Turning One Headphone into another?

[Dreadhead]

Hi everyone:

I'm thinking of getting a pair of these:

Stereo microphones, USB Microphones, Preamplifiers, Digital Recorders, Cables and more at Rock Bottom Prices from The Sound Professionals - Great deals on Microphone, Preamplifier, Digital Recorder, Cable and more!

and then using the convolver plugin for Foobar to turn a pair of headphones (say HD650s) into a pair of R10s (if I can get a measurement on a pair). I can sort of do the same thing already with my equalizer but a convolution based method would capture a lot more of the intricacies than my method would. I think it's likely that it could even capture the effects of the tubes in the line etc which would be interesting.

I think this would be a lot of fun but is anyone else interested?

[Torpedo]

Kind of not. There's no way you can mimic by EQ, reverb, delaying, phase changing, etc the distortion pattern, impulse response, sensitivity, energy storage and many other features of a pair of cans to make them sound like anything else. There's much more to transducers own sound than frequency response. However I'm sure you'll have a lot of fun trying

[Dreadhead]

Kind of not. There's no way you can mimic by EQ, reverb, delaying, phase changing, etc the distortion pattern, impulse response, sensitivity, energy storage and many other features of a pair of cans to make them sound like anything else. There's much more to transducers own sound than frequency response. However I'm sure you'll have a lot of fun trying

Actually convolution does capture phase response and impulse response, that's the point over straight equalization. I don't know about "energy storage" but I assume you mean the driver response which again should be captured. My understanding from everyone that I have spoken to is that with the Smyth system they could not tell the difference between the speakers and the headphones which is pretty damning. It's nice to have transducers that naturally do these things but if you can work out a way to get your system to do it anyway (with some work) I can't see why not.

Tubes will likely be a bigger issue. I'm thinking it might capture the response at one volume but not at another where the effects will differ slightly.

I have already done this with my headphones just through equalization and they all end up sounding about the same (to a lot smaller margin than people would like to believe, and my bet is that the margin is mostly due to technique and equalizer limitations). I've even done it to R10s

Anyway it's all fun. I don't expect people to want to buy a "black box"that turns their headphones from one to another but I can at least build myself one

[Torpedo]

The issue because I believe it won't work is that no matter how much you try to make an impulse response fast by software/hardware if the phones cannot match it. The same goes for intrinsic driver distortion pattern, it will add its signature over the one you're faking using that device.

[Dreadhead]

The issue because I believe it won't work is that no matter how much you try to make an impulse response fast by software/hardware if the phones cannot match it. The same goes for intrinsic driver distortion pattern, it will add its signature over the one you're faking using that device.

Interesting. I don't agree (as long as you have a powerful enough amplifier with enough control) but a valid point none the less.

I think the biggest weakness will be the variation with volume but I am not sure how strong that is.

After Christmas I think we may find out.... I have lots of Fourier transform math to check though to make sure I can invert the response measured with the set of phones you own to be the one measured with the phones you want but it appears to be true.

[Torpedo]

I'm sure you'll have a helluva fun playing with that toy, but there are physical limitations that the device cannot overcome. Had you measured yourself several cans and had you seen their different responses, then you'd be as sure as I am that some things can't be "recreated" by software. Yet another whole story is that you can hear that and that you're able to tell the difference between the real thing and the recreation, which I can understand most people won't tell, maybe even myself hahaha.

Keep us posted on your findings, it will be a nice read

[grawk]

I think if the software compensates for the affect of both headphones, it stands a pretty good chance of working.

[Torpedo]

Come on mate, if your transducer just can convert into sound 60% of the energy that reaches it spending 2ms in doing so, and you're trying to fake the response of a phone capable of converting 80% in just 1ms, it's impossible you can match that no matter how hard you try. It's like pretending to match the acceleration of a 2 strokes engine by overloading a 4 strokes one.

[aerius]

It's absolutely impossible to make this

look like this with any sort of electronic signal processing known to man.

The problems, that being resonances, echoes, stored energy, and so forth are mechanical and physical in nature, and can't be compensated for in the electrical domain. To alter those properties you have to go in and make physical changes to the device, for example, adding foam to soak up echoes or changing the driver material to help with resonances and stored energy.

To use an example, let's say you have a subwoofer with a shitty cabinet design, the cabinet rattles like hell whenever there's a note between say, 30-80Hz. There is no way to get rid of the cabinet rattle with signal processing, you have to go in and make physical alterations to the cabinet. It's the same thing with headphones & speakers, though usually on a much degree. I can't make a Quad ESL-57 sound like an Klipsch or Altec horn system & vice versa. I can't make a B&W 805 sound like an 801. I can't make a Lowther sound like a Wilson Sophia or Watt/Puppy.

[CarlSeibert]

I imagine you could make a more or less perfectly performing headphone mimic a lesser one. But if you had a more or less perfectly performing headphone, why would you.............

-Carl

[Dusty Chalk]

So how were you planning on doing this -- get the characteristics of your headphones, get the characteristics of the target headphones, take the inverse of your headphones, and then impose the target headphone "image" on the inverse of the extant headphone image? What software were you going to use to do that? I don't think Sony's Sound Forge Acoustic Mirror does all that. Matlab?

[JBLoudG20]

Mathematica, maybe, but never Matlab.

