|
Welcome to the Piano World Piano Forums Over 3 million posts about pianos, digital pianos, and all types of keyboard instruments. Over 100,000 members from around the world.
Join the World's Largest Community of Piano Lovers
(it's free)
It's Fun to Play the Piano ... Please Pass It On!
|
|
65 members (aphexdisklavier, bobrunyan, anotherscott, AaronSF, apianostudent, beeboss, brdwyguy, benkeys, 14 invisible),
2,188
guests, and
386
robots. |
Key:
Admin,
Global Mod,
Mod
|
|
|
Joined: Aug 2017
Posts: 83
Full Member
|
OP
Full Member
Joined: Aug 2017
Posts: 83 |
Hi
In regard to recording solo piano (CFX Concert Grand in to Garageband) what bit depth and sample rate would be best recommended?
|
|
|
|
Joined: Apr 2007
Posts: 7,268
7000 Post Club Member
|
7000 Post Club Member
Joined: Apr 2007
Posts: 7,268 |
You will be OK with 16-bit 44.1kHz.
|
|
|
|
Joined: Apr 2019
Posts: 2,067
2000 Post Club Member
|
2000 Post Club Member
Joined: Apr 2019
Posts: 2,067 |
Hi
In regard to recording solo piano (CFX Concert Grand in to Garageband) what bit depth and sample rate would be best recommended? Depends on the bit-depth of your samples. If they are 24-bit or more, I'd go with 24-bit and then convert the final mix to whatever resolution necessary. Many samples are usually 16-bit 44.1kHz which is CD-Quality but for pianos, good libraries should provide 24-bit samples as well. More dynamic range is audible for ears but there are many factors involved. Read this: A good link to read!
|
|
|
|
Joined: Oct 2013
Posts: 3,868
3000 Post Club Member
|
3000 Post Club Member
Joined: Oct 2013
Posts: 3,868 |
See http://drewdaniels.com/audible.pdf ... a rendering at a CD quality is enough for most people. I see 24bits more useful when recording an audio input. Since we set the gain with an headroom which avoids clipping. A 16 bits record could have only 15 or 14 useful bits (higher bits at 0). There were here also a workshop about audio quality : https://youtu.be/BYTlN6wjcvQ
Last edited by Frédéric L; 07/10/20 10:50 AM.
|
|
|
|
Joined: Aug 2009
Posts: 1,625
1000 Post Club Member
|
1000 Post Club Member
Joined: Aug 2009
Posts: 1,625 |
Hi
In regard to recording solo piano (CFX Concert Grand in to Garageband) what bit depth and sample rate would be best recommended? If it were my project, I would record at the native sample rate of the library, which might be 44.1, but I would always record at 24 bit. 24 bit affords you the opportunity to process the recording to 16 bit and the end result will be better.
|
|
|
|
Joined: Oct 2013
Posts: 3,868
3000 Post Club Member
|
3000 Post Club Member
Joined: Oct 2013
Posts: 3,868 |
When we record a virtual piano, the actual record is in MIDI events. You can do whatever you like with it (changing/mixing perspectives...)... the 16/24bits setting will happen late : when you want to make your WAV file. All processing will be made with the internal DAW bit depth (32bits or even 64bits with Cubase 9.5) and the bit reduction will be the last step.
I think the original poster means « rendering the CFX Concert Grand ».
Last edited by Frédéric L; 07/10/20 11:34 AM.
|
|
|
|
Joined: Aug 2009
Posts: 1,625
1000 Post Club Member
|
1000 Post Club Member
Joined: Aug 2009
Posts: 1,625 |
When we record a virtual piano, the actual record is in MIDI events. You can do whatever you like with it (changing/mixing perspectives...)... the 16/24bits setting will happen late : when you want to make your WAV file.
I think the original poster means « rendering the CFX Concert Grand ». When you have your midi recording; do you have an over-all volume slider that raises all the notes volume?
|
|
|
|
Joined: Oct 2013
Posts: 3,868
3000 Post Club Member
|
3000 Post Club Member
Joined: Oct 2013
Posts: 3,868 |
When you have your midi recording; do you have an over-all volume slider that raises all the notes volume? You can have such a slider in your DAW and in your virtual piano (Garritan CFX have it). Then after the record, you can note the headroom before clipping, adjust the volume to reduce it and render the record to a file with a nearly full dynamic. Note that Garritan CFX has a limiter. If you set a volume high enough, you would better use the CFX volume since the limiter will reduce or avoid the effect of clipping. Or you can use a Garageband Limiter.
