Re: phase measurement

[Karen young]

Hi Daniel,

Thank you so much for all the info, very useful.  

I tried the Keep M in N block, straightforward and works great! Now I also prefer using this block.

Thanks again

Karen

On Wed, Dec 28, 2022 at 2:39 PM Daniel Estévez <daniel@destevez.net> wrote:

Hi Karen,

Good you got this working.

I now see that something I said is wrong. I said: to get FFT bin 128, we
should use Skip Head 127. This is not correct. We need to skip FFT bins
0 through 127 (both included), which is a total of 128 samples, so we
should use Skip Head 128.

There is another thing I was misunderstanding. The Keep 1 in N block
keeps the last element out of every N, rather than the first element of
every N, which is what I expected. This explains why you need to use
Skip Head 129 rather than Skip Head 129 as the above paragraph would
seem to indicate.

Since I find this a bit confusing, I prefer to use the Keep M in N
block, which keeps the first M elements out of every N, starting with an
initial offset. I find it more straightforward to think in terms of
keeping the first element than in terms of keeping the last element.
Here, using M = 1, N = 1024, and an initial offset of 128 will get you
the FFT bin you want.

Best,
Daniel.

On 28/12/2022 16:31, Karen young wrote:
> Hi Daniel,
>
> Thanks a lot. This is so helpful. I tried both methods you suggested.
> The phase is correct now.
>
> For "Shift: No", compared to the Skip Head value 127,  value 129 has the
> max magnitude. I guess it is a zero-based system. Please correct me if
> wrong.
>
> Thanks and best regards
> Karen
>
>
>
> On Tue, Dec 27, 2022 at 12:42 PM Daniel Estévez <daniel@destevez.net
> <mailto:daniel@destevez.net>> wrote:
>
>     On 27/12/2022 17:49, Karen young wrote:
>      > Hi,
>      > Attached is the model I built for phase measurement.
>      > Since the samp_rate (5M) is 8 times of signal source freq
>     (625k),  the
>      > phase should be expected as shown in the table, while the
>     measured phase
>      > is only correct when there is half cycle or full cycle delay.
>      > What can I do to make it correct?
>      > Delay step    Expected phase delay    Measured phase delay
>      > 0     0       0
>      > 1     -0.79   -1.5
>      > 2     -1.57   -4.7
>      > 3     -2.36   -1.5
>      > 4     -3.14   -3.14
>      > 5     -3.93   -1.57
>      > 6     -4.71   -1.57
>      > 7     -5.5    -1.57
>      > 8     0       0
>
>     Hi Karen,
>
>     I think that the problem might be with your Skip Head block. I see that,
>     in combination with the Keep 1 in N block, you're using it to select the
>     correct FFT bin in the vectors output by the FFT blocks. However, it
>     seems to me that you're not selecting the correct bin.
>
>     Since you're using "Shift: Yes" in your FFTs, the DC bin would be at
>     index 512. The bin corresponding to the positive frequency of the cosine
>     (you're using a real cosine, so it has a positive and a negative
>     frequency component, if that makes sense), would be at index 512 +
>     1024/8 = 640. Therefore, I think it's necessary to use 639 as the number
>     of items to skip.
>
>     If you used "Shift: No" in the FFTs, the the positive frequency of the
>     cosine would be at index 1024/8 = 128, so the number of items to skip
>     should be 127 rather than 129.
>
>     To make sure that you're selecting the correct FFT bin, I suggest you
>     look at its magnitude and compare it with a plot of the magnitudes of
>     all the FFT bins. This will show if you're in fact selecting the correct
>     bin where the maximum of the FFT is, or something else.
>
>     Best,
>     Daniel.
>
>