Without getting envolved with what is right or what is wrong....I usually export my files to the same bit rate minimum 24bit and import the file into the mastering setup and then export my final to 16bit. I don't thing many consumer have devices that will play anything much higher than 16bit anyway...

John
You as usual are correct, but this time I use the MOSTLY qualifier. Malts are for drinking. I have about 20-25 of them at home at any given time. I dont drink much, but if we were ever attacked by zombies and the supply chain from Scooterland was severed, I need to protect me and my family.

I was quoting George Massenberg.

Even in the Analog days if I ever had to change a word length I would cut up bits of audio tape and reconnect them with bits of adhesive tape. See there's always been bits and words of varying length.

Go on John. It's up to you now. I dare you to be sillier.

It would only float if I dithered too much. But that was regardless of word length or how many bits were used.

I never tried it with any pro gear but often attempted to do it with just a 4 track tape on a glass block with a scalpel and would experiment with different adhesive tapes trying to find one that would stick to the audio tape but without the glue dissolving over time. The operative word being "attempted"; very amateurish. Pretty much like my digital efforts. All fun though.

ORIGINAL: Mr. Ease

So John,

Joking aside, you did state Don was correct. Do you have any response (or does indeed DonM with whom I was originally debating the point) to the questions I asked in my previous post.

I am not posting for the good of my health here, just pointing out something I think is misunderstood. I don't care for myself but I don't think this forum is a place where anyone wants misinformation to stand uncorrected. Something that you apparently stand for around here too!

I'm just trying to achieve a true definitive answer. Just responding to the humour (which I fully appreciate!) does not really get us there.

Mr. Ease et al..

Sorry that I have been to busy to read any follow up threads, and I haven't even read before posting so apologies in advance!

As I mentioned before, I have appreciated this thread and most particularly your posts when they sought to clarify the need for dither in 2 byte source files. Clearly worth discussing. Be careful that your responses don't sound like you've just come from Starbucks with a double caffe!

Here's where reading the posts may help me before posting myself, but here goes:

I'm not sure if you read some of the earlier threads that many in this thread have participated with regard to the application of dither. Most particularly my history is one being an advocate for it's use - with one very important preamble. Material Matters.

Some of the folks I speak with often have recently been this very discussion wrt the use of dither and the discussions have been pretty similar to our (without the caffiene sp.)

So I'll go back and read the posts - my goal - identical to yours - is to find the most appropriate choice given the source and the intended outcome. Keep in mind however - (this comes from my class, so you're now getting graduate credit in Sound Design) This is both science and art. The goal you seek may be mathematically correct but not relevant in the domain of art. Many folks argue on the topic of dither - I love the irony since in fact dither itself is noise.

I suspect that our friend Bob Katz would agree with you. I infrequently correspond with him - but feel sure if I sent him a condensed form of the question he'd respond. I respect Bob, and apply his methods (including dither choices) 90% of the time. Last month Bob posted (I believe at GS) a great set of guidelines for comparisons and educating the test in such circumstances.

Now to condense my art and Bob Katz thinking. If you are working with 16 bit source files that are already mastered and are pop/rock in nature with RMS above -32dbfs and peaks between -3dbfs and DBFS - (the RMS is generous if the material represents today's mix/masters - the rms is probably -18dbfs or higher) you'll never hear any dither algorithm.

I think I totally agree with all of your math (again apologies I haven't read the threads) - with respect I probably have better examples proving your point that I use in my classes - however the material is 4 member female sacred choral with very delicate room reverb tails and a well mic'ed Jazz quartet - both I recorded with the sole point of exaggerating the need for dither in many situations.

I'll repost if I get back to the forum in the week with some time.

Thanks.

-D

I am still with you Mr. Ease.

I think the main point to understand is that the word length increases non-destructively inside a daw. Essentially, if you import a 16 bit file in sonar and lower the volume .1 dB, you have potentially increased the word length to 64 bits. Check it out, use a bit meter.

Think about it, take a single digit number like 1 and divide it by 3 and what do you get? 0.3333333333333333~ etc. The same thing works in binary. Simple truncation might give you only 0 when there should be an average of 0.33~ using a combination of 1 and 0. That is where you need dither to help with the rounding off. Personally, I would rather have something dithered twice, (the increase in noise is negligable), than risk quantization distortion from the added processes.

ORIGINAL: rumleymusic

I am still with you Mr. Ease.

I think the main point to understand is that the word length increases non-destructively inside a daw. Essentially, if you import a 16 bit file in sonar and lower the volume .1 dB, you have potentially increased the word length to 64 bits. Check it out, use a bit meter.

Think about it, take a single digit number like 1 and divide it by 3 and what do you get? 0.3333333333333333~ etc. The same thing works in binary. Simple truncation might give you only 0 when there should be an average of 0.33~ using a combination of 1 and 0. That is where you need dither to help with the rounding off. Personally, I would rather have something dithered twice, (the increase in noise is negligable), than risk quantization distortion from the added processes.

