First readStream contains invalid samples

[Jan]

Thanks for this very interesting project!

I have been playing with Soapy and an Airspy when I noticed that the FFTs on my first samples did not always came out as expected. Looking closer I noticed that the 1st few complex samples always contained some 0s and from thereon on never again. This seems a bit strange and almost like either the first buffer has some wrong data in it or I am reading it wrong. Hopefully this is something I am doing wrong.

I allocate the buffers as such:

	vector<vector<char>> buffMem(1, vector<char>(SoapySDR::formatToSize("CF32") * mtu));
	vector<void *> buffs(1);
	for (size_t i = 0; i < 1; i++) buffs[i] = buffMem[i].data();
	int flags(0);
	long long timeNs(0);
	vector<vector<float>> sampleBuff(5 * 2048, vector<float>(2));		// Holds the IQ samples

then read a couple of sample with:

	for (int l = 0; l < 4; l++)
	{
		// Read stream
		int elementsRead = device->readStream(st, buffs.data(), mtu, flags, timeNs);

		cout << elementsRead * SoapySDR::formatToSize("CF32") << " bytes have been read. Flags = " << flags << endl;

		for (int i = 0; i < buffMem[0].capacity(); i += SoapySDR::formatToSize("CF32"))
		{
			deciCnt++;

			if (((i + 8) <= buffMem[0].capacity()) && deciCnt == 1)
			{
				// Bytes packed back into floats
				float ival;
				float qval;
				unsigned char ci[] = { buffMem[0][i + 0], buffMem[0][i + 1], buffMem[0][i + 2], buffMem[0][i + 3] };
				unsigned char cq[] = { buffMem[0][i + 4], buffMem[0][i + 5], buffMem[0][i + 6], buffMem[0][i + 7] };
				memcpy(&qval, &cq, sizeof(qval));
				memcpy(&ival, &ci, sizeof(ival));

				// Store the IQ values
				sampleBuff[sampleCount][0] = ival;
				sampleBuff[sampleCount][1] = qval;
				sampleCount++;

				deciCnt = 0;

			}
		}
	}

I am reading 2048 (MTU) elements at a time.

Output of the first few samples

0	9.75201e-007	0
1	-1.63161e-006	0
2	2.91294e-006	0
3	-5.7514e-006	0
4	1.00109e-005	0
5	-1.65721e-005	0
6	2.5934e-005	0
7	-4.02493e-005	0
8	6.21305e-005	0
9	-9.65394e-005	0
10	0.000161523	0
11	-0.000401581	0
12	-0.000173937	-0.000727539
13	-0.000135679	0.000232104
14	0.000481409	-0.000474067
15	-0.00017729	0.000708774`
...

Any idea what could be the cause of this?

As a side question is there some builtin byte packing function or is this pretty much the way everyone else also do it?

[Youssef]

The IQ conversion routine in libairspy has a delay of 12 samples.

[Jan]

Ah okay, that is an easy fix for me then. Thanks!