[rsaavedra]

Quote:

Originally Posted by Tim D True, but graphs are only really useful when compared with other measurements made under the same conditions. Can they get close or have the exact same reference as what was done who knows how long ago?

I assume they would use the same measuring equipment and under similar conditions they have established in all the measurements they've made. I think these measurements have just a couple of years, my guess is they still have and use the same measuring equipment, though only Headroom can say.

 

[Andrea]

Quote:

Originally Posted by rsaavedra I just find it somewhat paradoxical they went through the trouble and work of painstakingly measuring for instance Ipod earbuds, while the R10, PS1's, HP-2's are still missing. By the way a huge missing entry on my original list. The HP-1000 should of course be in that list.

Perhaps it's not so paradoxical but it reflects their thinking on certain headphone brands' reliability and/or they take into serious account the reasonableness of the price-performance aspect, which I would find shareable.

 

[Ref]

It would also be nice to find some Philips headphones there

 

[rsaavedra]

Quote:

Originally Posted by Andrea Perhaps it's not so paradoxical but it reflects their thinking on certain headphone brands' reliability and/or they take into serious account the reasonableness of the price-performance aspect, which I would find shareable.

Yes taking into account price-performance of course makes sense to choose headphones to measure, but come on, we are talking about Headroom, they are headphone geeks extraordinaires, it's paradoxical they have gone through the work of measuring so many inexpensive headphones while completely leaving out the cream of the top models in several brands.

 

[donaldekelly]

While their chart of the Sennheiser hd 212 seems right on (TOO MUCH BASS)

the chart of the HD280 pros seems off and is different from Sennheiser's own chart.

Maybe they should recheck all or some of the charts.

 

[saint.panda]

Quote:

Originally Posted by rsaavedra I do believe the log scales are incorrectly marked on their graphs online in general.

Yes, I’d also prefer a truly logarithmic scale so it would offer a better representation of what we actually hear.
edit: actually they're correct, my bad.


As for the R10, Qualia graphs, I’d very much like to see them although more out of curiousity. That reason is that, at times, my hearing seems to drastically differ from what the Headroom graphs supposedly tell me although, as so often, maybe the fault is to be looked for in the interpretation rather than the results themselves. I’m pretty sure Headroom knows that they’re doing but why is that nowhere in the world do I hear a 6-8Khz dip on the Ultrasone HFI650 compared to the HD 650 whereas the midrange measurements seem to reflect my actual listening results in a much more accurate way? Since I don’t understand the full extent behind these phenomena I can only offer rehashed opinions: Different people have different ears, different transfer functions, different perceptions of sound, different ways of wearing a headphone and I don’t really consider my ears the best RTA available either. But what is it that I really hear? If frequency amplitudes would be the sole answer, than the RS-1 should sound like the PX100 but I do find that the RS-1 has more pronounced highs. But hey, maybe Headroom wants to boost their PX-100 sales with that.


Now, Headroom certainly isn’t to blame for providing the date. But wait, those graphs are something people like myself can easily understand. “Look, there’s a scientific measurement that will tell me everything about a headphone without me actually having to listen to it. It’s measured date, so it’s got to be correct, right?”, “Oh no, those results don’t correspond to what I hear at all. Screw those graphs.” Yes, exaggerated statements but they have been running through my mind more than just once.


Those graphs should be taken with a huge grain of scientific salt and regarded as a complementing side dish. It’s just as wrong to simply ignore science when it’s available as dispelling it just because they don’t match with “reality” and as in all things audio both approaches are probably even dangerous.


Their graphs have been very useful to me for EQing some of my headphones so I can more easiliy compare headphones and assess their "audiophile" qualities without mistaking detail for brightness (7Khz and fingers moving up the fretboard...).


Nevertheless, I think it would be appropriate to put up a warning that the frequency responses don’t necessarily correlate with the end results and most of all, should not replace actual listening sessions. Anyway, I’m sure it’s just all that dummy’s fault.


Finally, I’d like to applaud Headroom that despite all the cost-intensive and from a vendor’s point of view probably faintly rewarding hassle (at least people will think that these guys know what they are talking about), they have been taking and providing their measurements to and for the public. Why would they take data from a Sony CD3000 which they don’t even sell? That’s something only true headphone aficionados would do.


Ahh, whatever. Since I’ll never be able to afford a real R10, at least show me the graphs!!

