Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: DAB mp2 vs AAC+ ("DABv2") (Read 6575 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

DAB mp2 vs AAC+ ("DABv2")

I am just listening to the Nero AAC(+) encoder. Remarkeable results even at 20kbps! I am comparing those to twoLame mp2 files (regarding discussions on DAB current and future standards).

Are my parameters valid?
It seems that twolame is not offering the "DAB" input parameter, but that cbr is on default and rates up to 192 would be representative of DAB quality?
mp2:
twolame.exe -m stereo -b 128 E:\Temp\mp2testing\Bendiksen.wav E:\Temp\mp2testing\Bendiksen_st_128

AAC:
Using Neros AAC encoder (SSE optimised version) with the "streaming" (=CBR?) option as that would simulate radio transmission. At rates below 40kbps, the encoder switches to "HE-AAC v2 High Efficiency AAC version 2 ", which is not supported by DABv2 or AAC+ as far as I can see so I should probably force it to the older standard?
neroAacEnc_SSE2.exe -cbr 20000 -if E:\Temp\AACptesting\Bendiksen.wav -of E:\Temp\AACptesting\Bendiksen_20.mp4

http://www.digitalradiotech.co.uk/nero_mpeg4_aac_encoder.htm

For the record:
LC-AAC = Low complexity AAC = apples AAC?
AAC = regular (main profile) AAC?
HE-AAC v1 = AAC+ = AAC w/SBR = DABv2?
HE-AAC v2 = HE-AAC v1 w/PS?



I am hoping to get some ABX testing up and running but doing pre-testing with purely subjective means and semi-consious listening;-) It seems to me that separate stereo mp2 hits a quality breaking point near 128kbps, while AAC does the same around 20kbps??

regards
k

DAB mp2 vs AAC+ ("DABv2")

Reply #1
I have listened only to a jazz/fusion recording of Norwegian saxophone player Bendik Hofseth at a number of rates and settings. It features nice dynamics lots of percussion/picked and delayed electric guitar and an open mix.

The most revealing artefacts of the two codecs when the bitrate was lowered:
For AAC there was some kind of aliasing (saxophone), "phasing" on percussion, a slight "gating" effect at times (pre-ringing or something?).

For mp2 it was a "swooshing filtered sound" reminiscent of a stepped wah-wah :-)
The saxophone sounded very distorted for mp2 almost like a filtered noise-source. Progressive HF-loss.

-k

DAB mp2 vs AAC+ ("DABv2")

Reply #2
Remember that mp2 DAB is limited to 24kHz or 48kHz sampling. If you encode a 44.1kHz track from a CD without resampling, this is not indicative of DAB output.

The dynamic text of DAB detracts from the available bitrate, but I thought toolame could handle/simulate this? (Never tried it - just saw the switches in there).

In the UK, many of the 128kbps mp2 DAB stations are audibly transcoded, so it sounds even worse than you would expect. We only have a few stations at higher bitrates, but many at even lower bitrates.

I think the technical description for the situation in the UK is a f*ck up.

Cheers,
David.

DAB mp2 vs AAC+ ("DABv2")

Reply #3
Remember that mp2 DAB is limited to 24kHz or 48kHz sampling. If you encode a 44.1kHz track from a CD without resampling, this is not indicative of DAB output.

The dynamic text of DAB detracts from the available bitrate, but I thought toolame could handle/simulate this? (Never tried it - just saw the switches in there).

In the UK, many of the 128kbps mp2 DAB stations are audibly transcoded, so it sounds even worse than you would expect. We only have a few stations at higher bitrates, but many at even lower bitrates.

I think the technical description for the situation in the UK is a f*ck up.

Cheers,
David.

I wasnt aware of that. But I am guessing that a properly conducted resampling of a lossless source would lead to neglible quality loss compared to the losses seen in an encoder at realistic bitrates.