Matlab>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>Mathematica

[Dreadhead]

Actually out of pure laziness I will use Room EQ Wizard to get the IRFs and then do the math in Matlab since I have all my code already written in matlab and it has a WAV reading algorithm already. I'll then use the convolver plugin already built for foobar.

As far as lesser and greater headphones I've measured R10s and their frequency response is crap but they sound lovely. I don't know what their phase response is.

Btw I hate mathematica, I use maple for symbolics (not that I do that much these days, mostly pencil and paper)

[luvdunhill]

Mathematica is the suck. Octave is the win.

[JBLoudG20]

Mathematica is the suck. Octave is the win.

I'd use Octave, but I paid for a Matlab license, and need full compatibility. I have both installed, however.

[Dreadhead]

hmmm Octave looks good. I will likely use that instead since I would have to use the the Matlab from work and I would rather not.

[Dusty Chalk]

So you'll use Room EQ Wizard just to obtain the impulses? And you need more than EQ measurements to get the R10, you need the impulse. It's not just EQ, or else we're back to...oh, who was that guy on head-fi...not gloco, but it was around the time when people like me and MacDEF and kelly and gloco and jude and kwkarth were the most active people on head-fi...

I have my doubts that it will work, but it's definitely an interesting experiment, and I definitely encourage you to do it, if only because I'd be curious to hear the results. I really think you need to start with a full-frequency headphone like the Sony MDR-7506/-V6, though. If you tried to impose bass on a bassless headphone (like "those" R10s or the K1000), you'll just get distortion.

...matlab ...has a WAV reading algorithm already.

This was why I specifically asked about Matlab, Jacob. I knew it had wav interoperability, don't know if Mathematica or Octave do.

...and then foobar has an convolver plugin? So what does it take, just a wav file (or equivalent)? That's pretty cool...there's probably a HRTF impulse out on the web somewhere, you should try downloading it and plugging it in...instant crossfeed.

Edited December 10, 2008 by Dusty Chalk

[cclragnarok]

Hi everyone:

I'm thinking of getting a pair of these:

Stereo microphones, USB Microphones, Preamplifiers, Digital Recorders, Cables and more at Rock Bottom Prices from The Sound Professionals - Great deals on Microphone, Preamplifier, Digital Recorder, Cable and more!

and then using the convolver plugin for Foobar to turn a pair of headphones (say HD650s) into a pair of R10s (if I can get a measurement on a pair). I can sort of do the same thing already with my equalizer but a convolution based method would capture a lot more of the intricacies than my method would. I think it's likely that it could even capture the effects of the tubes in the line etc which would be interesting.

I think this would be a lot of fun but is anyone else interested?

You may be able to do EQ with phase adjustments using a convolver, but I don't think harmonic distortion can be simulated or canceled out with a convolver. With more complex processing, you may be able to handle harmonic distortion as well.

[Dreadhead]

So you'll use Room EQ Wizard just to obtain the impulses? And you need more than EQ measurements to get the R10, you need the impulse.

...and then foobar has an convolver plugin?

1) Room EQ calculates both and keeps track of phase. Rest assured it gives the IRF.

2) yes it has a convolver and it takes a wav file. If I get the mics I can use it for the HRTFs as well.

I have sorted out the math and have the inverse of the IRF etc working. Good times. I will see what I end up with later. Distortion should end up as both phase and amplitude response and hence I should capture it but we shall see. The guy who built the convolver has stuff to model tubes etc on his website.

[cclragnarok]

Convolution might be able to simulate some other attributes of "tube sound," but it should not be able to simulate harmonic distortion.

[Dusty Chalk]

Yeah, it's going to have a hard time capturing anything that is actually a function of the signal (function being in the strict mathematical sense of the word, one which takes as at least one of its inputs amplitude vs. time or whatever). It can only capture those elements that can be reversibly imposed on a signal. That'll include compression as well, which is just a mild form of distortion.

Just try to keep your distortions to a minimum, and factor that out of the equation as much as possible. I'd be curious enough to hear what kind of success you could get just getting the headphones right.

[Dreadhead]

It works muhahahahahah

I am listening to HD650s that sound exactly like SA5000s. This is so freaking cool.

[Dreadhead]

Convolution might be able to simulate some other attributes of "tube sound," but it should not be able to simulate harmonic distortion.

May I ask why not?

[cclragnarok]

May I ask why not?

Unlike some other posters, I was only talking about what convolution can do in theory.

EQ and phase changes are possible. Time delay is a kind of phase change, so it is possible. Combinations (sums) of these effects are possible, which include things like adding or canceling out echos/reverbs.

Harmonic distortion, which adds harmonic frequencies as a function of the fundamental frequency, is not possible. If the input to your convolver is a pure sine wave, the output can only be a pure sine wave with the same frequency.

[Dreadhead]

Unlike some other posters, I was only talking about what convolution can do in theory.

EQ and phase changes are possible. Time delay is a kind of phase change, so it is possible. Combinations (sums) of these effects are possible, which include things like adding or canceling out echos/reverbs.

Harmonic distortion, which adds harmonic frequencies as a function of the fundamental frequency, is not possible. If the input to your convolver is a pure sine wave, the output can only be a pure sine wave with the same frequency.

hmm well I will think about that, I had not thought about it that way but I can see where you're coming from. Luckily I'm into SS so I can't care about tube harmonic distortion

As far as turning the phones from one to another it appears to work though.