Latest Thread Images

Featured Sponsor Listings

Headphone CSD waterfall plots

Nov 10, 2011 at 5:11 AM Post #481 of 937


Quote:
moostrkraaft said:
This thread is so informative (and uniquely so) that it would be a disappointment if it didn't develop into its own website, or at least become well-organized like InnerFidelity (or perhaps become part of it?). The consolidation and organization of the information in this thread and its evolution could become one of the most, if not the most, valuable information databases for selecting one's next pair of headphones, and moreover keep the competition more honest :) Regardless of what happens, at least I hope the measurements continue. Thanks Purrin for the enlightening information.
Click to expand...


Well said. 
 
There are always matters of taste, but measurements as these should at least establish some benchmarks so that companies like Ultrasone, for instance, wouldn't get away with so much expensive crap 
tongue.gif

 
Again, thank you Purrin. I'm sure the amount of followers (sounds like we're a cult?) of this thread will rise however. I don't think much of the information here is that hard to understand. Basic application of the xyz-coordinates, mainly. 
 
Nov 10, 2011 at 6:50 AM Post #482 of 937

Quote:
n3rdling said:
In your impulse reponse plots, why does the SR009 pretty much stop at 0 on the way down, while the others shoot past 0?  Is it because they have more moving mass to halt? 
Click to expand...

 
Quote:
purrin said:
That's a good hypothesis. Let me look through my other data, i.e. ESP950.
Click to expand...

 
 
These are things I like to think about. I am a layman, but I will post what I think it happens. I bet some electronic engineers will read this and rectify what is wrong with my guesses.
 
Perhaps at rest mode there is tension on electrostatic and free movement in dynamics. 
 
To halt a dynamic transducer you would need a magnetic force from the coil inter-acting with the permanent magnet (a DSP would insert a feedback on the signal to decrease such distortion). 
 
On the other hand, electrostatics would not need that magnetic force, because the diaphragm tension (which is passive mechanic force) already offer an opposite force to halt the movement.
 
Perhaps that would also explain why dynamics offer a resistive load to the amplifier, while electrostatics offer a capacitive load. 
 
If the diaphragm is tensioned and polarized while going from an excited mode (there is current in the coil or voltage in the stator) to a rest mode (no current in the coil or voltage in the stator), then such polarization and movement to the rest position have to interfere (during that short moment of no signal) in the stator. 
 
Furthermore, really strong DC signal in electrostatic (only if something bad happens with the hardware in the signal chain or bad recording without low pass filters) would collapse the diaphragm into the stator, but then the diaphragm would lose its positive polarity, coming back to rest position, which would translate into AC in the signal path. 
 
p.s.: I am not considering cross-over interference since the vast majority of dynamic headphones behave like a full-range transducer.
 
 
Nov 10, 2011 at 9:21 AM Post #483 of 937
I have a different thought.. I'm a layman too, though.
 
Perhaps what happens in the 'stat is that first rising edge corresponds to signal rise, first falling edge corresponds to signal fall. Electrostats have thin membrane with low tension force value - not relative to it's weight, but in absolute value - and tension doesn't contribute much to getting the membraneback in place. So, the falling edge is completely controlled by signal. When the signal hits zero and subsides, air compressed by the movement rebounds and kicks the membrane back into rising again.
 
On dynamics and orthos, reaction force of the diffusor/membrane is much higher due to higher thickness and contributes to the falling edge much more. Thus, membrane reaches zero deflection before signal reaches zero and falling signal pushes it into reverse deflection, and negative peak is actually when signal comes to zero.  In dynamics, this should be compensated by back EMF, so it's not super drastic, while orthos have much less tight and less heavy diaphragms.
 
Would also be interesting to know how would altering properties of diaphragms work. If one would make an ortho with much thinner diaphragm substrate and hence less tension force, it would come to stop quicker than any other driver, though still being slower that an 'stat on following signal shape and having to use some damping to compensate for that.
 
Anyways, to be sure, we'd have to know what input looks like.
 
Purrin,
Can you provide us with the shape of the signal sent to headphones or at least it's parameters, as obviously it isn't ideal? Even better would be to see actual signal at the outputs of the amplifier, but it's problematic with high voltages of 'stats, so I wouldn't ask for those.
 
We can get some more information from step responce, which isn't 'plagued' by having falling edge right after rising one, which complicates some things. Just look at fronts of square wave responces.
 
Nov 10, 2011 at 10:49 AM Post #484 of 937
 
Quote:
armaegis said:
Hmm, looking at Tyll's impulse graphs it doesn't really look like the Stax he measured are particularly lopsided in impulse response compared to dynamics and orthos... I know they're not quite the same, but still, hmm...
Click to expand...


