[HiFiChris]

A month later than I actually intended (transferring money from my bank account to my PayPal account took longer than expected, the package got stuck at customs clearance for two weeks, my microphone input was not powerful enough so I had to buy another interface), I finally get reasonable results from my rig, though it is not 100% calibrated for raw data yet.

By the way, the StarTech interface Luke recommends has got a pretty good microphone input (it can be considered as ruler-flat from 20 to 20000 Hz) that has got lots of power for amplifying the mic signal, but its output is not the very best and has shows some non-linearity in the sub-bass and treble above 10 kHz, so results won't be the most accurate when using the built-in output (unless you take it into account when you build the FR compensation curve).

By the way, does anybody know whether it is somehow possible to apply two FR compensation curves at the same time in Arta? I have searched and tried, but nothing found nothing yet.
It is just that I have built a compensation curve for raw measurements that was modeled after the IEC 711 coupler, but have also made a diffuse-field compensated curve based on 

Hammershøi's & Møller's averaged eardrum response curves from 2008; and now I don't want to create a third compensation file by hand as it would take some more hours to calculate and add it to the "IEC 711 raw" calibration file.​

 

[Brooko]

  A month later than I actually intended (transferring money from my bank account to my PayPal account took longer than expected, the package got stuck at customs clearance for two weeks, my microphone input was not powerful enough so I had to buy another interface), I finally get reasonable results from my rig, though it is not 100% calibrated for raw data yet.

 
Nice to hear you'll be up and running shortly
 

By the way, the StarTech interface Luke recommends has got a pretty good microphone input (it can be considered as ruler-flat from 20 to 20000 Hz) that has got lots of power for amplifying the mic signal, but its output is not the very best and has shows some non-linearity in the sub-bass and treble above 10 kHz, so results won't be the most accurate when using the built-in output (unless you take it into account when you build the FR compensation curve).

 
Yep - mine was same.  I already built my compensation curve to essentially return a ruler flat response (basically just used a loopback - and then inverted the figures that were out to return to a flat line).  Guess you did same?  Gives me more confidence with raw data measurements now anyway.
 

 By the way, does anybody know whether it is somehow possible to apply two FR compensation curves at the same time in Arta? I have searched and tried, but nothing found nothing yet.
It is just that I have built a compensation curve for raw measurements that was modeled after the IEC 711 coupler, but have also made a diffuse-field compensated curve based on 

Hammershøi's & Møller's averaged eardrum response curves from 2008; and now I don't want to create a third compensation file by hand as it would take some more hours to calculate and add it to the "IEC 711 raw" calibration file.​

 
As far as I know - no.  But it's easy enough to apply the values mathematically.  Apply your original curve (loopback etc).  File export (as a CSV).  Open in excel. Convert to 2 columns.  Add your new curve data to the old values.  Concatenate to get back to CSV again.  Save as a mic file.
 
Chris - would you be willing to share your compensation curve values for the IEC 711 standard compensation?  I would be very interested!

 

[HiFiChris]

 

By the way, the StarTech interface Luke recommends has got a pretty good microphone input (it can be considered as ruler-flat from 20 to 20000 Hz) that has got lots of power for amplifying the mic signal, but its output is not the very best and has shows some non-linearity in the sub-bass and treble above 10 kHz, so results won't be the most accurate when using the built-in output (unless you take it into account when you build the FR compensation curve).

 
Yep - mine was same.  I already built my compensation curve to essentially return a ruler flat response (basically just used a loopback - and then inverted the figures that were out to return to a flat line).  Guess you did same?  Gives me more confidence with raw data measurements now anyway.

No, I just use the Geek Out IEM 100 as output instead. Flat FR + low output impedance (important for multi-driver IEMs).
Haven't measured the StarTech's output with multi-driver IEMs yet, but it could be that it has got a rather high output impedance wherefore measurements with all IEMs that don't have a flat impedance response are not accurate.
 

