Search This Blog

Tuesday, April 9, 2013

Accudio™: Accurate Audio Player by Goldenears


Intro
Released in August 2012, Accudio™ is an iOS music player developed by Goldenears. In addition to its powerful 9-band parametric equalizer, this app also comes with brilliant acoustic enhancement features called "Reference/Simulation mode". As long as headphone measurement data are stored in the database consisting of frequency response measured from over 300 headphones, users can tune any of their headphones to match Goldenears' proprietary diffuse-field target, or simulate them into nine different kind of headphones from various brands.


As many of inexpensive portable headphones lack in quality, such features shall definitely improve the user experience by a great degree. However, the legitimacy of Goldenears' claims is yet to be known. By analyzing Accudio™'s sound enhancement features in detail, their technological validity will be assessed.




Testing methodology
The version of Accudio™ used in this analysis is 1.1.1, and an iPhone is used to drive headphones. The frequency response deviation from the reference setup is less than 1dB from 20 Hz to 20 kHz.


An occluded ear simulator complying ITU-T Rec. P.57, IEC 60318-4, ANSI S.325 is used, with the measurement procedure conforming IEC 60268-7.


A parametric equalizer
Accudio™'s 9-band parametric equalizer is one of the equalizer apps of the highest quality available in the iTunes app store. This function that utilizes IIR allows users to set equalization parameters at any nine points within the frequency spectrum from 20 Hz to 20,000 Hz in -15 dB to 15 dB scale. The bandwidth can also be adjusted from 28.9 to 0.2 in Q factor, or 0.05 to 5.0 in octaves.


However, despite its high resolution, regrettably, a touch-to-set function, which is commonly implemented in other equalizer apps, is missing, leaving non-expert users out of the range of usage.


Above is the result of amplification/attenuation of 15 dB in Q of 28.9, at 20 Hz, 200 Hz, 2 kHz, and 20 kHz, and clearly indicate Accudio™'s equalizer operates exactly the way it is supposed to.


Moreover, even with an extreme filtering, the equalizer does the job very precisely. As the signal processing is able to handle highly detailed equalization target, this parametric equalizer is recommended for even quite demanding purposes.


A reference emulator
One of the notable features of Accudio™ is its "Reference mode". By calculating the difference that the headphone of choice has over the Goldenears' reference response, the app applies the differential gain to the user's headphone to closely match the designated target. Although technical justifiability of the target is highly disputable, it is no doubt the feature may be able to dramatically improve sound quality of some low-quality earbuds.


The reference setting can be adjusted into 21 levels in 5 categories, leaving the users with great control over what they'd like to hear. However, how Goldenears set the frequency and bandwidth of each parameters, which are conceptually vague to define, is highly questionable.


Nonetheless, Goldenears firmly believe the concept of their reference emulation does work, so Etymotic Research ER-4P, which is officially rated 4/5 for its equalization adaptability by Goldenears, has been used to assess the effect of the process.


And the result shows the sign of amplification below 100 Hz, and above 3 kHz, which is quite far from either flat or even their own target.


A headphone simulator
The idea of simulating a headphone using another set of headphones has been around ever since Voinier and Briolle (1992), but has never been able to be successfully commercialized due to lack of dsp technology in portable devices, as it involves with a heavy amount of calculation for accurate reproduction. One commercial implementation of such technique is BDNC's HMT(Headphone Modeling Technology), but their technology never made it to the market.


Currently there are total of nine simulation targets, which are indeed of very high quality. If a cheap earbud can be turned into any of these products, Accudio™'s "Simulation mode" will totally revolutionize the user experience in portable audio.


However, as seen above, a simulated ER-4S lacks a good amount of bass compared to the actual ER-4S, indicating the simulation model is far off in this case.


Compared to the actual Sennheiser HD600 & HD650, simulated models are overly emphasized in bass, not to mention the characteristic deviation between HD600 and HD650 is nowhere to be found. Thus, even with Goldenears certification in adaptability, it is clear that accuracy of the simulated result is not guaranteed at all.