I dont know how the mp2 efficiency is affected by working at 48 vs 44.1, but if the psy model is working properly, it should not allocate any bits to an area that contains no ekstra information (22.05 to 24kHz), and the top octave is probably receiving few bits in the first place?

But I will try to simulate those as well.

regards
Knut

DAB mp2 vs AAC+ ("DABv2")

Reply #4
Remember that mp2 DAB is limited to 24kHz or 48kHz sampling. If you encode a 44.1kHz track from a CD without resampling, this is not indicative of DAB output.

The dynamic text of DAB detracts from the available bitrate, but I thought toolame could handle/simulate this? (Never tried it - just saw the switches in there).

In the UK, many of the 128kbps mp2 DAB stations are audibly transcoded, so it sounds even worse than you would expect. We only have a few stations at higher bitrates, but many at even lower bitrates.

I think the technical description for the situation in the UK is a f*ck up.

Cheers,
David.


Mp2 sounds horrific at 128k stereo and still bad with I/S. To my ears it cannot / should not be used below 192k.
wavpack hybrid 256k -hx4

DAB mp2 vs AAC+ ("DABv2")

Reply #5
I wasnt aware of that. But I am guessing that a properly conducted resampling of a lossless source would lead to neglible quality loss compared to the losses seen in an encoder at realistic bitrates.

I dont know how the mp2 efficiency is affected by working at 48 vs 44.1, but if the psy model is working properly, it should not allocate any bits to an area that contains no ekstra information (22.05 to 24kHz), and the top octave is probably receiving few bits in the first place?


The resampling part is nearly lossless. (Though in practice many radio stations do it in the analogue domain, sometimes more than once!).

However, you can't assume that a psychoacoustic codec works equally well irrespective of the sample rate. Changing the same rate can have a huge effect.

Lame, tooLame, FhG mp3 etc etc etc switch to 32kHz, then 22.05kHz, then... etc as you decrease the bitrate for good reason. If there was no issue, they could just stay at 44.1kHz and lower the low pass filter frequency. If you think that _should_ work, over ride the default behaviour and listen to the result!

With fixed block sizes, at higher sample rates, you have more blocks per second, and each one gives poorer absolute frequency resolution. You could make a slightly better trade with variable block sizes, but most codecs support two at most!

Even without that limit, you would still have more data coming in, which would mean you would have to throw more away - you'd need a perfect psy model for there to be no overall effect.

mp2 doesn't have infinitely variable block sizes or a perfect psy model, so sample rate is an issue. The performance hit could be even worse for encoders that are tuned at 44.1kHz but not 48kHz!

Cheers,
David.


Mp2 sounds horrific at 128k stereo and still bad with I/S. To my ears it cannot / should not be used below 192k.


Read it and weep...

http://www.digitalradiotech.co.uk/dab/digi...o_bit_rates.htm

(from http://www.digitalradiotech.co.uk/index.htm - a great site for comment on digital radio)

or

http://www.wohnort.demon.co.uk/DAB/uknat.html
http://www.wohnort.demon.co.uk/DAB/ukloc.html

Cheers,
David.

DAB mp2 vs AAC+ ("DABv2")

Reply #6
[Mp2 sounds horrific at 128k stereo and still bad with I/S. To my ears it cannot / should not be used below 192k.

In my limited experiment I thought that 128 stereo was acceptable for twoLame -> winamp.

Is that due to our playback-chain/pickyness being different, or are you factoring in less ideal radio signal chains consisting of resampling, processing, recoding etc?

-k

However, you can't assume that a psychoacoustic codec works equally well irrespective of the sample rate. Changing the same rate can have a huge effect.

Cheers,
David.

Does this mean that the efficiency loss is significantly greater than 48:44.1?

-k

DAB mp2 vs AAC+ ("DABv2")

Reply #7
Does this mean that the efficiency loss is significantly greater than 48:44.1?


Not worse - it's typically about that.

Unless the psy model is better tuned for 44.1 than 48.

Cheers,
David.

 
SimplePortal 1.0.0 RC1 © 2008-2021