitors for their headphone amps as much as they did 10 years ago. A measurement at 1 kHz may very well give you everything you need to know. Or, if you really doubt it, you could take measurements at 60 Hz and 10 kHz as well. 

 

[castleofargh]

thanks a lot yuriv, you've done all the work for me. \o/
I've tried only a few impulse responses on a few sources right now, and yes most of me refraining to test some of those other ways myself comes from fear of breaking a device(because I have to admit, my electrical understanding is very limited, so I tend to stick to rule 1: "don't be an idiot" ^_^). but the idea of adding a resistor at all time looks pretty good to me as far as staying safer is concerned.
 
I've seen a measurement box design that has a more complicated design and apparently some protection, so maybe I'll get the components and try it.
 
thanks again, you've tried all I was thinking about and more, that's very nice of you.