All,
I'd like to share some
ES100 data measured using Audio Precision 2722 right next to my desk.
[Test condition]
- ES100 3.5mm output
- No load
- EQ, DCT off
- AK4375a / sharp roll-off / 1x
aptX (with Galaxy S8 source device)
It's just a lossy waveform codec using no psychoacoustic model.
It uses QMF(Quadrature mirror filter) to split the input into a certain number of sub bands.
And then it encodes each sub band with the limited bit allocation;
more bits in the lower frequency band, less bit in the higher frequency band.
SBC(SubBand Codec) also split the input into sub bands, but it allocates the same number of bits per for all sub bands.
That's the one of the main difference between SBC and aptX.
SBC is a very simple codec with low complexity.
The low complexity nature was the primary reason for having SBC as mandate Bluetooth A2DP codec.
And CSR released the new codec aptX with the key idea of allocating the different number of bits per each subband;
more bits in the lower frequency band, less bit in the higher frequency band.
So typical frequency response of a single tone would be as above.
The noise floor is not flat across frequency ranges because the quantization noise from aptX codec loss makes the floor.
aptX-HD performs better but the frequency response is similar to aptX classic.
AAC (with iOS)
It's a complicated codec based on the psychoacoustic model.
As you see below, the 1KHz and 10KHz tone is not sharp but shaped like a tower or triangle.
The key idea of the psychoacoustic model is that;
the signal or noise below the shape(i.e., Masking Curve) is not audible and not critical.
Eventually we just need to allocate as less bit as possible for the target signal below the masking curve.
So, the codec like AAC using the psychoacoustic model requires frequency analysis, the masking curve calculation for the time-varying input continuously,
which cost a lot of processing power and power consumption.
Since it's not a waveform coder, measuring THD+N for the signal through AAC codec has no meaning.
Only perceptual domain performance would make sense.
Anyway, AAC is more complicated than mp3, showing the better perceptual performance than mp3 at the same bit rate.
USB (lossless, with iMAC USB)
It shows the baseline performance of ES100 HW.
Unfortunately, ES100 USB supports 16-bit only, the overall THD+N is limited by the 16-bit PCM (-96dB), not by H/W.
Please take a look at them all above.
You will find there's no such a design flaw or HW defect as SubMash insist.
If ES100 have any H/W issue, all the BT codecs, and USB would result in the same defect.
Because all of them use the same H/W resource including DSP, DAC, and Power.
SubMash mentioned AAC & USB sounds OK.
Then it's not an H/W issue.
And if we have the problem with aptX,
how could we have ES100 in the Qualcomm's official aptX product list?
They have their own qualification and certification process for aptX enabled products.
We sent ES100 to them for their testing and finally had ours on the list.
https://www.aptx.com/products/radsone-earstudio-es100-bluetooth-receiver-headphone-amplifier
Lastly, I just took a look at the SubMash test results and graphs from his mobile phone screenshot.
It seems that he ran the test with his mobile phone.
Anyway, as I see them, those results are showing the tremendous and unacceptable noise.
If those graphs are indeed real and represent ES100 output,
the sound would be considerably worse than the expected,
and everyone would definitely notice the defect right at the start.
We can't deceive many thousands of ES100 users.
The users are not fools and would notice the that much of conspicuous defect very easily.
Thanks and Regards,
WS
