Waveform Fun! Amp/Headphone output comparison

[purrin]

I captured the output from two different amps with the same headphone (HD800). The process to capture the waveform is indicated below:
 
PC--USB-->PS Audio PWD2 DAC-->[amp?]-->HD800--sound-->panny capsule mic-->ADC-->wavefile
 
Obviously a significant factor will be the quality of the ADC, but we will have to live with that. The three outputs also have slightly different offset, and obviously the HD800 will introduce some gross errors. But let's just look at the waveforms for fun! The three waves are the following not in order: 1) original wav file; 2) amp A; 3) amp B. I will say no more about the amps used and let you guess which is which.
 

 

[scootsit]

Quote:

I captured the output from two different amps. The process to capture the waveform is indicated below:
 
PC--USB-->PS Audio PWD2 DAC-->[amp?]-->HD800--sound-->panny capsule mic-->ADC-->wavefile
 
Obviously a significant factor will be the quality of the ADC, but we will have to live with that. The three outputs also have slightly different offset, and obviously the HD800 will introduce some gross errors. But let's just look at the waveforms for fun! The three waves are the following: 1) original wav file; 2) amp A; 3) amp B. I will say no more about the amps used and let you guess which is which.
 

I have a few questions about these, and a lot of your plots, actually.
1. Are the x and y axes the same in all 3? The last one looks zoomed slightly differently.
2. Is there raw data associated with these? I would love to do some statistics.

 

[EthanWiner]

Quote:

Originally Posted by purrin /img/forum/go_quote.gif
The process to capture the waveform is indicated below:
PC--USB-->PS Audio PWD2 DAC-->[amp?]-->HD800--sound-->panny capsule mic-->ADC-->wavefile

 
If your goal is to assess the amplifiers, you are seriously polluting the results by including a loudspeaker and microphone in the measurement. Simply route a wire from the output of the D/A back into the input of the A/D, then record.
 
--Ethan

 

[sphinxvc]

Awesome.  Middle one looks closer to the original.

 

[purrin]

The goal is to assess amplifier + headphone. (Edits made to thread). If you've ever seen klippel system results, transducers do not create static loads or behave nicely, so I wanted the transducers to be part of the equation. Given the complex dynamic interactions between amp and transducer, I don't think we should ever separate them totally. However, will do an amp or DAC into dummy load one of these days. 
 
I agree that the mic, ADC, and environmental noise will be not insignificant pollutants. Unfortunately the microphone/ADC will need to be part of the equation for this test. The question is even with these pollutants, do we see any differences? Can we see characteristics which might indicate which one is the original, and which of the others are more similar to the original, and how they are similar. The devil is in the details. And more data sets are needed.

 

[purrin]

Quote:

I have a few questions about these, and a lot of your plots, actually.
1. Are the x and y axes the same in all 3? The last one looks zoomed slightly differently.
2. Is there raw data associated with these? I would love to do some statistics.

 
They are the same. DC offsets are slightly different.

 

[scootsit]

Quote:

 
They are the same. DC offsets are slightly different.

Cool. I've often considered with headphone waveforms and what not, if we could do some comparisons mathematically. Are there tables of data or only plots?

 

[purrin]

Quote:

Awesome.  Middle one looks closer to the original.

 
Bottom one is the original. Now which amp is better? 

 

[sphinxvc]

Ha!  
 
/ I second the request for mathematical comparisons.  At this level it's too difficult to see the differences, wouldn't you say?

 

[muad]

How was the original measured?
 
to me the top and bottom are very close... but to what control?

 

[mikeaj]

Cool.  Could you (at least at some point) upload a similar picture with DAC -> amp -> ADC, removing headphone and microphone?  I can see the motivation behind wanting to go through the transducers though.
 
I wonder how much difference, if any, you could capture on a very low-Z amp between DAC -> amp (loaded with headphones) -> ADC, and DAC -> amp (loaded with resistor) -> ADC.
 
 
As for the pictures, I have neither golden ears nor golden eyes, so I doubt my ability to eyeball and evaluate differences.  But guessing is fun.  If the bottom one is the original, then
 

Spoiler: Warning%3A%20Spoiler!

Top looks better (closer to bottom) than the middle.  What's with the middle one?  Lower bandwidth, rolled off FR, what?  Looks can be deceiving, since potentially some of the change of "details" may be just in ultrasonics.  It would help to know the sample rate of the recording.

 

[purrin]

Quote:

Ha!  
 
/ I second the request for mathematical comparisons.  At this level it's too difficult to see the differences, wouldn't you say?

 
Yes. That's one of my points. Difficult to see differences, but I still think we can see some differences. Taking into account very low magnitude lower frequency oscillation and offset, I think one waveform is more precise and closer to the original at the bottom.
 
Math and more data points will come in time.

 

[purrin]

Quote:

Cool.  Could you (at least at some point) upload a similar picture with DAC -> amp -> ADC, removing headphone and microphone?
 
I wonder how much difference, if any, you could capture on a very low-Z amp between DAC -> amp (loaded with headphones) -> ADC, and DAC -> amp (loaded with resistor) -> ADC.
 
 
As for the pictures, I have neither golden ears no golden eyes, so I doubt my ability to eyeball and evaluate differences.  But guessing is fun.  If the bottom one is the original, then
 

 
Will do. These things take an extraordinary amount of time! I've got a few more data sets, most of which I haven't looked at in detail, but I wanted to post these two as a teaser. Always a good thing to get us thinking!
 
I agree with your assessment.
 
I actually though the middle one was the original at first. (I had my wife rename the files and keep track which was which.)
 
No answers now, but will post more data sets.

 

[Steve Eddy]

Instead of looking at signal graphs, run them through Bill Waslo's Audio DiffMaker.
 
se

 

[mikeaj]

Yeah, this isn't easy.  Take your time, especially when headphone measurements are involved.  I don't even want to think about the environmental noise and other challenges with these setups.