|
|
Post by jin167 on Sept 4, 2017 21:03:00 GMT -6
interesting comments. make your own decision. I like this video, VERY interesting! My take is: We can make a choice of a simpler workflow (no dither) and create inaudible truncation distortion artifacts. Or we can choose a slightly more complicated workflow (insert an extra plugin), and create inaudible dither noise. I had chosen the former, and after this thread I've decided to go for the latter now. Good on you. Like I said, form your own opinion on this matter but understanding how technical stuff works (and I assume you do since you found the video interesting) comes before making comments about the sound (especially for things like dither). How something sounds is a very subjective matter. Math? Not so much. Making comments about how something sounds without having a proper understanding of how that sound is produced is, in my opinion, a waste of time. |
|
|
|
Post by Bob Olhsson on Sept 5, 2017 11:01:32 GMT -6
The idea of not needing 24 bit dither has been widely promoted and people repeat what they've read or heard in a video. On my Duntech monitors I can almost always hear a difference and I've demonstrated this to others numerous times.
|
|
|
|
Post by mrholmes on Sept 10, 2017 19:12:52 GMT -6
I checked, and I have the Waves IDR360 and I think there's a dither option in the L3 and L3 mastering limiter. Does it matter which of those I use, any particular options?I think thee's an optional dither setting in Logic, but I've always bypassed it. Should I use that or the Waves? You can download a lot of free dither plug ins at airwindows |
|
|
|
Post by wiz on Sept 10, 2017 19:20:26 GMT -6
I checked, and I have the Waves IDR360 and I think there's a dither option in the L3 and L3 mastering limiter. Does it matter which of those I use, any particular options?I think thee's an optional dither setting in Logic, but I've always bypassed it. Should I use that or the Waves? You can download a lot of free dither plug ins at airwindows thats what I use cheers Wiz |
|
|
|
Post by lcr on Sept 11, 2017 4:06:43 GMT -6
I started my testing first trying Pro L, Downloaded the good dither trial per recommendation in this discussion and now Im using that. I will report back after further use. So far I like the results. Not sure how to describe the improvement, smoother, cleaner maybe?
|
|
|
|
Post by mrholmes on Sept 11, 2017 10:26:59 GMT -6
You can download a lot of free dither plug ins at airwindows thats what I use cheers Wiz Jupp its fair to give Chirs a few bucks a month via his Patreon. Becasue nearly all his plug ins are now free of charge. he hopes for more patreon payers... ans some of his plugs are genius strikes. |
|
|
|
Post by johneppstein on Sept 11, 2017 21:30:19 GMT -6
Dumb question... why are the DAW manufacturers not doing this automatically for us? Brad Because they don't think it will do anything to sell more product to vast majority of their customers, most of whom are utterly ignorant technically. They'd much rather spend the development budget on flashy (but really fairly unimportant, technically) bells and whistles that will move more product to amateurs who have more disposable cash than brains or technical knowledge. Remember, people who are technically savvy constitute a very small - and dwindling - minority of people in the gear market these days. Furthermore, the corporate marketing people don't understand things like dither (and have no interest in learning) and are probably afraid of confusing and/or scaring off their customers. |
|
|
|
Post by johneppstein on Sept 11, 2017 21:36:06 GMT -6
The idea of not needing 24 bit dither has been widely promoted and people repeat what they've read or heard in a video. On my Duntech monitors I can almost always hear a difference and I've demonstrated this to others numerous times. But most of those people probably think that Beats are premium listening phones. After all, they paid $200-$300 for them... That's if they're not listening on $3 earbuds. They probably can 't hear the difference. |
|
|
|
Post by nightstalker on Sept 13, 2017 17:29:38 GMT -6
My question about dither.
If say I have a 60 track daw session using my symphony mkii and I sum it out to my dangerous 2buss+ (and any additional outboard)
Do I add dither only to the 16 summed outputs from my symphony
Or
Do I add dither on each of the 60 tracks individually.
I assume all 60 tracks but hoping the former.
|
|
|
|
Post by Bob Olhsson on Sept 13, 2017 17:44:03 GMT -6
ONLY dither the summed output channels that are feeding converters.
|
|
|
|
Post by nightstalker on Sept 14, 2017 1:01:50 GMT -6
ONLY dither the summed output channels that are feeding converters. Thanks for chiming in Bob. So in my case I'll be adding dither to only the 16 physical outputs that go into the summing amp that will feed back into the daw for a 2trk capture. I'm happy with that as that's not a workflow killer for my setup. Ive always assumed this but after reading through this thread with different setups from others I started to doubt myself. |
|
|
|
Post by Quint on Oct 22, 2024 6:33:40 GMT -6
To bring this thread back from the dead, I've been thinking about dither, as it relates to HW inserts. I know that, out of this thread, came at least somewhat of a consensus that dither is generally needed prior to HW inserts, due to changes in gain, among other things.
