And both Rtings and Oratory's data show a broadband (more audible) increase between around 600 Hz and 2 kHz, near where our hearing is most sensitive.
Indeed, it’s the only point of commonality they have between them, if you think this is a good idea to directly compare them (it isn’t). Even then it’s quite inconsistent. But you don’t need to invoke the occurence of a firmware update to explain these differences, pads breaking in or coupling will do nicely given the dates of publication, methods, range and magnitude involved.
Ex here, APM with pad compression on my own head :
But I’ll reiterate that I would not expect that behaviour to be exactly the same on someone else’s head or on a fixture.
Rtings' HATS starts to become less reliable above this in the mid to upper treble, but still shows a general increase in energy here as Oratory's does.
Since your proposition is that Rtings’ data is evidence that a firmware update caused the differences they observed :
Yep places like Rtings have measured differences after firmware updates.
You’ll first and foremost have to contend with these issues :
- Why is the Sanity check trace different from the others ? We don’t know with which firmware it was measured, but given the date of publication it’s one of the earlier three already there.
- Why are all of these traces inconsistent ? Are Apple’s engineers a bunch of lunatics who can’t decide on a direction to follow ?
- Why, given that the medium to high-Q features at higher frequencies on the APM are not static features but vary with coupling, would they modulate the FR with medium to high-Q filters ?
I'm including for good measure a later trace they measured in December 2022.
I’m also going to object to the idea that some of these differences would necessarily result in what I’d term a perceptually brighter pair of headphones (it isn’t like the differences are only a increase in SPL relative to firmware 3B71), but hey, this is just my subjective impressions, which you shouldn’t take seriously.
Now you can continue to believe in whatever you want, personally I’ll just stick for now to the rather more reasonable proposition, in light of Rtings’ methodology and this :
That it's just more plausible that we're dealing with here with averages to averages differences (and I'm presuming here that Rtings performed averages for the latter firmwares, but we don't know), and that if any difference arose from a firmware update, it’s buried deep under the noise from the former phenomenon.
As Oratory's pdfs are an average, it's no surprise the later one would show less of an increase over the earlier one here than Rtings,
It’s a stupid idea to directly compare them given their different origin, but since you seem to need traces to be directly in the same graph to properly read them :
I’ll let others decide whether they consider the difference Oratory observed to correspond to “less of an increase”, let alone, outside of the 1-3kHz range, an overall increase in the first place. As far as I’m concerned I’d rather go for “an inconsistent result” first and foremost.
I’ll also let them decide whether or not they think this is all the sign of a deliberate change from Apple’s, but I think that anyone with some sense would be rather bothered with the inconsistency rather than conclude anything from it.
but the fact it still shows an increase means the later measurement in isolation must have had even more energy in this region than the second pdf shows, as this will have been brought down by averaging with the earlier measurements from the first pdf.
And the effect on perceived tonal balance is evident in the slope of the error curve (top dotted line) in Oratory's second pdf being less steep (indicating brighter tonality) than the first, and the the predicted preference rating increasing to 62% from 56%. Again the differences here, in slope and rating, would be larger between the earlier and later measurements in isolation, as the second pdf is an average that includes the earlier measurement.
The slope and preference ratings in Harman’s model aren’t just influenced by whether a pair of headphones actually is "brighter" or "warmer" than another, but also simply by the smoothness of the error curve. You can see an example here with the Sony MDR-MV1, using the same data from Aregina, progressively smoothed :
What you see in terms if slope and preference rating in regards to Oratory’s pdf is mostly related to that phenomenon (ie the February 2022 trace being smoother to begin with) and not an evidence that the later AirPods Max are indeed sloped brighter than earlier models :
The February 2022 pdf from Oratory isn’t necessarily smoother in the treble because later APMs are smoother in that region (they’re not), but just because more of this happened :
(given that the average perfectly overlaps with the December 2020 publication, I think that only one sample was contained in it, but I could be wrong here).
In other words had more samples (including with broken in pads) and individual traces been averaged as early as December 2020, the curve would have been smoother, and would have had a slightly brighter slope number as a result.
I'd like to make it very clear given the caliber of discussion on that subject that I'm likely to have here on ASR from some members that I do not show these examples as a way to "massage" the data, advocate using increased smoothing, or consider that the 1/3 smoothed traces are this time evidence that no change happened, far from it, but as a an
illustration of how the factors that make up the preference score can be influenced by how smooth the curve is to begin with, and that you can't really conclude from the difference in slope and preference ratings seen here that latter APMs are brighter than earlier ones.
You’ll find several pre-filled spreadsheets on the web to actually experiment on your own with that model, but I don’t think that you’re actually interested in learning anything anyway, since you didn't even take up my proposition a while ago to assist you to gain access to Harman's articles (that offer still stands).
All of this matches with yours and multiple others' perceptions of the difference in the APM over time, becoming brighter and with a better overall tonal balance (and so higher preference).
I know that your reading is conveniently selective but that’s not really exactly what they all wrote. In fact the respective experiences of
@acbarn,
@oleg87 and
@barreleye are quite a bit different between each others,
@barreleye in particular. I'll also note that the difference they observed was described as "dramatic" for example, while I'd put the differences between Rtings' and Oratory's traces in the "audible" category, I wouldn't call them "dramatic", but that's just me.
I’m also a bit annoyed to have been left out of the party as this is how my current pair differed when I received it in November 2022, and how it measured more recently in may 2023 (same pair, same pads - already broken in) :
Given the uncertainty related to averages to averages variation, and in-ear mic positioning variation, I can’t conclude from this that any change happened out of a firmware update for me during that time.
Again, I don't want to deny someone else's experience, and it would be interesting if we could get to the bottom of it, but as far as a firmware update deliberately causing a change in signature is concerned, I'll remain quite skeptical until proven otherwise. I can only speculate on the potential causes for their experience. Could some specific units have been temporarily affected by a series of firmware updates (
@barreleye) ? Could 2023 units, which to my knowledge we've yet to see a measurement of, truly be different ? I don't know.