Another thread about summing in scope

[Mike Goodwin]

Knowing weather or not it is best to use FLT ASIO drivers is important to me and has seemingly gone completely uncovered. It seems to me to be a very fundamental question that almost all users of the scope environment can benefit from, if even just by a few rounding errors.

my experiment shows that both 24bits and 32bit float have the same resolution after conversion, but the rounding is different.
while you get more accuracy on the quiet parts of 32bit float, you get less accuracy on the loud parts. its a give and take situation.

i think you cant tell the difference anyhow, and it all gets converted to 64bit float inside the daw anyhow.

so to answer your question - if you are recording through converters - record at 24bits integer to keep exactly the same audio that got into the card. if you are recording a synth - record in 32bits float - to keep your sound as much in tact as you can (well it gets quantized from 64bits anyhow.) all the rest is irrelevant.

Thanks Red MuZe. When you say "recording a synth" I take it you are refering to a synth in the Scope enviroment. If so is recording a soft synth from my DAW (running at 32-bit) would be the same situation?
I am never recording from the analog world these days. I do not have room for the instruments in my studio and will not for a LONG time. I will be moving into a 6.5' by 9' room in about a year That is besides the point.

So if I understand you right....
The bottom line is that if you want to have better accuracy in the louder portions of audio it is best to use the 32-bit flt ASIO drivers when going to and from your DAW. Is that correct?

[ReD_MuZe]

Thanks Red MuZe. When you say "recording a synth" I take it you are refering to a synth in the Scope enviroment.

no sorry. i meant recording a vsti calculated in float. actually for a scope synth id favor 24bit int which is the quality i hear it in through the DACs.

If so is recording a soft synth from my DAW (running at 32-bit) would be the same situation?
I am never recording from the analog world these days. I do not have room for the instruments in my studio and will not for a LONG time. I will be moving into a 6.5' by 9' room in about a year That is besides the point.

do you have room in your studio for a microphone?

So if I understand you right....
The bottom line is that if you want to have better accuracy in the louder portions of audio it is best to use the 32-bit flt ASIO drivers when going to and from your DAW. Is that correct?

[/quote]
no the oposite. 24 bit has a nicer rounding algo for louder parts. but you DONT want to get into rounding discussions this is WAY more convoluted than bit counting, and even top scientists don't agree on which method is best for what.
so basically as i said the difference is inaudible. it is the same resolution and you cant tell the difference. go with whatever you like. once you turn any kind of knob on a processor you loose much more data than the difference between 24bit int and 32bit float.

so again. dont worry about it.
i use 24bit int. i like the number

[Warp69]

simple.

becouse 24bits has slightly better error correction at the top end.

look at this pic:

32bits float are better for silence (the stuff you don't hear) and teh difference between 24bits and 32bits is just by one point.

when you go to the high range (the stuff you actualy hear) 24 bits is on the upper hand with a nicer rounding algorithm that makes rounding errors of + - 64 points while 32bit float reaches errors of 128points.

both have the same precision and different rounding algos. this is the experiment i did on reaper. comparing 32bit float drivers and 24bit integer drivers.

Could you please explain what you do to get those numbers?

Lets take the positive large 32bit float numbers :

32bit int (Scope) -> 32bit float (Driver) -> 64bit float (Reaper) -> 32bit float (Driver) -> 32bit int (Scope)

Then you compare the 32bit input with the 32bit output, right?

1.431.655.765 as 32bit input -> blah blah blah 32bit float blah blah blah -> 1.431.655.680 as 32bit output

And there's a difference of 85 points between those numbers, when using 32bit float drivers? If thats the case - would you please make the exact same test (same numbers) with 24bit int drivers, so we could compare the different results.

And the same goes for the 2.000.000.120 test - It would be appreciated if you could make that test with 24bit int drivers.

Thank you.

[voidar]

I say screw the soft parts, the loud ones are more important as they are what will cut in a mix.
If you can use ASIO 32 drivers with Live then go for that.

I wonder if this all could explain why some say VDAT sounds better compared to ASIO based recording software? How valid could this be?

