DigitalSignal Capture Accuracy

[Glen]

Hello,

What accuracy can we expect for DigitalSignal.Capture on SC13048? I’m seeing values ranging by up to 800ns between falling edges from a known signal. The known signal is being sent by a frequency generator and confirmed by scope. It seems like the captured values are always at least 150ns too long.

On 20260D I’m seeing values being ~30ns larger than expected, but very stable.

Any config that could be wrong to cause that or tips to improve?

Thanks,
Glen

[Dat_Tran]

Hi, we just checked,

SC13x could be ~150ns ~6.5MHz
SC20x could be ~40ns = 25MHz.

For SC13 we use PA2 to generate pwm
and capture from PA1.

Send us your code if you have different result.

PWM code:

 // sc13xxx PA2
            var pwmController = PwmController.FromName("GHIElectronics.TinyCLR.NativeApis.STM32L4.PwmController\\14");
            var pwmchannel = pwmController.OpenChannel(0);


            pwmController.SetDesiredFrequency(6000000);

            pwmchannel.SetActiveDutyCyclePercentage(0.25);

            pwmchannel.Start();

Capture code.

 var pin = GpioController.GetDefault().OpenPin(SC13048.GpioPin.PA1);

var dsig = new DigitalSignal(pin);

 dsig.OnCaptureFinished += Dsig_OnCaptureFinished;

while (true)
{
                if (dsig.CanCapture)
                {
                    dsig.Capture(1000, GpioPinEdge.FallingEdge, false);
                }

                Thread.Sleep(10);
}

private static void Dsig_OnCaptureFinished(DigitalSignal sender, double[] buffer, uint count, GpioPinValue initialState)
        {
            if (count == 0)
                return;

            var t = 0D;
            for (var i = 1; i < count; i++)
            {
                t += buffer[i];
            }

            t /= (count - 1);

            var freq = 1000000000.0 / t;

            Debug.WriteLine("freq = " + freq + " Hz, ns= " + t + ", count " + count);
        }

[Glen]

Hi Dat,

Thanks for looking into this. Results from your code are consistent with mine. When doing a capture on a 2500Hz (400us) signal with 1000 samples using SC13 I get the following results (ignoring the first value):

Average = 400237
Median = 400238
Max = 400550
Min = 400088

Any chance of increasing that clock rate for the capture (maybe even as a special mode on the CPU)? This is an awesome feature but the current rate leads to errors of up to 3-4Hz at the frequencies we need to measure.

Thanks,
Glen

[Dat_Tran]

This mode use DMA, everything is done by hardware.

Let me read more on document to see anything we can improve, but so far I don’t think we can. We give DMA a buffer and bring this buffer to user when DMA done.

On SC20 it is more accurate you mean? Is the source generate clock correctly? :)).

[Dat_Tran]

Hum, I used code above to generate PWM, change to 2500Hz.

max min is very accurate on SC13.

Can you check source generator?

[Glen]

Hi Dat,

I do see much better results on SC20.

In your example is the source / capture on the same SC13? I setup a 2.5kHz PWM on PA2 on SC13. I’m seeing a similar issue with accuracy on the scope and protocol analyzer. The output frequency is 3-5Hz off. If you were going from SC13 to SC13 maybe the PWM / Capture accuracy cancels out and that’s why you got such perfect results?

Saleae single wave:

Salese average freq:

Scope stats:

To me it seems like the PWM and Capture use a similar accuracy timer behind the scenes. Can the frequency of that timer be cranked up in the hardware config?

Regards,
Glen

[Dat_Tran]

Interesting.

Good thing to me that there is no thing we can do so far, look to me hardware issue.
Bad thing to you is, something you need to do :)).

We will spend a bit more to see what we can help.

[Glen]

Thanks Dat.

No real surprise here: but FYI this is from SC20260 PWM at 2.5kHz 2.49997kHz measured.

[Glen]

Just closing this one off… The issue was that we were testing on FEZ Flea so I’m assuming this was just a lack of xtal. It works great in production with a good xtal and on the SC13048Q dev board. Sorry for the false lead!