Different measurement techniques. I mount the headphones on a big sponge to isolate driver behavior. Tyll uses a dummy head.
 
Quote:
nevod said:
Purrin,
Can you provide us with the shape of the signal sent to headphones or at least it's parameters, as obviously it isn't ideal? Even better would be to see actual signal at the outputs of the amplifier, but it's problematic with high voltages of 'stats, so I wouldn't ask for those.
 
We can get some more information from step responce, which isn't 'plagued' by having falling edge right after rising one, which complicates some things. Just look at fronts of square wave responces.
Click to expand...


I'll look into it. I don't have a 'scope, but I'll grab an impulse response of a loopback with the electrical signal.
 
 
Nov 10, 2011 at 12:49 PM Post #485 of 937
Someone would tell me which frequency was used in that impulse?
 
When I wrote my last post I was thinking about a low frequency impulse, which is affected by the diaphragm tension, according to Chu-Moy: 
 
[size=small] (...)[/size]
[size=small] The Transducer Frequency Response[/size]
[size=small]  [/size]
[size=small]
electro2.gif
 
Figure 2[/size]
[size=small] Figure 2a shows the mechanical-equivalent circuit of the transducer. m is the mass-per-unit area of the diaphragm, S is the suspension of the diaphragm in the transverse direction, and 2Rm is the damping in the mid frequencies from the impedance of the air. Fo is the peak force-per-unit area on the diaphragm.[/size]
[size=small] The electrical-equivalent circuit in figure 2b converts m to an inductance of M = jwm henries, the suspension S to a capacitance of S-1farads, the damping to 2Rm ohms, and the force to a voltage Fosin(wt)V. The high frequency response of this circuit is constrained by the inductance M, which is directly proportional to the diaphragm's mass-per-unit area. Therefore, the lighter the diaphragm, the smaller the inductance and the wider the frequency response.[/size]
[size=small]  [/size]
[size=small] (...)[/size]
[size=small]  [/size]
[size=small] http://gilmore2.chem.northwestern.edu/projects/showfile.php?file=cmoy3_prj.htm[/size]
Click to expand...

 
At mid-range frequencies, impedance of the air plays its role for both dynamics and electrostatics (the resistive component). At high frequencies, the diaphragm mass seems to be the critical variable.
 
Certain techniques might help dynamic drivers to reach rest position faster: light voice coils, different coil diameters, different diaphragm patterns and densities (which seems like a vortex), with damping dense edges to halt the diaphragm at rest position at low frequencies (and thin center to improve high frequency response). 
 
 
 
Nov 10, 2011 at 1:20 PM Post #486 of 937
It's an MLS (max length sequence) impulse response. It sounds like white noise (basically think of it as white noise). An impulse response is extracted from playback of the MLS through mathematical voodoo that even I don't understand (I bet arnaud can explain it though).
 
In other words, all frequencies up to a certain point (based on length and sampling rate) are excited in the impulse response. This is why we can mathematically calculate entire CSDs from it. It's pretty wild and rather fascinating. One of these days I will make YouTube video on how this works.
 
Nov 10, 2011 at 1:56 PM Post #487 of 937

 
Quote:
purrin said:
 
I'll look into it. I don't have a 'scope, but I'll grab an impulse response of a loopback with the electrical signal.
Click to expand...



 
This is the electrical loopback of the impulse response for those who are interested. The ideal looks much better than reality.
 
 
      Quote:
anaxilus said:
Lol, me and RD talked about the coherence issue between the two drivers awhile back.  Seems you can see the two distinct red humps separated by a nice ringing ridgeline.  We both liked them though as far as thinking they had something potentially special, just failed in proper execution.
Click to expand...

 
I really do think a K340 type hybrid headphone has some serious potential. It just wasn't developed enough. Too bad we don't have companies like AKG of the '70s anymore that are willing to experiment on wierd stuff. This is all Mr. Kardon's fault. Think what they could do if they had all the measurement tools and technology available today.
 
Nov 10, 2011 at 2:33 PM Post #488 of 937


Quote:
purrin said:
 
I really do think a K340 type hybrid headphone has some serious potential. It just wasn't developed enough. Too bad we don't have companies like AKG of the '70s anymore that are willing to experiment on wierd stuff. This is all Mr. Kardon's fault. Think what they could do if they had all the measurement tools and technology available today.
Click to expand...


I thought you might at least get a kick out of those.  Have you seen a breakdown of the K340?  It has 7 diaphragms per channel!  1 dynamic driver, 5 passive radiators and a stat tweeter.  This is the biggest version of this picture I could find.  Quite an engineering marvel especially considering it came out in the late 70's.
 