Goldenears certified 
As it is possible that the headphones with higher adaptability might perform better, their certified headphones, T-PEOS H-100 and LG HSS-F450 Quadbeat have been analyzed.


Unlike what the official data suggests, the problematic dip @ 8 kHz and the peak @ 10 kHz have not been neutralized with "Reference Mode" setting. The high frequency bandwidth does not improve either.

Again, nothing is too different from the result of ER-4P, strongly suggesting the sign of inaccuracy.


Conclusion

Upon analyzing Accudio™'s acoustic enhancement features in great detail, the developer's claims have been fully assessed. While the parametric equalizer is indeed of a professional quality, it turns out "Reference/Simulation mode" highly lack in accuracy. It is especially disappointing, that the developer has not implemented a high resolution filtering process, which is critical for accurate simulation, in "Simulation mode". One way of doing so is to fully utilize the time-domain characteristics of their measurement data, which requires a lot of processing power. Maybe the current mobile technology may be still too far from achieving the means of true simulation.

All in all, the idea furnished in designing Accudio™ is innovative and creative, yet the developer's execution has proven to be quite poor. And with the price tag of $4.99, which is rather expensive for a music player app, it is doubtful that the app could give a strong impression to the potential buyers.


References

T. Voinle and B. Francoise, “Transfer Function and Subjective Quality of Headphones: Part 1, Transfer Function Measurements,” presented at the 11th International AES Conference, Portland, Oregon (May 29-31, 1992).

14 comments:

  1. I'd pay $4.99 just for a 9 band parametric equalizer on my android phone...

    ReplyDelete
    Replies
    1. The EQ is very nice. I hope they release the equalizer-only version with less price LOL

      Delete
  2. well. interesting but your detailed but clear analysis contradicts your rather harsh conclusion. audiophile people spend hundreds of dollar for cables, just for "feeling something different" from their headphones. comparing to that, this app is very solid investment. i may get one if i had god-forsaken i products.

    ReplyDelete
    Replies
    1. When something that was guaranteed turns out to be the exact opposite, the product is inevitable to meet with harsh criticism; it is same with Accudio as my money was not invested to "feel something different".

      If this happens in retail industry, it is indeed a false advertisement, and customers would ask for something called "refund" or even a "lawsuit".

      Delete
    2. well, i got your point. and if somebody says the same point, that's understandable. but problem is, you not gonna win any lawsuit(refund is different matter though). because it is a "simulation", so a kind of imitation, not exact reproduction. nobody sue the company designed a flight simulation for not 100% accurately reflecting real flight. so it's very hard to say GE lied. you can say their simulation is not satisfying (but compare to what? considering their low price), but nonetheless like you said, their 'reference mode'is good. so at leat they satisfied 50% of your high exectation. and you said, in your outstanding review of sennheiser ie800, that ie800 is a masterpiece, even though a half of sennheiser's advertisement of ie800 is false. so why not give a fair share of compliment to Accuido(? weird name though). we can't say that 50% of lies is ok for 1000 dollars product but not for 6 dollars'!

      Delete
    3. It is a matter of what you value the most: Personally, I don't like being played by false claims by the manufacturer, just like any other customers. Price is only a part of the equation.

      Delete
  3. Thanks for review!

    ReplyDelete
  4. One thing I've learned working with EQ is to observe the real world distortion at the band of interest. If the bass response starts dropping sharply below the speaker's fundamental resonance, it is rarely ever a good idea to compensate it with EQ. You might be able to get a 6dB boost flattening out the sub-bass drop a bit, but higher than that distortion tends to rise sharply, and IMD will throw mud in the precious mid-range.

    I think EQ might be more useful in gobbling up some uncontrolled mid-hi frequency resonance if the filters can handle high Q notches well. Simply not having anything there tends to sound much more forgiving than a long ringing trail.

    ReplyDelete
    Replies
    1. According to Sound Reproduction, written by Toole of Harman International,

      "The right kind of equalization sounds just fine, and electronics can provide an option equivalent to natural acoustical manipulations."

