solderdude
Administrator
measureutternutter
Posts: 4,886
|
Post by solderdude on May 19, 2014 6:37:50 GMT
Tonal balance seems to be spot on.
The HF roll-off above 16kHz (also present in the M50) is a shame but most people do not hear much above 16kHz anyway.
Also there are peaks at 8, 10 and 13kHz
I may order one from Thomann and send it back if I don't like it.
These newer higher definition files differ substantially on a number of headphones compared to the 128kb MP3 I downloaded yesterday from their website.
|
|
|
|
Post by c61746961 on May 20, 2014 21:00:03 GMT
I hacked together a tool to plot Audacity's frequency analysis numbers against the source file and got these results: DT-990 600 ohm:  ATH-M50x:  ATH-M40x:  I used the 'metal' sample of the recording to generate the frequency analysis. Sub-bass response probably isn't accurate. |
|
|
|
Post by c61746961 on May 20, 2014 21:15:47 GMT
SRH440:  DT1350:  V-Moda XS:  IMO, in this context, both Audio-Technica pairs seem remarkably well-behaved. |
|
solderdude
Administrator
measureutternutter
Posts: 4,886
|
Post by solderdude on May 20, 2014 21:23:18 GMT
Well... the DT990 isn't that far from the truth.
I think the M50x has moar bass than what the plots suggest (at least the older M50 I have was very bloated)
The M40x seems to have slightly more bass than the M50x which could be strange.
The peaks are indeed very small, just a few dB.
Haven't seen many headphones this 'straight' and wonder how it would measure on a rig.
Only one way to find out though.
The lows aren't accurate for the same reason Purrin's older plots weren't accurate in the lows either as there are too few 'samples'
If the M40X is really that flat and extended it should be one of the best headphones out there, providing it doesn't show ringing or other deal-breaking behaviour which doesn't show up in the FR
The DT1350 seems 'accurate' and looks quite similar to my stock measurements.
The 440 is indeed a bit bass shy and sharpish in the treble.
This, together with the sound samples make me curious.
Will be waiting for your comments before buying one just to hear how good they are.
Nice tool you made there !
Can you 'export' the data from Audacities analyser ?
When you have the M40X please start a thread about it.
At that price (around E 100) it should be a no-brainer but wonder what the catch is as they could have sold it with another type number on it for 'flagship' prices (between E 300 and E 700) in that case and would probably sell even better than the M40x may do.
|
|
|
|
Post by c61746961 on May 20, 2014 21:43:06 GMT
Yes, the dialog has an 'Export' button which generates a plaintext table of values, then I used matplotlib for Python to generate the plots. I used 8192 samples for the analysis, is there any other setting which would be sufficient for better LF comparisons?
|
|
solderdude
Administrator
measureutternutter
Posts: 4,886
|
Post by solderdude on May 21, 2014 4:43:39 GMT
I also have the option for 16384 samples and when using the last 2 'sections' together show a reasonable 'pink noise'.
This doubles the amount of (LF) samples but might make the resulting plot 'noisier'.
Perhaps playing with other analysis methods is an idea (haven't tried that yet)
The sample rates do have to be converted though to be the same.
Do you also use Ubuntu ?
|
|
|
|
Post by c61746961 on May 21, 2014 4:54:59 GMT
Yeah, I tried with up to 65536 samples and the results were more or less the same, the biggest difference comes with the choice of windowing function, the least roll-off being that of the 'rectangular' window. I'm thinking of performing the fourier analysis myself straight from the source files, or maybe explore the wavelet transfom to find the spectrum instead.
I'm not currently using ubuntu, but I have and enjoyed it quite a bit.
|
|
solderdude
Administrator
measureutternutter
Posts: 4,886
|
Post by solderdude on May 21, 2014 19:18:52 GMT
Maybe listen to the M40X first, the plots won't differ that much from your current efforts.
When you have 'approved' it I am likely to buy one myself if only to measure them and to see how they could be improved (filtering the small peaks passively).
|
|
|
|
Post by c61746961 on May 23, 2014 22:12:51 GMT
So I went ahead and produced an impulse response with Voxengo Deconvolver and used that in ARTA to produce the following plots: FR:  CSD (there's some HF artifact going on):  So there's definitely some ringing going on, not unlike other dynamics in the $100-$300 range. |
|
|
|
Post by c61746961 on May 23, 2014 22:13:25 GMT
Here's the M50x for reference: FR:  CSD:  |
|
solderdude
Administrator
measureutternutter
Posts: 4,886
|
Post by solderdude on May 24, 2014 6:17:15 GMT
Impressive work !