 

 By the way, does anybody know whether it is somehow possible to apply two FR compensation curves at the same time in Arta? I have searched and tried, but nothing found nothing yet.
It is just that I have built a compensation curve for raw measurements that was modeled after the IEC 711 coupler, but have also made a diffuse-field compensated curve based on 

Hammershøi's & Møller's averaged eardrum response curves from 2008; and now I don't want to create a third compensation file by hand as it would take some more hours to calculate and add it to the "IEC 711 raw" calibration file.​

 
As far as I know - no.  But it's easy enough to apply the values mathematically.  Apply your original curve (loopback etc).  File export (as a CSV).  Open in excel. Convert to 2 columns.  Add your new curve data to the old values.  Concatenate to get back to CSV again.  Save as a mic file.

Good idea, Paul. Haven't considered good old Excel for this but wanted to do it by hand based on drawing both curves. I'll do that when I find time to.
 
 

Chris - would you be willing to share your compensation curve values for the IEC 711 standard compensation?  I would be very interested!

Sure, but my compensation curve isn't really finished yet and it is only rough. I tried to kind of model my curve based on the raw data from innerfindelity, goldenear, rin choi and headroom with the IEMs of my collection that are in their database.


Just a rough "quick and dirty" compensation for raw data (with the linear output from the GOIEM100):
 

20 +0
40 +0
70 +0
1000 -2.5
2000 -8
3000 -10
3600 -11
4000 -11
5000 -12
6000 -10
7000 -14
8000 -20
9000 -26
10000 -28
15000 -7
17000 -12
20000 -10 

 
Close enough for the beginning, though at least mids and upper treble need some refinement.

Doesn't look too bad with the UE900 (orange), but my UERM (green) doesn't have such an extreme peak (although it has got a steep peak for sure):




The same goes for the Ivery IS-1 by the way.

Could be that the insertion depth is not enough or my raw calibration in the upper treble is still quite off. Gotta try some more on the weekend. If you/anybody else finds out what it is, please let me/us know. 

By the way: is there a way to align the graphs at a certain frequency (usually 1 kHz) in Arta?

 

[Brooko]

   
By the way: is there a way to align the graphs at a certain frequency (usually 1 kHz) in Arta?

 
Yep  - in the frequency screen, go Edit > Scale Level.  Then enter a value.  Mine usually measure at around 140 dB, so I scale everything back -70 dB. Then for the second graph, drop it back by 70 as well, and then if you're still a few dB off, simply repeat with a fine tuning value.  It moves the whole freq graph.

 

[castleofargh]

wazaaaaaaahhhh.

thanks hifichris for mentioning this topic to me.
 
my question, why ARTA?
my own attraction to REW has to do with the fact that I'm a EQ junky and REW has some nice EQ simulation feature. but even then, I tend to find the UI of REW more convenient overall.
 do you guys have trouble with REW? or is it simply that ARTA was suggested on the veritas website?
 
I'm really intrigued because I feel like I must be missing something obvious for everybody else to go for ARTA.

 

[HiFiChris]

  wazaaaaaaahhhh.

thanks hifichris for mentioning this topic to me.

Btw, I think I know what I'm going to do tomorrow to get a curve that (hopefully) matches the raw data of other sites: measure my ER-4S (without pseudo-calibration applied), build the diffuse-field compensated curve based on the graph and then build the raw graph based on that one (I might need your guys' assistance for the latter as my Excel skills are not the best).


As for the Arta question:
For my part, I can say that I use Arta because I find it easier to operate and navigate and prefer the UI.

 

[castleofargh]

well it's pretty easy(though not fast or fun) to record an online curve by frequencies, do the same with whatever you record with the veritas and use amazing math skills like + and - to get the curve you need to apply.
the 2 main problems with that method are:
1/ there is almost no hope that the point of resonance in the veritas will be the exact same as with the gear used for the online measurement(those stuff wouldn't cost so much if it was easy to mimic). so that might end up with 2 places having the wrong compensation as soon as we use another IEM.
2/ it's pretty much guaranteed that 2 pairs of the same IEM model will have a few DBs of variation and so will the one you use vs the one measured online. for that, the ER4 is clearly a good idea because the accuracy is known to be pretty high, an even nicer thing is that some ER4(PT I think) came with a measurement of the very pair \o/. but I don't remember how precise was the db axis on the graph.
 
 
to everybody, when the default tip doesn't allow to go deep enough, what's the best course of action? use whatever depth available, or use another tip? both can have very significant changes depending on the type of tech. I tend to go to the right depth with even another brand of tips if necessary(I at least stick to silicone), but how relevant is such a measure?