 

[JaZZ]

The scale on HeadRoom's FR graphs is logarithmic, but the 20...200...2,000...20,000 references are really confusing -- I would have preferred 10...100...1,000...10,000 instead.

The lines should be read as follows:

20...40...60...80...100...120...140...160...180...200...400...600...800...1,000...1,200...1,400...1,600 ...1,800...2,000...4,000...6,000...8,000...10,000...12,000...14,00 0...16,000...18,000...20,000

Yes, it would be cool to have Qualia, R10 and HP-1000 graphs, but I guess HeadRoom only includes headphones available with them.

 

[rsaavedra]

Quote:

Originally Posted by JaZZ The lines should be read as follows: 20...40...60...80...100...120...140...160...180...200...400...600...800...1,000...1,200...1,400...1,600 ...1,800...2,000...4,000...6,000...8,000...10,000...12,000...14,00 0...16,000...18,000...20,000

Ah read that way the scale does make sense, didn't think of that sequence, but now that you mention it, it makes perfect sense.

Along the same lines, the scale in the harmonic distortion graphs having jumps from 300 to 3KHz should be read:
300, 600, 900, 1200, 1500, 1800, 2100, 2400, 2700, 3000.

Definitely using 10-100-1K-10K would be much easier to read though.

 

[rsaavedra]

Quote:

Originally Posted by JaZZ Yes, it would be cool to have Qualia, R10 and HP-1000 graphs, but I guess HeadRoom only includes headphones available with them.

Not really, they have graphs for the Orpheus also Sony CD3K, which aren't for sale on Headroom.

 

[Tim D]

The scale is fine, the freq. labeling may be off? I find it bizarre how people complain about how it doesn't reflect what they hear especially towards the frequency extremes. Like most people even know what a 6 or 8 khz tone really sounds like off the bat. Unless they can prove to me that they can identify what frequency a tone is by ear how can people say 'that dip or rise doesn't make sense?'. Now I am sure some people *can* do that, but I'd have to assume it is not within the typical curriculum of education for most people. I can hear high frequency pitches from the TV for example, but had no idea it was around 16khz or whatever until someone else said that was the frequency of sound typical TV flywheel components or whatever give off. Can anyone tell me right now what the frequency composition of a typical police siren is by memory? If not, then how meaningful is it that you can read that graph and say...boy I don't hear that rise or pitch at 7 khz whatsoever!

For the most part most headphones recommended here are pretty flat in the midrange. The highs are typically heavily smoothed results. Not to mention it is usually the high freqs (or even the lowest freqs as well) where ears can differ the most. Since most of the graphs only differ at extremes, most *good* headphones typically only differ the most at extremes (as far as freq/amp response), and most ears typically differ the most at the freq extremes, how people end up over-analyzing the data is beyond me. The same thing for speakers, high freq response is what changes the most as axis or positioning is changed.

Another thing to consider is if they did freq response curves of most good tube amps they will *still* be mostly flat (and if its SS most definitely flat), yet it does not reveal in entirety much of what the amp would sound like or even mention harmonics, distortion, speed, etc. So if you compared a tube amp freq curve to a SS amp, and saw maybe 2-3 db difference at most, you can safely assume that all the sonic characters you *can* hear between the two is not accounted for on *that* particular graph.

Bottom line, some of the worse headphones like the Sony 600's measure poorly even in the midrange and the graphs will show it. If a graph shows a headphone to be incredibly poor in certain areas...most likely there is *something* there to account for it. Likewise if an overwhelming majority of head-fiers had a certain consensus as to what a headphone sound like, most likely there is *something* there to account for it. What you shouldn't do is look at very slight differences or nuances in graph data and assume it must equate to very overwhelming results in sound, just like you probably wouldn't/shouldn't make the biggest deal out of one persons 'meet' impressions. It is just interpretation of data. All of the consensus or opinions held on head-fi is also just 'data', what you make of any of it is up to you.

There was a time where Headroom wanted to open up the test equipment for people to send stuff in but they never got around to it yet. Again, I'd have to imagine its a lot of work (if you want good results that is).

 

[markl]

Headroom doesn't carry Sony cans, I think that's the primary reason they haven't been measured. Some time ago (years) Tyll got to hear some R10s at a meet (probably his own World Of Headphones tour but I forget) and he concluded that subjectively, they were probably better than the HD600 (then his and Headroom's favorite cans). The post where he says so is stored somewhere here on Head-Fi.