So, if you're using a HW inserts plugin, such as those available in Logic, S1, Cubase, etc. which have gain staging built into the HW inserts plugin, unless those HW inserts plugins have dither actually built into them (post gain) internally, you're not technically dithering after the last gain stage, even if you place a dither plugin immediately before the HW inserts plugin.
If so, then it makes me inclined to not use the gain staging within said HW inserts plugins at all, and instead just use a trim plugin, followed by a dither plugin followed by the HW inserts plugin, with no gain staging being done in the HW inserts plugin at all. And then another trim plugin after the HW inserts plugin.
This particular nuance on HW inserts and dither never got discussed when this thread was going a few years ago. Thoughts?
|
|
|
|
Post by Dan on Oct 22, 2024 7:28:45 GMT -6
To bring this thread back from the dead, I've been thinking about dither, as it relates to HW inserts. I know that, out of this thread, came at least somewhat of a consensus that dither is generally needed prior to HW inserts, due to changes in gain, among other things. So, if you're using a HW inserts plugin, such as those available in Logic, S1, Cubase, etc. which have gain staging built into the HW inserts plugin, unless those HW inserts plugins have dither actually built into them (post gain), you're not technically dithering after the last gain stage, even if you place a dither plugin immediately before the HW inserts plugin. If so, then it makes me inclined to not use the gain staging within said HW inserts plugins at all, and instead just use a trim plugin, followed by a dither plugin followed by the HW inserts plugin, with no gain staging being done in the HW inserts plugin at all. And then another trim plugin after the HW inserts plugin. This particular nuance on HW inserts and dither never got discussed when this thread was going a few years ago. Thoughts? that’s the right way to do it. Fixed point operations need dither. Float to fixed needs dither. Fixed to float does not and it’s a nice fact that all 24-bit fixed numbers map to an exact 32-bit float value. The reverse isn’t true. Float to float, the rounding error cannot be dithered but it is much less audible if at all, than truncation distortion, which can make the sample value incorrect because the systems has no way to no if the bottom bit is going up or down. Ideally the daw would do this for you but they don’t. There are some plugins with undithered fixed point truncation in them too. Infamously the brainworx plugins. There are still new brainworx plugins with truncation distortion in them. It’s a simple mathematical error that is absolutely ridiculous to have in 2024 when daws no longer run on primitive fixed point dsp chips like pro tools tdm did |
|
|
|
Post by Quint on Oct 22, 2024 7:43:14 GMT -6
To bring this thread back from the dead, I've been thinking about dither, as it relates to HW inserts. I know that, out of this thread, came at least somewhat of a consensus that dither is generally needed prior to HW inserts, due to changes in gain, among other things. So, if you're using a HW inserts plugin, such as those available in Logic, S1, Cubase, etc. which have gain staging built into the HW inserts plugin, unless those HW inserts plugins have dither actually built into them (post gain), you're not technically dithering after the last gain stage, even if you place a dither plugin immediately before the HW inserts plugin. If so, then it makes me inclined to not use the gain staging within said HW inserts plugins at all, and instead just use a trim plugin, followed by a dither plugin followed by the HW inserts plugin, with no gain staging being done in the HW inserts plugin at all. And then another trim plugin after the HW inserts plugin. This particular nuance on HW inserts and dither never got discussed when this thread was going a few years ago. Thoughts? that’s the right way to do it. Fixed point operations need dither. Float to fixed needs dither. Fixed to float does not and it’s a nice fact that all 24-bit fixed numbers map to an exact 32-bit float value. The reverse isn’t true. Float to float, the rounding error cannot be dithered but it is much less audible if at all, than truncation distortion, which can make the sample value incorrect because the systems has no way to no if the bottom bit is going up or down. Ideally the daw would do this for you but they don’t. There are some plugins with undithered fixed point truncation in them too. Infamously the brainworx plugins. There are still new brainworx plugins with truncation distortion in them. It’s a simple mathematical error that is absolutely ridiculous to have in 2024 when daws no longer run on primitive fixed point dsp chips like pro tools tdm did Yeah, it's super annoying that we have to think about this. The DAW should take care of all of this. Supposedly, Harrison Mixbus properly addresses dither, but they're the only one I've heard of that is actually maybe addressing it. I inquired about this with Luna, and got a response that didn't inspire confidence. The same can probably be said for all of the DAWs, unfortunately. |
|
|
|
Post by EmRR on Oct 22, 2024 8:06:22 GMT -6
I probably said it previously in this thread - one mystery, with a clue, is the DA chips in products like the MOTU 16A. Those interfaces are seen as 32 bit pipes, the chips are 32 bit processors with 24 bit outputs, so one might assume they are dithering.