 

[HiFiChris]

Many tired evenings, sleepless nights and many, many hours later, I now get something that imho looks quite good and tendentially comparable to IEC 711 measurements with applied diffuse-field target curve.
Though I thought castleofargh's compensation/pseudo-711-calibration files would help me, I found myself to have to make an entirely new one, as the areas where I tried to adjust his results to match my mic's response ended up in a huge mess - that's why I had to continue working on the calibration/compensation I began before, and now I think that my results are quite good and don't require (much) more adjustment (at all, please also keep in mind when you compare different graphs from various sources, that they may have a different scaling).

Just four of them, all made with the included medium silicone tips:
 

---

Shure SE846 with white filters:                                            Westone W4R:
 
 ​

---

 ​

DUNU DN-2000J:                                                               Logitech UE900(S):​

 
 ​

---

 

[Cotnijoe]

Did you create your own compensations for it?
 
It looks to me like your veritas accentuates the bass like mine does 

 

[HiFiChris]

Yes, I did.

I find them actually quite comparable and they fit more or less between Tyll's and Rin Choi's results. Good for a general overview and comparison, but it will never be a real 711 coupler. 

 

[shotgunshane]

Did you create your own compensations for it?

It looks to me like your veritas accentuates the bass like mine does 

His scale is showing 5db increments instead of 10, so it appears more accentuated than most graphs. I like the 5db increments more, as it's easier to read and see the impact. It's a shame calibration files weren't supplied like originally intended.

Anyway good work HiFIChris. Maybe you guys can go into a little detail on how you calibrated step by step. I've struggled with mine. I think everyone using a Veritas needs to do this, otherwise the graphs posted are very misleading to the uneducated reader who assumes it's accuracy.

 

[HiFiChris]

Anyway good work HiFiChris. Maybe you guys can go into a little detail on how you calibrated step by step. I've struggled with mine. I think everyone using a Veritas needs to do this, otherwise the graphs posted are very misleading to the uneducated reader who assumes it's accuracy.

It isn't perfect, but closest to where I could get at the moment.

The way I did it was quite "simple", but it took up a lot of time. Since I got the Veritas, I spent almost every evening trying to get closer and closer to the diffuse-field IEC 711-like response.
 
The first thing one needs are a couple of IEMs that are stored in several online databases (i.e. Rin Choi, Golden Ears, Innerfidelity and Headroom). I measured my IEMs' responses, stored them and took a look at the raw measurements on those sites, then I tried to get as close to what I considered as the "golden mean", as each IEM and side doesn't measure identically, which is just logical (for example, I know that many IE 800s had some imbalance in the treble, whereas mine has about zero mismatch from the mids to the upper treble).
From that, I built something that was kind of a replication of the IEC 711's raw data.

Before my Veritas even arrived (it spent almost 2 weeks at the customs clearance), I built a diffuse-field compensation file after Hammershøi & Møller's eardrum response curve from 2008 and applied it using Excel and some maths.
Using the applied DF-compensated, pseudo-IEC-711-calibrated compensation file, I measured even more of my IEMs that were stored in at least one database and tweaked the compensation file until I got what you can see now.
That was a hell of work and plenty of trial and error, but my results are what I'd consider as quite good at the moment.

I could share my compensation file, but as castleofargh also said to me, it will most likely not help you, as each Veritas might have slightly different response curves. I guess when trying to build the pseudo-raw file, I started from scratch for at least three times by starting all over again.

 
Using this (my current "calibration" + compensation), ...
 

  0    +0
100    +0.8
200    +0.8
300    +1.1
400    +1.8
500    +2.4
600    +2.75
700    +2.4
800    +2.2
900    +1.4
1000    +0.7
 
1500    -0.3
1600    +0.1
 
1700    +0.6
1800    +1
 
2000    +1.7
2650    +5.5
3000    +5.8
3250    +4.76
3600    +1.946
4000    +1.4
5000    +2.0
6000    -0.95
7000    -6.55
8000    -16.05
8500    -19.025
9000    -16
10000    -13.25
15000    -4.47
17000    -17.16
18000    -14
20000    -10

... I get that:



Green is the uncalibrated, uncompensated response of my Veritas, orange is the pseudo-IEC-711-calibrated, diffuse-field compensated response of the same IEM (Sennheiser IE 800).