These points I have already raised but the issue is possibly more complex as Don is indicating. That is why I posed the question about levels of attenuation. With 13 dB of attenuation, any previous dither applied to the original 16 bit file would be lost when converted back to 16 bits. With your example of 0.1 dB attenuation however, the two dither components could add and increase the amount of noise added by dither by up to 6 dB. There are other factors involved in whether the dither may become intrusive such as the dither algorithm used in each case (E.g. POWR1, 2 or 3). For myself, I believe that dither should still be applied (most certainly if effects or other tracks are added) but the choice of dither algorithm is probably more critical.

That is why I am interested in Don's response as to the criteria where he believes dither should NOT be applied as there are certainly scenarios (as I indicated) where the original dither is lost and the arguments to use no dither become moot points.

These points I have already raised but the issue is possibly more complex as Don is indicating. That is why I posed the question about levels of attenuation. With 13 dB of attenuation, any previous dither applied to the original 16 bit file would be lost when converted back to 16 bits. With your example of 0.1 dB attenuation however, the two dither components could add and increase the amount of noise added by dither by up to 6 dB. There are other factors involved in whether the dither may become intrusive such as the dither algorithm used in each case (E.g. POWR1, 2 or 3). For myself, I believe that dither should still be applied (most certainly if effects or other tracks are added) but the choice of dither algorithm is probably more critical.

That is why I am interested in Don's response as to the criteria where he believes dither should NOT be applied as there are certainly scenarios (as I indicated) where the original dither is lost and the arguments to use no dither become moot points.

I think the question is whether dither used twice creates more noise than truncation at any level. I seems that whether you decrease the volume 13 db or .1 db the dither noise added would theoretically be a similar volume in relation to the music (actually greater the more you decrease the volume). I would think however, that a good noise shaping algorithm like POW-r 3 would be able to hide that noise rather well, and that distortion created by the lack of dither might be more noticeable. I guess the only way to tell is to do a test.

Just a gentle bump in the hope that Don will find the time to respond.

Cripes, I didn't realise it was a competition for brownie points! :<).

To be honest I am sure there is something in what Don is saying. If you import an already dithered 16 bit file then there is absolutely nothing to discuss if you then export it with no change whatsoever. If it is attenuated by 13 dB then the original dither becomes mathematically insignificant so an export with no dither would be simply truncated and there is a strong case to use dither - particularly as the signal is now lower in level and the truncation artifacts will be relatively louder. With less than 12 dB, some proportion of the original dither will still be present. Clearly my opinion in this grey area is that dither should be used but it is surely a sliding scale - the more attenuation is applied, the more dither becomes necessary.

Of course if other processing is applied, say compression, because for instance the average loudness on one of the pre-mastered tracks is significantly lower than the other imported tracks (c.f. O.P.'s situation) then I would argue that dither must be used on this track.

As Don has much experience and respect around here, I am interested to hear his comments on this. I fully appreciate his comments on listening versus maths. This is why I have put different (maybe abstract) mathematical cases up to make a clear case for dither/ no dither scenarios. It is the grey area that is possibly of most interest though as that is the area where most practical examples would probably sit.

Let's face it, the need for dither is not a requirement of mathematical accuracy, rounding optimises that quite happily. Dither only arrived due to perceived psycho acoustic benefits.

Just as a side note, as it has been mentioned hereabouts. I still cannot believe that dither is necessary when converting to 24 bit from either 32 or 64 bit floats. This is purely because the dither (in the 2 LSB's) is still way below the inherent noise of any 24 bit D-A converter. 1 LSB of truncation OR rounding noise is around -140 dBFS and insignificant to even the best converters.

I think you have missed a lot of what has been discussed here. If you are interested please read from post #10 onwards for more detail. There IS a difference between 0.1 dB and 13 dB when it comes to whether you will be summing the original dither to any post processing dither you might add.

No, I read it. I'm don't agree with all of it. But now it seems to be more of a competition of ego, so good luck on the arguments.

ORIGINAL: DonM

ORIGINAL: Mr. Ease

Just a gentle bump in the hope that Don will find the time to respond.

Richard:

Sorry I am late to respond - I have been terribly ill - a tremendous flu has kept me in bed since last week with a temperature of 102.5 degrees - sorry I don't know the centigrade conversion

I will get some time hopefully when I recover to revisit this important issue - thanks to all for haning in there!

-D

Hi Don,

Sorry to hear of your illness - I hope you are recovering well. I had noticed you had not been posting on the forum but thought you may just be busy or away. I'm sure either would have been preferable to reality!

Come back when you are fit and healthy. I agree it's an important issue but it certainly isn't urgent.

To Rumley - I don't think either Don or I are needing ego massages here... fcarosome brought that up and my response was intended to reflect that but perhaps it wasn't worded very well! For me it isn't about winning or losing, it's about sharing and/or gaining knowledge.

ORIGINAL: Mr. Ease
For me it isn't about winning or losing, it's about sharing and/or gaining knowledge.

Agree.

I read the whole thread in a breath, and I perceived some ego was raising, perhaps I was just wrong.

Cheers
fc