[ReD_MuZe]

in the pic i posted there is already the comparison with the 24bit drivers...

on the top range there is a rounding error os 128 points for both drivers.
the 24bit driver rounds to the nearest neighbor while the 31bit float is rounded to the highest neighbor which gives a possibility for larger errors...
on the quiet numbers the situation is reverse. 32bits is more accurate than 24bits

you can d/l reaper here:
http://www.cockos.com/reaper/
and use pulsars drivers to run the test.

i didn't know how to show this since i only had logic on my pc but indeed reaper is a great solution you can really probe it in realtime :>

voidar -> this rounding effect is neglectable you cant hear it anyhow.
and yes VDAT is 32bit integer fully normalized audio. this is the only way for now to "retain the sound as it is" in scope. however, i don't share the opinion that it sounds better. i cant tell the difference, and my ears are in top shape

[voidar]

Red_Muze:

Your numbers did baffle me as my testing with REAPER hardly gave me those differences, but since you used Logic I guess this is why. Also I tested the 24/32 bit integer driver vs. the 32-bit float driver and it stayed within 1-2 points the whole range it seemed. The float driver performed better at the lower registers though and diverged more at the top range, as it should right?

You're entitled to your opinion of course and I don't know what to believe these days myself, though I do remember long ago thinking there was an audiable difference. Anyway, I too am looking for the "best way" to interface the DSP to 3rd party software so interesting topic.

[ReD_MuZe]

i used reaper not logic. logic on pc cant do 32bit float. re-read what i wrote

the 32-bit float driver and it stayed within 1-2 points the whole range it seemed. The float driver performed better at the lower registers though and diverged more at the top range, as it should right?

i dont understand did it or didn't it stay within the points?
it maybe that you have asio monitoring on, and then you are just getting the int signal from scope (it baffled me at first because i got it when i tried testing the 24bit drivers and has no lost data ehehe!)

again. why are you depending on numbers to decide? use your ears. you cant hear any difference? well then its not important is it?

[voidar]

I did have record monitoring enabled (ASIO monitoring doesn't work), so it's a full ASIO cycle delay. It's all live feed-through, no recording.

ASIO2-32 driver performed within 1-2 points the whole range.
ASIO2-FLT performed worse in the higher range. I don't remember the points.
VDAT performed the best but could diverge by 1 point.

Sure, one should use the ears, and that is what people do when they claim A sounds better than B etc.

Ignorance is bliss, but when you're aware of all the tech, it can surely destroy your wibe .

[ChampionSound]

[...]
and yes VDAT is 32bit integer fully normalized audio. this is the only way for now to "retain the sound as it is" in scope. however, i don't share the opinion that it sounds better. [...]

But what about the record inputs of the STS samplers? In theory they are the only other solution besides VDAT to record scope audio "as it is". No ASIO or Wave drivers involved. It isn't the most user friendly way to record audio, but might have equal quality as VDAT, no?

Sorry for going slightly OT ...

[voidar]

I tested STS too. It didn't perform just as VDAT, but it is be neglible really. Might just be the gain control in the sampler.

[Warp69]

Could someone please enlighten me regarding setup of REAPER?

ASIO1-32 and ASIO-Flt both get truncated to 24bit in REAPER when I record - how do I prevent that?

When I use ASIO1-32 as Asio driver, REAPER shows Scope 24bit on the armed track.

Thank you

[ReD_MuZe]

reaper does not recognize the 32bit int driver. you can only use the float driver in reaper.
the 32bit driver acts just like 24bit - because it cant work on 32bit int,
but the float driver has a different behavior - as expected.

also i found that in order to get the right depth i had to restart reaper after i changed the asio drivers.

i did the same test with synthmaker - same results.

but guys i think you are blowing this way out of proportion. recording to STS!?
your time is better spent learning some musical theory, arrangements, counterpoint, and mixing...

[Mike Goodwin]

do you have room in your studio for a microphone?

So if I understand you right....
The bottom line is that if you want to have better accuracy in the louder portions of audio it is best to use the 32-bit flt ASIO drivers when going to and from your DAW. Is that correct?