 

[Cotnijoe]

Thanks for the info hifichris. Thats actually super helpful info. Ive been meaning to calibrate my own but i havent gotten the chance to

 

[castleofargh]

  Many tired evenings, sleepless nights and many, many hours later, I now get something that imho looks quite good and tendentially comparable to IEC 711 measurements with applied diffuse-field target curve.
Though I thought castleofargh's compensation/pseudo-711-calibration files would help me, I found myself to have to make an entirely new one, as the areas where I tried to adjust his results to match my mic's response ended up in a huge mess - that's why I had to continue working on the calibration/compensation I began before, and now I think that my results are quite good and don't require (much) more adjustment (at all, please also keep in mind when you compare different graphs from various sources, that they may have a different scaling).

Just four of them, all made with the included medium silicone tips:

---

told you, nobody had any actual use with my compensation, and I made sure to give the simple version with few points so that people could alter it without going mad in a 0.4hz increment. but in the end it's just more practical to go from scratch
 
@ shotgunshane
 
how I made at least the starting point of my compensation curve:
 
I took an IEM, calibrated the software with it and the veritas in loop(automatic calibration), so that I ended up with a flat FR.
then I went to find the online measurement. I mostly stuck with headroom because you can change the vertical scale, but innerfidelity is ok too once you take 1step up on the graphs as 0.2db I believe.  I just noted some values, put a fancy minus sign in front of the values in db, then used amazing math skill in excell so that the smaller values would be -0.1(so adding the same number to all values). and that's about it. why -0.1 and not 0, because REW seems to ignore the point when it has 0 or positive values. dunno for ARTA.
if your IEM was exactly identical to the one measured online and you're sure about the values you've taken from the online graph, then you're done.
 
I wasn't so lucky.
I did that for a bunch of IEMs, applied those compensations to my IEM with the "all in the loop calibration", kept only those that seemed relevant and created an average(I don't know what parts of this you can do in ARTA, but I'm guessing most).
then I exported that average FR as .txt so that I would have plenty of data points, and went back again in excel to turn it into all negative values close to 0 as to limit the db variations from reality. I use that alone,  I have no calibration at all but that file, it takes everything into account as it was made with flat FR using everything in the loop from the start, so it includes amp and ADC correction(at least for FR).
 
I did the same for several IEMs because sadly several IEMs of the same model don't measure exactly the same. and then I had like "calibration for se215", "calibration for hf5", calibration for RE400" ... and I made another average of that. 
 
TBH the result wasn't too far away from what I obtained doing only the very first part with 1 IEM(HF5). at least the 200hz-7khz was virtually identical.
 
of course I believe one of the most important points is to determine an insertion depth and stick to it with all IEMs(when you can). the video on veritas mentioned about 15mm deep, I couldn't get 95% of my IEM down there with the default tip, for me the cone shape stops somewhere in the 11 or 12mm I would say(I don't have anything fancy to measure with precision, I used tweezers to determine a point and mark it on the outside). I align the IEM with the mark outside, put a super small piece of blue putty where the top of the veritas should be, and try my best to be around there. it must look pretty half assed compared to "rambo statistical averager" when doing the compensation curve, but I haven't thought of a better way(well I did but I don't have the tools).
 
 
not sure that will help anyone either but just in case. ^_^

 

[HiFiChris]

 
 
I took an IEM, calibrated the software with it and the veritas in loop(automatic calibration), so that I ended up with a flat FR.
then I went to find the online measurement. I mostly stuck with headroom because you can change the vertical scale, but innerfidelity is ok too once you take 1step up on the graphs as 0.2db I believe.  I just noted some values, put a fancy minus sign in front of the values in db, then used amazing math skill in excell so that the smaller values would be -0.1(so adding the same number to all values). and that's about it. why -0.1 and not 0, because REW seems to ignore the point when it has 0 or positive values. dunno for ARTA.
if your IEM was exactly identical to the one measured online and you're sure about the values you've taken from the online graph, then you're done.

Does anybody know whether the same procedure is also possible in Arta?