Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Herbeck]

Moderator Note:
During WUMP 25 the subject of digitally reducing gain to avoid clipping the DAC came up.

Here is that discussion:

Converted track to 32bit Float and used a look ahead limiter with very subtle settings to deal with the digital overs. I'd be very interested to hear how others dealt with this. My main concern was clipping the Crookwood Transfer DAC.

I reduced the song by 6 dB (Bit transparent) -6.020599913279624 dB.
Samplitude doesn't seem to like that many digits. But I know Reaper does both in the mixer fader and even in the built in plugins such as the fader on the EQ. You can use Bitter VST and see if you can find a fader plugin that works in Samplitude. The song should be 16 bits even after the -6.020599913279624 dB reduction.

http://www.stillwellaudio.com/?page_id=33

Herbeck

[djwaudio]

I reduced the song by 6 dB (Bit transparent) -6.020599913279624 dB.

That's really interesting, I use the TC 6000 this way.  In some algos, it shows a green light when the bit pattern is retained.  I hadn't realized that some DAWs had some capacity for this.

[Herbeck]

That's really interesting, I use the TC 6000 this way.  In some algos, it shows a green light when the bit pattern is retained.  I hadn't realized that some DAWs had some capacity for this.

Most gain faders would work if they allowed enough digits. I suspect that the faders on the free ReaPlugs VST will work fine with the "-6.020599913279624 dB thing" in any DAW.

http://www.reaper.fm/reaplugs/index.php


Herbeck

[Hermetech Mastering]

Thanks very much for addressing my question Herbeck. Sorry, I'm not too mathematically inclined here, but why should reducing it exactly -6.020599913279624 dB be bit transparent at 16 bit fixed, whereas, say reducing it by exactly -6.0 dB wouldn't be?

So are you saying that if you do it this way, then there's no need to convert it to 32bit float before applying the gain reduction (or limiting, as in my case)?

[pmx]

10*log(2^2) = 6,02... dB

1 bit in audio addresses that much information of the digital signal, so reducing a signal by 6 dB is not exactly the same as reducing the bit depth by 1 bit.

[Hermetech Mastering]

OK, I get that, but can it really be "bit transparent" if you are throwing exactly one bit away? Or is the part you are throwing away by reducing the gain way down in the noise floor and not worth concerning yourself with? Sorry, I'm probably just over-thinking this.

[Herbeck]

The phrase "bit transparent" is probably the wrong technical term, I'm not to good with words. But it sounds a little bit more transparent than "a normal gain reduction" and that is the cool thing about it.

Cheers,
Herbeck

[Hermetech Mastering]

Cool, next time I get a 16/44 track for mastering which already has clipping/overs (unfortunately seems to happen quite a lot), I'll try the -6.020599913279624 dB instead of my usual "convert to 32 bit float and use a lookahead limiter" approach.

I'm still not really sure how or why -6.020599913279624 dB would be more transparent than just -6.0 dB, something to do with the maths? If you say it sounds more transparent I'll take your word for it and do a couple of tests myself.

[pmx]

i have a hunch that truncation distortion plays a role here. is dave collins on this forum? he knows for sure...

[Herbeck]

i have a hunch that truncation distortion plays a role here. is dave collins on this forum? he knows for sure...

He's been known to reduce the feed to the DAC by 6dB for some sources.

[djwaudio]

What is happening is, you are retaining the "bit pattern", moving it to the next least significant bit. So, no it's not technically bit transparent, but closer than a typical gain change.

It doesnt work with just 6.0dB, it seems to require the decimals in my test with Reaper.

The whole point is to be able to adjust the level to feed the analog chain without an extra stage of truncation or dither. 

[Herbeck]

The whole point is to be able to adjust the level to feed the analog chain without an extra stage of truncation or dither.

This is a very good point. But I also prefer the sound of the -6.020599.....dB change over the -6 dB change, but that's just me and my opinion.

Herbeck

[Hermetech Mastering]

Thanks Dana and Herbeck for clearing that up, I will give it a try soon!

[djwaudio]

i have a hunch that truncation distortion plays a role here.

Well, you have to be careful, because you'll now have a 17 bit signal, according to my bit meter. 

[Herbeck]

Well, you have to be careful, because you'll now have a 17 bit signal, according to my bit meter.

Maybe this technique only works properly with a 16 bit signal, since a 24 bit signal reads as 25 bit after the gain change and would cause truncation in the DAC. Do you have the means to compare with the TC 6000?