SonicStage, MD Simple Burner and Hardware encoder comparison

[Avrin]

Following this discussion: http://forums.minidisc.org/index.php?showtopic=15499 I decided to check the properties and operations of encoders available in the two programs and my RH-MZ10. Following are my findings, which are by no means scientific, and can contain some numeric errors. The track used for frequency analysis was again Paul Desmond's "Take Ten".

The main findings are that most encoder modes feature a cut-off frequency, which differs for different bitrates.

Now the results:

ATRAC3plus in SonicStage 3.4.01.13062:

Bitrate - Cut-off frequency

352 kbit/s - none

320 kbit/s - none

256 kbit/s - none

192 kbit/s - 18.0 kHz

160 kbit/s - 16.5 kHz

132 kbit/s (ATRAC3) - 17.5 kHz

128 kbit/s - 15.5 kHz

96 kbit/s - 15.5 kHz

64 kbit/s - 15.5 kHz

48 kbit/s 13.0 kHz

MD Simple Burner 2.0.05.25150

Bitrate - Cut-off frequency

HiSP - none

HiLP - 15.5 kHz

48 kbit/s - 13.0 kHz

Recording via optical in on MZ-RH10

Bitrate - Cut-off frequency

HiSP - 19.5 kHz

HiLP - 15.5 kHz

So it seems that SS, SB and hardware encoders are somewhat different, and recording on the unit in HiSP does not provide full frequency range, but is probably less prone to artifacts at higher frequencies. Also, in ATRAC3plus the frequencies, which are higher than the cutoff one are not completely empty, they contain some low-level noise and harmonics.

Now the MP3 results. The most interesting fact here is that the High/Normal drop-down box DOES have effect on encoding, changing the cut-off frequency for MP3s. I always suspected that this switch had nothing to do with the ripping process. Especially if we remember that the new SoundForge ATRAC codec also includes this switch. Its effect on ATRAC encoding is yet to be seen, although I think that it actually affects encoding quality, without changing the cut-off frequency though. All MP3 bitrates in SS encode in Stereo, except 96 kbit/s, which encodes in Joint Stereo.

Bitrate - Cut-off (High) - Cut-off (Normal)

320 kbit/s - none - none

256 kbit/s - 21.0 kHz - none

224 kbit/s - 20.8 kHz - none

192 kbit/s - 20.5 kHz - none

160 kbit/s - 20.0 kHz - 20.2 kHz

128 kbit/s - 15.5 kHz - 16.0 kHz

112 kbit/s - 13.7 kHz - 13.7 kHz

96 kbit/s - 13.2 kHz - 11.6 kHz

In MP3 there is almost nothing at frequencies higher than the cut-off one. Except for some occasional very low level noise at lower bitrates (128 and below).

[pata2001]

Very interesting findings. It would be great if we can also see the difference between High and normal for Atrac.

Excellent post.

[dex Otaku]

http://forums.minidisc.org/index.php?s=&sh...indpost&p=95858

http://forums.minidisc.org/index.php?s=&sh...indpost&p=95884

Nice work, otherwise. You are a welcome addition to the fora.

[Sparky191]

HiSP - 19.5 kHz on the HiMD unit is not good

[dex Otaku]

HiSP - 19.5 kHz on the HiMD unit is not good

I'd say it's a matter of priorities. The idea with location/live recording at a rate that is supposed to provide transparent encoding [i.e. no or as few as possible audible encoding artefacts] is to give priority to the most important [i.e. audible] bands, in the range between about 500Hz - 5kHz [the vocal range, where most musial instruments also present most of their energy].

Psychacoustic data reduction is intended specifically to take advantage of average characteristics of human hearing. It's well-established that a fair majority of people 25+ years old can't even hear 19.5kHz as it is, so the bandwidth that would have otherwise been consumed encoding something that most people aren't going to hear is devoted to the range that has the most energy in it.

In other words: setting a "decent compromise" cutoff for the highs also means setting the average amount of artefacting that occurs in the bands where the most energy usually is, and which are the most important to present with as little distortion [artefacting] as possible.

This is not to say that "all recording should only be done to 19.5kHz because people are deaf." This is to say that this is a consciously chosen compromise made to ensure a base-level of quality during realtime encoding.

[Sparky191]

Why does the compromise only exist on the HiMD unit?

[dex Otaku]

There are several possible reasons why, and they are not mutually-exclusive:

* a desire to ensure a base-standard for encoding quality from a live source

* the cutoff of the unit's ADC

* the limits of the unit's input preamp

My thought is that it's probably a combination of all three. Each section before encoding even takes place has its limits.

[Sparky191]

"base-level/standard"

Yahwhoozywhatsit? :~

[dex Otaku]

base standard = lowest common denominator. It's like saying, "for recording on units with gen1 of this encoder, this will be the sonic profile of every recorder." This includes everything from the quality of the ADC used to the quality of the analogue sections of the input to the quality of the encoder. Everything in the stream has its own effect on the final output.

This is why [iMO] every HiMD recorder that has come out thus far has exactly the same input section. Same ADCs, same preamps. The codec might change slightly between models released at different times [if they've continued working with it, it should hopefully improve with newer models, at ATRAC SP did with the original MD], but my guess is that this hasn't happened yet [HiMD is only 3 years old]. If such is the case, everything in the recording section of every HiMD model made thus far is identical, so recordings made on each models should only vary slightly based on the mass-production tolerances of the parts they use.

Base-line, or base standard. Lowest common denominator. There is only one quality.

[Sparky191]

But if you record in PCM on the portable, then encode to HiSP in SS you will not have this - 19.5 kHz with your HiSP files.

[dex Otaku]

Good point, but you still missed my point completely.

[Sparky191]

Actually I don't get it.

I think you are saying the cut off is there. So that there should be little variance between recordings made on different portable units. That there is a cutoff at all is because of something on the units themselves.

However I'm just surprised that there would be a cutoff at all, if there isn't for PCM on a portable. Whats so different about a HiMP in HiSP mode vs PCM?

[dex Otaku]

[banging head on desk]

[Sparky191]

Maybe I'm missing something Would

* the cutoff of the unit's ADC

* the limits of the unit's input preamp

Not effect PCM recording the same as ATRAC?