Fun to see how it is possible to derive FR using music.
In essence (when looking over a certain time period) all frequencies will be present and have the character of pink noise.
The M50X seems to look a lot like the old version I measured except for the part below 500Hz.
The early ones were MUCH darker/bassier than the newer models.
Acc to many the 50X and 50 seem to sound the same.
For E 100.- the M40X doesn't look that bad
The treble 'peaks' and ringing may be reducible and the missing sub-bass can be compensated I reckon.
The ringing around 5k and 10k should be addressable.
Perhaps a similar trick as the K550 is possible.
I damp (and thus reduce) the treble part and subsequently reduce the mids and bass with a simple electronic filter creating a flat FR again but with damped highs.
Penalty is reduction in efficiency.
Otherwise the analysis looks promising.
I plan to order one on Monday.
Should be in by Tuesday.
Then measure it and start working on it.
Could become a great headphone if the ringing can be reduced.
good find...
I heard the M40fs a couple of times and it sounds bad (not hifi) so never would have looked at the M40X
The M50 was too bassy and fat sounding to me.
good for monitoring but not as hifi-ish super deal as some make them out to be.
|
|
|
|
Post by c61746961 on May 24, 2014 14:33:08 GMT
Impressive work ! Fun to see how it is possible to derive FR using music. In essence (when looking over a certain time period) all frequencies will be present and have the character of pink noise. The M50X seems to look a lot like the old version I measured except for the part below 500Hz. The early ones were MUCH darker/bassier than the newer models. Acc to many the 50X and 50 seem to sound the same. Yeah, the M50x does sound slightly bassier and boomier in the recordings nonetheless, I wonder what the roll-off is all about, it doesn't show in most other graphs, perhaps missing energy in the recordings? That's what I thought, 5 dB less in highs means less ringing in the audible zone. Cool! I sure hope the sub-bass can be raised to neutral levels since I listen to a lot of electronic music with important content in that range. I did a CSD of the HD25-1 II and it shows a nastier, longed lived ridge in the mids zone, if people can live with that (and a hyped, scratchy treble) and will constantly praise and recommend that pair, I can't see this one being worse. Haha, you will probably get to listen to it before I do. I surely hope so, since the price is right and the build is good and convenient. I listened to the M40fs recording as well and it sounds nasty and cuppy for sure, I did listen to an older M50 and liked everything about it except the fat bass and slight sibilance. |
|
solderdude
Administrator
measureutternutter
Posts: 4,886
|
Post by solderdude on May 24, 2014 15:11:31 GMT
Yeah, the M50x does sound slightly bassier and boomier in the recordings nonetheless, I wonder what the roll-off is all about, it doesn't show in most other graphs, perhaps missing energy in the recordings? The derived plots would account for less bass energy as it will only display a difference. When analysing the original recording there seems to be a lot of LF energy present. It does worry me a bit though as the M50 is quite bassy and both the 40X and 50X do not sound bassy in the recording. I sure hope the sub-bass can be raised to neutral levels since I listen to a lot of electronic music with important content in that range. Will have to be active filtering to raise the subbass, pasively only the treble can be addressed. I did a CSD of the HD25-1 II and it shows a nastier, longed lived ridge in the mids zone, if people can live with that (and a hyped, scratchy treble) and will constantly praise and recommend that pair, I can't see this one being worse. I never understood the hype over the HD25/amperior etc. I measured and listened to one and simply didn't like it. It isn't a neutral headphone at all. Tried one on different occasions and while it has some appealing properties (forward sound and solid bass) the highs are anything but nice. Doesn't show that much in 'gritty' rock music though. I do like headphones that are more universal and sound good with any type of music. In my opinion 'flat' headphones fit that bill but they also ruthlessly show poor recordings. |
|
|
|
Post by c61746961 on May 24, 2014 16:24:22 GMT
Will have to be active filtering to raise the subbass, pasively only the treble can be addressed. If the highs are attenuated 6-12 dB, do you think a 'U' filter to lower the mid-bass and midrange could fit inside the cups? Yes, I share that sentiment. |
|
solderdude
Administrator
measureutternutter
Posts: 4,886
|
Post by solderdude on May 24, 2014 16:49:28 GMT
Can't tell yet if and what filtering can be done passively.
It depends on the impedance characteristics and amount of attenuation that is allowed.
Also the components can not take up too much space as I figure there isn't much room in the cups after damping felt etc. has been applied.
Sub-bass improvement can only be done actively.
|
|