      And according to Voinle and Francoise (1992),

      "One most important results of this study is the finding that the quality of a headphone of poor or fair quality can be considerably improved simply by simulating on it the acoustical characteristics of an excellent headphone. This can be done in real time, with 200 taps FIR filters at a sampling frequency of 44.1 kHz."

      With headphones, even with a slight displacement, various factors such as the ear canal geometry & acoustic leak caused by the amount of hair totally change the sound, not to mention IEMs' acoustic intereference from 6 kHz to 10 kHz range cause by insertion depth related resonance.

      My rule of thumb is that equalization won't improve the bandwidth & can't drive the headphone over its excursion limit. Unless there is a high resolution inverse-filtering, which can only be made by binaural scanning, available at disposable, a simple low-pass/high-pass filter should do the job along with low-Q (preferably less than 10) parametric equalization or 1/3oct equalization. Trying to tame small ripples/high-Q notches won't contribute to the overall sound quality at all.

      Delete
    2. For IEMs, given using the same tips with consistent insertion for the same person, my experience is that the resonance frequencies does not shift around much. So if you can determine those frequencies yourself, it would be nice to EQ them out using "high-Q notches"

      Delete
    3. My take is that if there is nothing to excite the ringing frequency, it might as well appear absent to the ear, hence the high Q notch filters at resonance freq. I sort of took this from treating not so optimal recordings where there is apparent resonance from room nodes/mic, bandpassed and a few notches in the mid-hi frequencies before mixing can do wonders. But I agree there is nothing that can be done with excursion limits, maybe some negative impedance drive voodoo to cancel out some DC resistance of the coils will help to a certain degree, but nothing groundbreaking.

      Delete
  5. This comment has been removed by the author.

    ReplyDelete
  6. I find your assessment interesting. First, I do agree it is one of the best eq's on the ipod, period. I love it. As for the reference mode. I literally bought the er4s based on the accudio simulation mode of the er4s via the westone 4r. I thought, the 4r sounded so good on hi fi with no adjustments and then compared the hifi to all the simulation modes. I found the er4s simulation mode was almost identical to the hifi setting, but with a little less sub bass. After buying the er4s, I found it to be exactly what it has represented. I then compared my pfe112, 4r, er4s, earpods, etc. all with hifi mode and simulation modes between each other. While they aren't identical sounding, they give a very good approximation of the differences in frequency alone. That should be understood. This is only simulating frequency response.
    In fact, I've replicated every "hifi" setting of these earphones using the custom eq settings and switching between hifi and custom until I have is exact. I literally get the eq to the point where I can't tell the difference. Once this is done, my eq looks almost exactly like the graph compensation they show in their previews. When I take that exact eq setting to my computer eq application, the results are identical.
    Also, regarding the hd600 comparison, this i found very odd. I don't have the er4p, but I compared my hd600 to my er4s using the hd600 simulation. It wasn't identical, but again, extremely close, and all the main "difference" in the sound were present. When I switch the er4s simulation to the hd650 I hear the high end reduction and the warmer tone. Your graph of the differences doesn't even appear to show any high frequency reduction. I don't see any difference, while my simulation showed a clear (although small) difference. I've noticed accudio sometimes lags and/or doesn't switch settings every time until you switch back and forth to refresh it. Is it possible you measured the same setting with that simulation? Just curious.
    I'm sure they aren't perfect simulations, but again, I bought the er4s based on the simulation and was very happy to get the sound I expected. Putting my er4s on pfe112 simulation also shows the signature treble imbalance. Any idea why we would be getting such different results? The biggest improvement I hear is with my er4s on reference hifi mode. It brings the bass up almost exactly the amount I see it lacking on graphs. That is about 6-7db at 20-30hz, up to about 5db at 50hz. and in a smooth gradual curve the levels off no higher than 90-100hz. I matched the setting with custom eq and got those values and it sound excellent. It even opens the treble up a tad making it more transparent ever so slightly... :-o

    ReplyDelete