Last edited by Frédéric L; 07/10/20 11:43 AM.
|
|
|
|
Joined: Aug 2009
Posts: 1,625
1000 Post Club Member
|
1000 Post Club Member
Joined: Aug 2009
Posts: 1,625 |
When you have your midi recording; do you have an over-all volume slider that raises all the notes volume? You can have such a slider in your DAW and in your virtual piano (Garritan CFX have it). Then after the record, you can note the headroom before clipping, adjust the volume to reduce it and render the record to a file with a nearly full dynamic. Note that Garritan CFX has a limiter. If you set a volume high enough, you would better use the CFX volume since the limiter will reduce or avoid the effect of clipping. Or you can use a Garageband Limiter. Appreciate the ‘schooling’; I would still render at 24 bit but I also know that piano recordings benefit from no processing. Thanks
|
|
|
|
Joined: Aug 2019
Posts: 1,395
1000 Post Club Member
|
1000 Post Club Member
Joined: Aug 2019
Posts: 1,395 |
When we record a virtual piano, the actual record is in MIDI events. You can do whatever you like with it (changing/mixing perspectives...)... the 16/24bits setting will happen late : when you want to make your WAV file. All processing will be made with the internal DAW bit depth (32bits or even 64bits with Cubase 9.5) and the bit reduction will be the last step.
I think the original poster means « rendering the CFX Concert Grand ». Small clarification: MIDI is converted to audio at VSTi (CFX) plugin output. If you apply any plugins after that (FX, EQ, reverb, etc), they will deal with audio, not MIDI. The highest quality natively supported audio (especially bit depth) out from VSTi in such a case would mean a higher final result. If you record directly the CFX output without any postprocessing, then you can record it with any (CD or better) audio quality natively provided by CFX. And because CFX as far as I know supports only 44.1KHz /16bit, then this is the only output option worth consideration.
Wise men speak because they have something to say; Fools because they have to say something. (falsely attributed to Plato) Vlad, Adult beginner
|
|
|
|
Joined: Oct 2013
Posts: 3,868
3000 Post Club Member
|
3000 Post Club Member
Joined: Oct 2013
Posts: 3,868 |
Yes reverb or other effect are applied on the audio path, not the MIDI path, but generally a DAW works at a given bit depth (32 bits for example), and when we choose a bit depth it is only about what is recorded or rendered on the disk.
I have seen the 24bits preference in GarageBand. I don’t think it applies to internal audio streams. (CFX to reverb). This would mean an added complexity to any Audio Units which instead of supporting one default bit depth (32 or 64 bits since it is a common CPU register size), would have to support 16 or 24 bits and others...
|
|
|
|
Joined: Aug 2017
Posts: 1,756
1000 Post Club Member
|
1000 Post Club Member
Joined: Aug 2017
Posts: 1,756 |
The 24-bits are almost certainly padded (least significant bits equal to 0 or 1) to 32-bits or even 64-bits. Or something of the sorts.
Kawai ES8, Roland RD2000, Yamaha AG06 mixer, Presonus Eris E5 monitors, Sennheiser HD598SR phones.
|
|
|
|
Joined: Aug 2019
Posts: 1,395
1000 Post Club Member
|
1000 Post Club Member
Joined: Aug 2019
Posts: 1,395 |
The 24-bits are almost certainly padded (least significant bits equal to 0 or 1) to 32-bits or even 64-bits. Or something of the sorts. I don't think so. For generic numbers yes, such conversion would be trivial, but we are talking about audio encoding where both 2^32 − 1 (for 32 bit resolution) and 2^24 - 1 (for 24 bit) mean 100% sound level (loudest possible), and 0 means no sound. Conversion between 32 and 24 bit values, or between 24 and 16, is not precise. Conversion from 16 to 32 bit would be precise, but not vice versa. With all that said, as long as your final output is 16 bit, these conversions would be hardly noticeable, not by human ears IMO. Just more CPU load.
Last edited by VladK; 07/10/20 03:10 PM.
Wise men speak because they have something to say; Fools because they have to say something. (falsely attributed to Plato) Vlad, Adult beginner
|
|
|
|
Joined: Aug 2019
Posts: 1,395
1000 Post Club Member
|
1000 Post Club Member
Joined: Aug 2019
Posts: 1,395 |
Upconversion between 44.1, 48, 96 kHz would be more tricky, because computer never knows which approximation to choose for best result. There are so many approximation algorythms, computer would make decision based on history of past deltas, and if there is big enough buffer it might also use future deltas for analysis. Threshold might be dynamic or static, etc. This is where FPU is used.
Last edited by VladK; 07/10/20 03:26 PM.
Wise men speak because they have something to say; Fools because they have to say something. (falsely attributed to Plato) Vlad, Adult beginner
|
|
|
|
Joined: Aug 2017
Posts: 1,756
1000 Post Club Member
|
1000 Post Club Member
Joined: Aug 2017
Posts: 1,756 |
Vladk, sorry but I think you are thinking of interpolation. No interpolation at all. Simple complement of two's math. Who programs at hardware level (like me) knows what I mean (you know, there are 10 types of people, those that think binary and those that do not ).
Last edited by EVC2017; 07/10/20 03:46 PM.