"no the oposite. 24 bit has a nicer rounding algo for louder parts. but you DONT want to get into rounding discussions this is WAY more convoluted than bit counting, and even top scientists don't agree on which method is best for what.
so basically as i said the difference is inaudible. it is the same resolution and you cant tell the difference. go with whatever you like. once you turn any kind of knob on a processor you loose much more data than the difference between 24bit int and 32bit float.

so again. dont worry about it.
i use 24bit int. i like the number "

Yes I have a couple of microphones but I cant say I have use one in the last 12 months This being said I need to know that I am getting the best rounding when recording my 16 month old daughter for the first time

After reading your post for a second time this morning I did see that I was misunderstanding you but thank you very much for clarifying. It seems for what I do I would go with the 24 bit drivers in order to hear no disernable difference So what you are saying is that I would be better off buying a product like this http://www.musicdirect.com/product/73105

Now again just to me totally clear. you feel that these errors are far to small to be heard, but we are talking about single tracks and only one pass in or out of the Scope environment. Do you feel that even with many (32+ tracks or so) stacked up, passing in and out of the environment via ASIO drivers there would still be no "worthwhile" difference? I don't want to put you on the spot. I am just looking for a friendly opinion. I know, I know just listen and figure it out for yourself. That is the truth in the end. And I to agree that none of this tech knowledge will make up for a lack of ability.

All the best

[Warp69]

I thought REAPER supported 32bit drivers.

But unfortunately in my setup - REAPER truncates every 32bit driver to 24bit, including 32bit float.

1.431.655.765 = 1010101010101010101010101010101 -> 1010101010101010101010100000000 = 1.431.655.680
311 = 0000000000000000000000100110111 -> 0000000000000000000000100000000 = 256

And every thing below 256 (32bit int) translates to 0 - just like normal truncation to 24bit.

Since you have REAPER working with 32bit float setup, I would appreciate if you could show me how I prevent truncation to 24bit.

To others that have 32bit float working - which DAWs are you using?

Thank you.

[tgstgs]

@Mike Goodwin:
audio in scope is 32bit integer;
so if you record at 32bit integer there is NO algo between and the only way to record your doughter best!
-------------
@Warp69
your math is absolutely correct but tricky coders there are
--------

its said:
you take a 32bit integer value;
you convert to 24bit integer;
convert to double for internal audioprozessing as Voidar told us;
convert back to 24bit integer for output;
and convert back to 32bit integer in scope;
i dont think this is realy happening here;
---------
lets just look at the 24bit test to stay simple
lets say we have a blackbox you put some values in;
well for the math:

here we go:

_you put in _ 2000000060
if i /256 i get 7812500,234375;
now forget the numbers below 1 as it would be with a bitshift and we have 7812500;
if i now *256 i get 2000000000;
is about your (ReD_MuZe Picture) result if you have in mind the positive 1 loss for symetrie;

_next: 1514251289/256=5915044,09765625
forget the rest 5915044*256=1514251264
again it fits your result;

but now it gets strange (tricky):

_ 137659207/256=537731,27734375
forget the rest: 537731*256=137659136
but you get 137659199

the only way to get this is:
137659207 has a leading 0 in hex format;
so do a bitshift (*16) and you get:2202547312/256=8603700,4375
forget the rest: 8603700*256=2202547200
do a bitshift back (/16) and you get 137659200

got it?
--------
of course all is done in bitshift;
multiplication and division would take to much circles;

hope this helps

bitshifting vibes from vienna

[ReD_MuZe]

yep the way to convert between types is bit-shifting which results in loss of data. (you move all bits to one direction, essentially clipping part of the signal)
there is no conversion to 24bits in reaper if you use the flt driver
(you need to set your project to be 32bit float in project settings)

[tgstgs]

loss of data = brutal use of words!
one with no knowledge of coding could think loss of data is bad;
but it depends on the task to do;
wouldnt it be cool to loose the noise floor of a recorded mic for example?
how should the blackbox look like if we want to realise such an effect;
_

i think its time to shift myself out here;

good vibes

[ReD_MuZe]

loss of data = brutal use of words!

no other way to put it.

one with no knowledge of coding could think loss of data is bad;

it seems that some with knowledge of coding could think its bad aswell

but it depends on the task to do;
wouldnt it be cool to loose the noise floor of a recorded mic for example?

getting rid of the mic noise-floor is the last thing we are trying to achieve when recording. and certainly not through bitshifting.

[tgstgs]

well then try cooling down;

sorry couldnt resist!!!!!!!!!!!!!!!!!!!!

over and out vibes from vienna

edit:
should be a joke!!!
hope you get it right!
-----
cooling down the system helps reduzing the noisefloor!


in deep respect!!

[voidar]

REAPER reports the driver format in the upper right corner. It says 32-bit for me using either 32 or FLT. REAPER supports 32-bit PCM drivers (!). Are you sure you have set the correct recording format? There are many supported, though only WavPack and AIFF supports 32-bit PCM recording.

The 32 driver sais "Scope 24-bit" though, but this is just a string.