 

[rsaavedra]

Quote:

Originally Posted by Tim D There was a time where Headroom wanted to open up the test equipment for people to send stuff in but they never got around to it yet. Again, I'd have to imagine its a lot of work (if you want good results that is).

Yes, actually for good results ideally you would measure at least a few headphones of the same model, and then you would have some other very important data item, unit to unit variability. But then, how much of that variability is really introduced by differences in the headphones, and not introduced by variations in the testings/measurings of them? (e.g. because of slightly different position of the headphones on the dummy head for instance). So lots of work indeed, and probably not worth it to that level.

One thing I think doesn't really help much in the graphs is the plotting of left channel vs. right channel, because of that same variability, it is not certain whether it's due to variability between drivers, or something introduced by the measuring. I think would be best to probably just measure let's say a pair of headphones of the same model, average all the measurements and list just one curve, probably with the variability info attached. Lots of work indeed no doubt.

 

[Tim D]

Speaking of unit to unit variability and potential 'payoff' for their efforts, one idea would be to cater to the truly neurotic and perform a higher level of 'QA' and 'hand-matching' on top of what is done from the Senn factory and sell the 'uber' Senns at a hand-matched premium with individual graphs and all. I'd figure most people however would rather be ignorant of how well matched their headphone drivers are, but I'm sure there could be a market there somewhere. Thus they offer a service that not many others can do by performing this extra 'hand-matching'. Wheter or not Senn would like one of their distributers to really reveal how 'good' their current supposedly 'hand-matched' criteria is is yet another matter.

Not a horrible idea since for all the people saying graphs don't matter, its probably only a QA or hand-matching process that differentiates drivers between the 580/600/650.

 

[bigears]

Some more data to get a clearer picture...
If Headroom could do just more 3 sets of measurements,
2 sets for the 2 Ultimate Ears models, and 1 set for the Stax model,
Those numbers would improve quick...

But what i REALLY would like to see is ...

[size=medium]Beyerdynamic DT880 with Cardas Upgrade Cable
Beyerdynamic DT880 with Balanced Cardas Upgrade Cable
Sennheiser HD595 with Balanced Cardas Upgrade Cable[/size]

Please, please, please ???

 

[Tyll Hertsens]

Wow! Lots of questions about graphs.

Quote:

Originally Posted by rsaavedra I think there are some extremely important high-end entries sorely missing in their database of charts

I agree. I’d love to add more headphones to the dbase.

Quote:

Originally Posted by rsaavedra I'd like to ask some Headfiers whether they would like to lend their Headphones to Headroom for a week or so, so that they could measure those headphones and add the corresponding graphs to their wealth of graphs online. Headroom is one company that lends gear for meets themselves after all, and there should be nothing to worry about in lending Headroom even the most expensive gear for this.

Thanks for your trust. I do believe that this could be accomplished this summer. Right now would be a bad time. Do you guys have an interest?

Quote:

Originally Posted by elerno I question the accuracy of the measurements to begin with.

We do measure headphone in accord with the ANSI and IEC standards. Headphone graphs are not like speaker graphs in that there is significant interaction and interdependence between the device under test and the test equipment itself. A person definitely has to get used to what the graphs look like. We have had occasional problems measuring headphones, especially at the beginning and with sealed cans. The PX200 was particularly troublesome. But Sennheiser was pretty satisfied with our measurements; I think you can feel pretty confident, too.

Quote:

Originally Posted by bigears 1) AKG K1000, is in Headroom site for sell I know this headphone is harder to measure due to his ergonomics, but shurely that would be interesting...

The test set-up for these cans is different; but I do tink we’ll get around to it this summer also. There are some other cans that have unusual set-ups: wireless cans and noise cancellers for example.


Quote:

Originally Posted by bigears Example: to compare the senn hd650 hd600 hd580 for frequency response... Error message: mumber of points x and y are not equal...

I guess I’m not sure of what you’re talking about.

Quote:

Originally Posted by bigears 3) Comparing the Ety´s for isolation the measurments falls off the graphic range...

We kida did that on purpose. When we include all the Ety data the graph becomes too insensitive to compare headphones that don’t have much isolation. We figure that essentially saying that IEM isolation is off the chart good is reasonable.

Quote:

Originally Posted by bigears The legend (if 4 phones) sometimes overlaps the graph...

Oooo.. Good call on that one. I don’t know if we can do anything about it though. Thed graphing module is a third party written piece of code and I’m not sure how much flexability we have to change things. I’ll have a talk with Matt.