Kawai ES8, Roland RD2000, Yamaha AG06 mixer, Presonus Eris E5 monitors, Sennheiser HD598SR phones.
|
|
|
|
Joined: Aug 2019
Posts: 1,395
1000 Post Club Member
|
1000 Post Club Member
Joined: Aug 2019
Posts: 1,395 |
Vladk, sorry but I think you are thinking of interpolation. No interpolation at all. Simple complement of two's math. Who programs at hardware level (like me) knows what I mean (you know, there are 10 types of people, those that think binary and those that do not ). Yes I am talking about interpolation. You are talking about binary conversion where 16 bit is subset of 24/32 bits, but in audio they both cover the same 0-100 dynamic range, just with different steps.
Last edited by VladK; 07/10/20 04:06 PM.
Wise men speak because they have something to say; Fools because they have to say something. (falsely attributed to Plato) Vlad, Adult beginner
|
|
|
|
Joined: Aug 2019
Posts: 1,395
1000 Post Club Member
|
1000 Post Club Member
Joined: Aug 2019
Posts: 1,395 |
[If conversion was done by complementing with zeroes, it would not improve precision during calculations, because it would not increase delta between two neighboring originally 16 bit numbers usable for rounding results during downconversion. Let look at 1 and 3. In 16 bit you would have 0000000000000001 0000000000000011 You propose to convert it also to 1 and 3 but in 32 bits 00000000000000000000000000000001 00000000000000000000000000000011 Such conversion is useless, because delta between them is still 2.
You need to convert 3 to 5 instead (complement every bit with 0) and get twice as much space for rounding when you convert back to 16 bits: 0000000000000000000000000000101 which would give you twise as much space for rounding errors and more precise downconversion back to 16 bits.
Simply speaking, if after manipulations we get 6, 5 or 4 (110 to 100) it is downconverted to 3, but if we get 3, 2, or 1 (011 to 001) it is downconverted to 1, zero is always zero of course.
Conversion to/from 24 would require calculations, simple bit operations would not work here.
Last edited by VladK; 07/10/20 04:32 PM.
Wise men speak because they have something to say; Fools because they have to say something. (falsely attributed to Plato) Vlad, Adult beginner
|
|
|
|
Joined: Aug 2019
Posts: 1,395
1000 Post Club Member
|
1000 Post Club Member
Joined: Aug 2019
Posts: 1,395 |
Actually, when we upconvert from 16 to 32, we should complement from right, not left, which gives even more delta space and flexibility. 1 would become 2, 2 would become 8, and so on.
Wise men speak because they have something to say; Fools because they have to say something. (falsely attributed to Plato) Vlad, Adult beginner
|
|
|
|
Joined: Aug 2017
Posts: 1,756
1000 Post Club Member
|
1000 Post Club Member
Joined: Aug 2017
Posts: 1,756 |
Hi, Vlad. Maybe we are talking of different things. I was assuming 24-bit raw samples and "converting" them to 32-bits (even 64-bits if that makes sense) by simply multiplying by 2^8 to have the same sample normalized to 32-bits. It was just a quick example, I did not want to elaborate. To be honest, I don't even think it makes sense multiplying by 2^8 if the samples are being processed or added to others. With the 24-bits, you may have up to 256 channels with the same resolution summed up (oversimplification). After all processing, including normalization to get the highest signal to noise ration) you may eliminate the least significant bits to obtain the required resolution (e.g. the CD-quality 16-bits) with the minimum possible distortion and noise (I mean noise added by digital processing, not previously exhisting in the signal source and preamplification before AD conversion).
About your example above (2, 3, 5 etc), you're right, you do not gain resolution. But interpolation would add distortion and/or noise as you do not know what was the real analog signal that led to the digital value.
Finally, I doubt the heavy lifiting in terms of processing is done in integer or fixed point math in these DAWs, but you seem to agree as you mention the FPU above.
Last edited by EVC2017; 07/10/20 04:44 PM.
Kawai ES8, Roland RD2000, Yamaha AG06 mixer, Presonus Eris E5 monitors, Sennheiser HD598SR phones.
|
|
|
|
Joined: Apr 2007
Posts: 7,268
7000 Post Club Member
|
7000 Post Club Member
Joined: Apr 2007
Posts: 7,268 |
Vlad, when you upsample, you add zeroes to the right (LSB, not MSB).
For instance 0000000000000001 (a 16-bit value)
would become
000000000000000100000000 (a 24-bit value)
This way you keep the dynamic range. Same for downsampling, you just remove LSB-bits. No need for truncations or any complex math. Unless I haven't understood what you mean.
|
|
|
|
|
|
Piano
by Gino2 - 04/17/24 02:34 PM
|
Piano
by Gino2 - 04/17/24 02:23 PM
|
|
Forums43
Topics223,408
Posts3,349,457
Members111,637
|
Most Online15,252 Mar 21st, 2010
|
|
|
|
|
|