 
I think AKG might have been the most innovative headphone company, but only for a short period of time.  I think the K1000 may have been their last interesting headphone of note. 
 
Today imagine a modern dynamic driver like the HD650 or Denon driver tuned to roll off treble sooner in exchange for producing flatter, deeper bass, paired with an electrostatic tweeter with a crossover that's implemented better.  I tend to think that a well designed ortho would meet similar goals and sound better once you've got such a complicated and expensive design, but still.
 
Nov 10, 2011 at 5:47 PM Post #489 of 937


Quote:
purrin said:
Thank you. Unfortunately, judging by the number of page views per day, I don't think a lot of people really care or really understand what I am trying to do. (The LCD# threads are far more popular!) I know however that I have a small audience that has expressed appreciation and that's who I do them for.
Click to expand...



To feel better, look at the number of views of the thread I started about headphone sound simulation to understand its behavior. I think there were about 2 people interested, and that is including me, lol ;)). Doesn't really distract me as I am doing it for self-education purposes as well though.


Quote:
purrin said:

PS1000

We all know about these (in)famous Grados and how they sound. A few people have asked about CSDs for these and RD was kind enough to provide me with one of his own plots:
Click to expand...


^^sounds about right
 
Quote:
purrin said:
It's an MLS (max length sequence) impulse response. It sounds like white noise (basically think of it as white noise). An impulse response is extracted from playback of the MLS through mathematical voodoo that even I don't understand (I bet arnaud can explain it though).
 
In other words, all frequencies up to a certain point (based on length and sampling rate) are excited in the impulse response. This is why we can mathematically calculate entire CSDs from it. It's pretty wild and rather fascinating. One of these days I will make YouTube video on how this works.
Click to expand...

 
MLS is a big buzz word, behind it there's really nothing special...
  1. It's just a peculiar series a random numbers (to ensure a true random sequence) which turns into effective white noise
  2. As part of MLS acquisition, the impulse response of the headphone is extracted from the cross-correlation function between input signal (what's fed to the preamp) and mic signal.
  3. Cross-correlation function might look scary as a word, but it is nothing more than how to sequences relate to each other.
  4. If the output is just the same as the input, for a true random signal, the cross-correlation function is all zeros and a one.
  5. The one is offset a little bit in time which is the propagation delay between in and out
  6. In case of headphone, propagation delay is minimal but headphone dynamics adds some ringing, hence it does not look like a sharp peak (your loop test)
 
Quote:
nevod said:
I have a different thought.. I'm a layman too, though.
 
Perhaps what happens in the 'stat is that first rising edge corresponds to signal rise, first falling edge corresponds to signal fall. Electrostats have thin membrane with low tension force value - not relative to it's weight, but in absolute value - and tension doesn't contribute much to getting the membraneback in place. So, the falling edge is completely controlled by signal. When the signal hits zero and subsides, air compressed by the movement rebounds and kicks the membrane back into rising again.
 
On dynamics and orthos, reaction force of the diffusor/membrane is much higher due to higher thickness and contributes to the falling edge much more. Thus, membrane reaches zero deflection before signal reaches zero and falling signal pushes it into reverse deflection, and negative peak is actually when signal comes to zero.  In dynamics, this should be compensated by back EMF, so it's not super drastic, while orthos have much less tight and less heavy diaphragms.
 
Would also be interesting to know how would altering properties of diaphragms work. If one would make an ortho with much thinner diaphragm substrate and hence less tension force, it would come to stop quicker than any other driver, though still being slower that an 'stat on following signal shape and having to use some damping to compensate for that.
 
Anyways, to be sure, we'd have to know what input looks like.
 
Purrin,
Can you provide us with the shape of the signal sent to headphones or at least it's parameters, as obviously it isn't ideal? Even better would be to see actual signal at the outputs of the amplifier, but it's problematic with high voltages of 'stats, so I wouldn't ask for those.
 
We can get some more information from step responce, which isn't 'plagued' by having falling edge right after rising one, which complicates some things. Just look at fronts of square wave responces.
Click to expand...

 
I have no proof at the moment, but I also believe the oscillation around zero for dynamic headphones is due to the suspended membrane which has some settling time.
  1. In case of a lightweight tensioned electrostatic diaphragm, indeed I can imagine that the ES force has sufficient control on the diaphragm to prevent it from going to much below zero point...
  2. One thing to note though is that the apparent ringing you see in the impulse response is often that of the "dominant" resonance in the headphone response.
  3. For example, most headphones have some more or less strong driver and / or acoustic resonances between 2 and 10kHz and the lower the frequency of the resonance, and the less damped it is, the more you'll see it in the impulse response. There might be other less pronounced resonances (with higher damping for instance), but they're not really visible from looking just at the impulse response.
  4. I am putting my thoughts together on a primer to relate all the time and frequency domain curves from some examples as part of write up on CSD for Tyll, I think it could be helpful
     
 
 
Nov 10, 2011 at 9:52 PM Post #490 of 937

Shure SRH840

I've haven't heard (or read) much about these other than these are highly recommended for what they cost. (Checking out the the impressions thread right now.)
 

