About double precision (AudioProcessor, AudioBuffer)

General JUCE discussion
1

Hello,
in AudioProcessor and AudioBuffer
it looks it can use double precision instead of float

enum ProcessingPrecision
{
    singlePrecision,
    doublePrecision
};

If it’s better, I want to use double precision

So, my questions are

  1. Is it better to use double precision in audio processing ?
    I have sufficient memories and HDD spaces
    My audio interface is Audient ID22 and I have RME Converter, native instruments, Dorico, Ableton Live etc

  2. If yes, I will really appreciate if you can advise me how to do this in C++ programming

Thank you

2

Is it better to use double precision in audio processing

Not always. Although the calculation almost takes the same time, double number requires a larger bandwidth and you cannot hear the difference. But there are cases where double precision is a must, e.g., a low pass/shelf at 10Hz.

If yes, I will really appreciate if you can advise me how to do this in C++ programming

Suppose we have juce::AudioBuffer<float> &buffer from processBlock() and pre-allocated juce::AudioBuffer<double> doubleBuffer, you can copy data as follows:

doubleBuffer.makeCopyOf(buffer, true); // copy data to double precision buffer
// process doubleBuffer;
buffer.makeCopyOf(doubleBuffer, true); // copy data back to buffer
1 Like
3

Hello, Zishu.
Thank you for your help

Is the source code you showed me about using AudioBuffer (float) in processBlock ?

Then, isn’t it possible use only AudioBuffer (double) in process processBlock
without copying or converting data from double to float ?

The link below talks about double version of processBlock

4

Yes it is OK if your host support that (I am not sure whether all hosts support double precision buffer input, maybe someone else can confirm this for you).

2 Likes
5

Thank you, I will search this forum about this issue.
So have a nice day and see you again, Zishu.
Take care

6

i use double precision a lot because bugs that are introduced due to float being too small can be really sneaky, so unless i need some performance optimisations i just keep things double

1 Like
7

Doubles are mostly useful for phasors and alike, but not for the audio signal itself. A simple float is just fine there.

1 Like
8

Hello, Florian
As you mentioned, I also like double precision.
I will consider double in my coding
Thank you and have a nice day, Florian

9

Hello, Michael
Thank you for your valuable information
You helped me a lot !
See you again, Michael