Re: Quite OK Audio (QOA)... anyone ?

Reply #75 – 2024-09-03 14:48:25

Now I've tried QOA with Melodi 6 (means "Melody 6") form Supper Zill (it's a very popular school bell program in Turkey, which is my country) and the result is awful. I'm started to think even 4bps won't be enough to get high quality with this sample, I think 5bps or 6bps can be needed. The funny fact is that the original version of this file is 8bps. (I think another ABX is not needed as I've already proved that QOA is not transparent.)

Re: Quite OK Audio (QOA)... anyone ?

Reply #76 – 2024-09-03 17:48:42

I studied the qoa code which is practically a sort of ADPCM with the choice of 16 types of quantizations. The decompression sound wave is also similar to ADPCM. Overall, a good job by the programmer who manages to obtain good results with psnr between 40 and 50 and above DB. In fact, I am delaying the release of ADC because everyone has complained that in some cases, especially at high quantization levels for engines based on data differences, this problem is encountered. I certainly don't want to be a professor but this phenomenon has been resolved by other codecs such as mp3, opus with a certain coding that limits such psychoacoustic inconsistencies. Many who criticize do not know that these codecs have a low psnr performance but a frequency cut and masking that deceive the ear. Happy with them  !.

Re: Quite OK Audio (QOA)... anyone ?

Reply #77 – 2024-09-04 12:44:31

And, I think I found the ideal values for 4bps, 5bps, and 6bps:

float dqt[16] = {0.25, -0.25, 0.75, -0.75, 1.5, -1.5, 2.5, -2.5, 4, -4, 6, -6, 8.5, -8.5, 11.5, -11.5};

float dqt[32] = {0.2, -0.2, 0.6, -0.6, 1.2, -1.2, 2, -2, 3, -3, 4, -4, 5, -5, 6, -6, 7, -7, 8, -8, 9, -9, 10, -10, 11, -11, 12, -12, 13, -13, 14, -14};

float dqt[64] = {0.1, -0.1, 0.3, -0.3, 0.6, -0.6, 1, -1, 1.5, -1.5, 2, -2, 2.5, -2.5, 3, -3, 3.5, -3.5, 4, -4, 4.5, -4.5, 5, -5, 5.5, -5.5, 6, -6, 6.5, -6.5, 7, -7, 7.5, -7.5, 8, -8, 8.5, -8.5, 9, -9, 9.5, -9.5, 10, -10, 10.5, -10.5, 11, -11, 11.5, -11.5, 12, -12, 12.5, -12.5, 13, -13, 13.5, -13.5, 14, -14, 14.5, -14.5, 15, -15};

Of course, the subarray of the lowest scalefactor should be edited for better efficiency in 5bps and 6bps. Can anyone update the code with these values? (I know C but this code is a little bit too complex for me. Also, I'd like to learn how to change the adaptation speed without causing bugs in the audio (I've tried it and caused bugs in the audio).)

Re: Quite OK Audio (QOA)... anyone ?

Reply #78 – 2024-09-13 12:31:07

By the way, the exact bit depths will be 1.066, 2.133, 3.2, 4.266, 5.333, and 6.4 bits per sample.

Re: Quite OK Audio (QOA)... anyone ?

Reply #79 – 2024-09-15 00:07:55

Also, I think using 32 scalefactors with a dequant tab of [1,-1] could be good for 1bps, which will result in 59 samples per slice and 1.085 bits per sample: scalefactor_tab <- round(pow(s + 1, 2.5)). Besides that, I think [0.75,-0.75,3,-3] could be better than [0.75,-0.75,3.5,-3.5] for 2bps.