|
|
|
|
Post by Dan on Oct 22, 2024 8:09:14 GMT -6
that’s the right way to do it. Fixed point operations need dither. Float to fixed needs dither. Fixed to float does not and it’s a nice fact that all 24-bit fixed numbers map to an exact 32-bit float value. The reverse isn’t true. Float to float, the rounding error cannot be dithered but it is much less audible if at all, than truncation distortion, which can make the sample value incorrect because the systems has no way to no if the bottom bit is going up or down. Ideally the daw would do this for you but they don’t. There are some plugins with undithered fixed point truncation in them too. Infamously the brainworx plugins. There are still new brainworx plugins with truncation distortion in them. It’s a simple mathematical error that is absolutely ridiculous to have in 2024 when daws no longer run on primitive fixed point dsp chips like pro tools tdm did Yeah, it's super annoying that we have to think about this. The DAW should take care of all of this. Supposedly, Harrison Mixbus properly addresses dither, but they're the only one I've heard of that is actually maybe addressing it. I inquired about this with Luna, and got a response that didn't inspire confidence. The same can probably be said for all of the DAWs, unfortunately. Yeah it’s patching around some monstrosity board that’s only good for faders and the bus but if we’re considering plugins to be inserts, we’re alway doing that. What comes with the daws is usually beyond awful for eq, dynamics, reverb, etc. it’s to the point that almost anything you can buy in 2024 is better with some choice exceptions like awful emulations (almost all of them and sometimes they’re even correctly emulating awful analog behavior), plugins with basic programming and mathematical errors, and some awful fx. |
|
|
|
Post by Dan on Oct 22, 2024 8:41:20 GMT -6
I probably said it previously in this thread - one mystery, with a clue, is the DA chips in products like the MOTU 16A. Those interfaces are seen as 32 bit pipes, the chips are 32 bit processors with 24 bit outputs, so one might assume they are dithering. we have no idea. Some interfaces do awful things internally. RME TotalMix truncates the output to 24-bits and the faders are truncated to 16-bit making it only good for monitoring mixes. I presume others are dithered or floating point. There’s always Paul Frindle posts and the old metric halo videos to tell of the warning signs of really bad code like rumble, zipper noises, and other artifacts. I stopped using some plugins because of these. |
|
|
|
Post by copperx on Oct 22, 2024 9:42:48 GMT -6
I probably said it previously in this thread - one mystery, with a clue, is the DA chips in products like the MOTU 16A. Those interfaces are seen as 32 bit pipes, the chips are 32 bit processors with 24 bit outputs, so one might assume they are dithering. Yes, lots of D/As are 32 bit float nowadays. Is dither necessary to go from 64 bit float (internal DAW bit depth) to 32 bit float? |
|
|
|
Post by damoongo on Oct 22, 2024 9:51:28 GMT -6
So let me get this straight - You guys want to add -140db of noise (dither) to HW inserts that will already be adding at least 100db of noise from the gear you're inserting? Dither isn't "correcting" any math. It's just covering it up with noise (which happens to be 40db lower than the noise floor of the high end analog gear you're inserting). Make it make sense... At 32bit float to 24bit, you AREN'T hearing truncation errors down at -144db in this HW insert scenario (or likely ANY scenario).
|
|
|
|
Post by EmRR on Oct 22, 2024 11:10:39 GMT -6
I probably said it previously in this thread - one mystery, with a clue, is the DA chips in products like the MOTU 16A. Those interfaces are seen as 32 bit pipes, the chips are 32 bit processors with 24 bit outputs, so one might assume they are dithering. Yes, lots of D/As are 32 bit float nowadays. Is dither necessary to go from 64 bit float (internal DAW bit depth) to 32 bit float? I'd say no offhand |
|
|
|
Post by gravesnumber9 on Oct 22, 2024 11:54:47 GMT -6
|
|
|
|
Post by Dan on Oct 22, 2024 12:16:27 GMT -6
Yes, lots of D/As are 32 bit float nowadays. Is dither necessary to go from 64 bit float (internal DAW bit depth) to 32 bit float? I'd say no offhand dithering that is physically impossible. it just rounds half to even and prevents dc offset. there has been some attempts to better round double precision to single precision but nothing that i know of works. i cannot really hear the 32-bit float rounding in many plugins. what i can hear is the filter structure crapping out when automating some eqs and 24-bit and 16-bit truncation affecting the perceived stereo image and immediacy of the audio. |
|
