Repeatedly playing contents of QByteArray through QAudioOutput

[mrjj]

My best guess is that it dont like the format but
I cannot see why not.
Also
http://stackoverflow.com/questions/32049950/realtime-streaming-with-qaudiooutput-qt
which seems same code, it seems it plays for him.
And you seem to check errors etc so Im out of suggestions for now :(

[nulluse]

I am somewhat suspicious of the repeated starting of the output from the same stream.
Should there be something to re-wind the stream to the beginning? Like this:

        else{ // restart from scratch
            s->resetStatus();
            audio->start(s->device());
        }

which does not actually help, I am just posting this as a question.
That was one of the reasons I wanted to create a local variable for the stream in both places: for initial start and re-start.

Or perhaps do I need to create a completely separate stream for reading from the buffer? I am getting impression that the stream is either at the end after writing data into the buffer, or when I want to re-start.

[mrjj]

Hi
I wondered the same.
Like it plays once and then are at the end.
You could maybe put sample generation in a function and create a new sample
at restart. mostly for test.

[nulluse]

No luck! I tried to add this to the state handler:

        else{ // restart from scratch
            delete buf;
            delete s;

            buf = new QByteArray();
            s = new QDataStream(buf, QIODevice::ReadWrite);

            for(float ii=0.0f; ii<360.0f; ii+=(360.0f*1000.0f/af.sampleRate())){
                int sample = ((int)(qSin(qDegreesToRadians(ii)) * 32768));
                (*s) << sample;
        //        char sample = (char)(qSin(qDegreesToRadians(ii)) * 128);
        //        buf->append(sample);
        //        qDebug() << (int)sample;
            }

            audio->start(s->device());
        }
        break;

Still no sound.
Also tried completely removing re-start and simply repeating the wave 1000 times to get 1 second of sound, but to no avail:

for(int i=0; i<1000; i++){
    for(float ii=0.0f; ii<360.0f; ii+=(360.0f*1000.0f/af.sampleRate())){
        int sample = ((int)(qSin(qDegreesToRadians(ii)) * 32768));
        (*s) << sample;
//        char sample = (char)(qSin(qDegreesToRadians(ii)) * 128);
//        buf->append(sample);
        qDebug() << (int)sample;
    }
}

It saved a 180,000 byte file which plays a nearly perfect tone when imported into Audacity (as far as I can tell with the cheap speakers).

[mrjj]

Ok so data is good
but it seems it dont like. I wonder if some of the format settings
is wrong but I really cant spot it.

Is it possible for me to have the project to play with ?

[nulluse]

@mrjj
You can grab it from here: qtcreator project
Thank you for your time!

[SGaist]

Hi,

I'd recommend taking a look at the spectrum example in the example of the QtMultimedia module. There's a small tone generator that should get you started.

[nulluse]

This is it, that project set me onto the right track:

s->device()->close();
s->device()->open(QIODevice::ReadOnly);

before starting playback fixed the issue! So that really was the stream at the end.
Super, thanks a lot!

Development is not over, as I suspect that closing and re-opening the device is time consuming and would cause jerky sound, but at least it is clear now why it was not working.

For now when I try to play the buffer with 1 sine wave repeatedly, using s->device()->seek(0); before restarting audio, the speakers are just clicking fast instead of playing a sine wave. This approach may be entirely unsustainable due to overhead involved in restarting both IO device and Audio output.

[mrjj]

Super
I tried with
http://doc.qt.io/qt-5/qbuffer.html#details
but had no luck.

[nulluse]

Overhead of seeking to 0 and re-starting audio must be huge!
When I am playing 1 second of 1000 waves and re-starting, there is about 1/10 second delay between the blocks.
This is on i7-2600 3.4 GHz machine. I will need to change the approach completely, but at least you two got me onto the right track - appreciate your help!

[mrjj]

@nulluse
You are most welcome :)
Its pretty huge delay.
Im surprised seek is that expensive.

funny, i have i7-2600 also. Old now but still a fine CPU

[mrjj]

Note
Adding just
s->device()->seek(0);
made it play for me.

[mrjj]

I was wondering if we subclass
QIODevice and implement
qint64 QIODevice::readData(char * data, qint64 maxSize)
If we then would endless supply data and avoid the
delay as we never restart then.
I have never subclassed QIODevice so not sure what minimum
override is.