Quote:

Originally Posted by bigears Currenty is 20hz, 200hz, 2000hz, 20000hz, Also is not clear what the other vertical lines are...

OK. I’ve been round and round on this internally and a lot of times it takes folks a while to get this so I’ll spew it out and you can not like it if you want (there are things not to like for sure), but the graphs are right.

We cannot make the graphs with 10, 100, 1000, 10000 as major tick marks because we would HAVE to start at 10 (which isn’t so bad) and stop at 100,000 which is a waste of space. When we go by 20’s the tick marks are (left to right) 20,40, 60,80,100,120,140,160,180,200, 400, 600, 800, 1000, 1200, 1400, 1600, 1800, 2000, 4000, 6000, 8000, 10,000, 20,000.

This is a perfectly legitimate log scale and going in 20’s is commonly done (in my experience) in the audio world. People don’t get it though. Sigh.

Quote:

Originally Posted by bigears As for Harmonic Distortion, is also not clear which frequencies are between 300hz and 3000hz.

300, 600, 900, 1200, 1500, 1800, 2100, 2400, 2700, 3000.

We have been thinking of different ways to show the distortion data so that they don’t overlap so strongly.

Quote:

Originally Posted by Tim D For one I'd like to know how THD or freq response changes according to input level or volume or how much variation there is between *same* make and model. Lets not forget the entire arena of headphone amps that can be measured.

You’re right there is a lot of work that goes into the graphs and there are a LOT of other things we could do with the equipment given endless amounts of time---like your suggestions above. But we do get something out of it---we hope. We do hope to get some good will out of publishing graphs. Sure, some folks rag on us, “Measurements don’t tell the whole story” and the like. It’s true, too, but measurements do tell a significant part of the story; and we do perceive a significant relationship between what we hear and what we see in the graph. We want to give that vision to people because we believe it will cause better headphones to be produced in the long run because of it.

Quote:

Originally Posted by rsaavedra My point is that if those few are missing just because they don't have them, then people should lend those headphones to Headroom, so measurements of them can be added to the database for completeness sake, especially for completeness on the high-end of things.

I’m repeating this because I agree so strongly. I do want to do more cans and we would love to get folks to send them to us for measurement. Bring it up again in June and we’ll move forward.

But I have to say that when we introduced the graphs there seemed to be a collective “Ho humm” from the community. I didn’t seem like people really appreciated the measurements and the work so we kind of let it sit with just the cans we carry.

True, but graphs are only really useful when compared with other measurements made under the same conditions. Can they get close or have the exact same reference as what was done who knows how long ago?[Tim D]
That’s the beauty of buying the real deal test equipment for the job: as long as it remains calibrated (and we have a test jig to test that on an ongoing basis) the measurements should be good and comparable over time. But we have gone back and measure the same cans over and over, over time and get very good results.

Quote:

Originally Posted by saint.panda I’m pretty sure Headroom knows that they’re doing but why is that nowhere in the world do I hear a 6-8Khz dip on the Ultrasone HFI650 compared to the HD 650 whereas the midrange measurements seem to reflect my actual listening results in a much more accurate way?

Maybe because your ears are a different shape than our test head. This brings up a very important point: Our measurement system is designed very carefully to deliver a normal (average) response. But people hearing systems differ widely (head size, pinna shape, ear canal impedance characteristics) so your particular response may be as much as … oh … 10dB different in places than our test system. It’s just not very likely. On the whole our system is very close to the “normal” response. Most folks hearing will vary very little from our measurements. But some will. Remember: headphones are not like speakers, there is no objective far-field frequency response to be measured. The sound you get is a combination of the headphones and your ear and head in them and the two cannot be disentangled.

On the up side, all our measurements are done pretty much the same way (IEMs are a bit different) so relative performance should be fairly close to what you experience.

Quote:

Originally Posted by rsaavedra But then, how much of that variability is really introduced by differences in the headphones, and not introduced by variations in the testings/measurings of them? (e.g. because of slightly different position of the headphones on the dummy head for instance). So lots of work indeed, and probably not worth it to that level.

We did a bit of experimenting there and, though there is some positional sensitivity it was pretty minor until you got to the extremes where it was actually hard to position the headphones that wrong on the head. The one general exception is that it seemed the sealed cans were a bit more positionally sensitive and the PX200 was particularly troublesome as it seals in the pinna and not around the ear.