Pre-measurement impressions

I don't think I'm a big fan of their tonal balance. Bass is good however and not overdone. It's hard to explain - they lack warmth - as if there is a little drop around 200-500Hz. I've been forcing myself to listen to these but I just can't quite get into them even though there doesn't seem to be anything horribly wrong. They seem to reproduce detail, yet they don't. I know this completely makes no sense! There is a bump at 4k that I find quite distracting. There seems to be a hard edge to the sound. The HD800 sounds laid-back in comparison. I can see a lot of people liking the SRH840, especially at it's price, but I personally don't like them.
 
 
These are not FR plots. The waterfall plots below only go down to 200Hz not 20Hz!
Closed headphones require compensation below ~2kHz which is not applied in these plots.
(In other words, closed headphones are going to look rolled off toward the bass)
 
 
Double ridges (of varying width and duration) at 4k and 5k. This explains probably what turned me off the most. The sudden dip from 500Hz down explains what I meant with "lack of warmth." Otherwise, it's very clean.
 
 

PS: Big thanks to 200poundsofamp for sending me a box of headphones. This kiddies, is how "community" works. Headphones don't automatically appear before me nor do manufacturers freely send me samples for review or independent measurements. Hmm. I wonder why.

 
Nov 10, 2011 at 10:47 PM Post #491 of 937

Yamaha RH-5Ma (modified)

RD modified these by inserting a circular layer of double-sided sticky foam and attached the ear cushions on top of this. 
 

Pre-measurement impressions

RD's mods effectively increases the distance between the ear and the driver, thus transforming a bassy muddy POS into something that is really quite nice. I've actually been using this modified headphone over the SR80 as my primary plug-into-motherboard headphone. This was really one of the pleasant finds out of RD's stash. They don't resolve anything at all though, have some minor issues in the lower treble, but I found their tonal signature pleasant.
 
 
These are not FR plots. The waterfall plots below only go down to 200Hz not 20Hz!
Closed headphones require compensation below ~2kHz which is not applied in these plots.
(In other words, closed headphones are going to look rolled off toward the bass)
 
 
 
The following are the same measurements on a solid plate (for a seal).
This gives us a better picture of the frequency response, but at the cost of some enclosure induced ringing effects that in my opinion we don't hear (ear/brain apparatus filters it out.)
 
 
M-Voda M80 (solid plate) for comparison purposes:
 
 
 
 
Nov 10, 2011 at 11:48 PM Post #492 of 937
Hey, thanks for measuring those.
 
Quote:
Originally Posted by purrin /img/forum/go_quote.gif

They seem to reproduce detail, yet they don't. I know this completely makes no sense!
Click to expand...


I know what you mean, I like them but they're kind of confusing. Really why I wanted to get them measured and see what's going on.
 
But I think I'm even more confused now! I really didn't think they'd look that good, and they never really seemed that fast to me either. Am I reading it wrong or are they really as fast as the AD2000?
 
I thought they'de be kind of a mess in the lower treble too but they seem fine. I guess it's that weird ridge.
 
 
Nov 10, 2011 at 11:57 PM Post #493 of 937
I do believe there is a 'Shure sound' as I've heard and owned quite a few of their products.  It's not quite for me.
 
Nov 11, 2011 at 12:10 AM Post #494 of 937
 
Quote:
200poundsofamp said:
Hey, thanks for measuring those.
 
I know what you mean, I like them but they're kind of confusing. Really why I wanted to get them measured and see what's going on.
 
But I think I'm even more confused now! I really didn't think they'd look that good, and they never really seemed that fast to me either. Am I reading it wrong or are they really as fast as the AD2000?
 
I thought they'de be kind of a mess in the lower treble too but they seem fine. I guess it's that weird ridge.
Click to expand...


Let me dig deeper into the data to see what's going on. The 840 is certainly a strange one. They way the measure you'd think they sound really good.
 
Nov 11, 2011 at 12:19 AM Post #495 of 937
I really liked the 840 when I had it, but it just wasn't comfortable for me. 
 
You must log in or register to reply here.

Users who are viewing this thread

Total: 5 (members: 0, guests: 5)
Back
Top