FLDIGI 4.1.18 watefall fft different from previous versions


iz1fks
 

Hello, in version 4.1.18 the waterfall looks worse than in 4.4.14. Setting are the same but something different is notieable. It looks like an issue with fft averaging (active after being lost in 4.1.17) but not sure about it. It could be the averaging is perfomed on a less number of samples or no so effective.
Anyone has noticed the same issue?


73 de Phil iz1fks/ai4yn


Richard Rohrer
 

Hi Phil,

I am seeing the same thing.  My waterfall is much fuzzier that in previous versions.   

73
Dick - KC3EF


Cliff
 

Somewhere around that time Dave reduced the resolution on the waterfall to reduce the CPU load caused by the higher resolution. I expect that is what you see.

73,
Cliff, AE5ZA

On May 23, 2021, at 06:34, iz1fks <iz1fks@...> wrote:

Hello, in version 4.1.18 the waterfall looks worse than in 4.4.14. Setting are the same but something different is notieable. It looks like an issue with fft averaging (active after being lost in 4.1.17) but not sure about it. It could be the averaging is perfomed on a less number of samples or no so effective.
Anyone has noticed the same issue?


73 de Phil iz1fks/ai4yn


iz1fks
 

thanks for reply, and see that's not just an issue here. Hope it could be fix in next release and other users could notify the problem.

73 de Phil, iz1fks


Dave
 

On 5/27/21 9:28 AM, iz1fks wrote:
thanks for reply, and see that's not just an issue here. Hope it could be fix in next release and other users could notify the problem.

73 de Phil, iz1fks


iz1fks
 

thanks very much for your reply Dave. I have installed 4.1.18.40 but I cannot see difference coparing with 4.1.18.0 with same settings of course.
I'm still with 4.1.14 that has a better waterfall resolution.

I attach a screenshot with 4.1.18.40 on the left and 4.1.14 on the right, even if it's not so clear as when the waterfall is moving.

thanks and 73 de Phil,iz1fks


Dave
 

My hardware and OS are not the same as yours, but there is no reason why the latest alpha should not give similar performance on your system

  waterfall set to 2x

Lubuntu 20.1
ThinkCenter m90p computer, using sysem Intel video system
1920x1080 LG monitor

waterfall parameters set to



73, David, W1HKJ

On 6/1/21 2:39 PM, iz1fks wrote:
thanks very much for your reply Dave. I have installed 4.1.18.40 but I don't see any difference in waterfall comparing with 4.1.18.0 (same settings of course) I'm
still with 4.1.14 with better waterfall
I attach a screenshot with a comparison 4.1.18.40 (left) and 4.1.14 (right) even it's not so clear the difference as it's seeing the waterfall moving.

thanks and 73 de Phil, iz1fks


iz1fks
 

I don't know Dave, my setting were very close to yours (tried also exactly the same ones) but a difference with waterfall of 4.1.14 is clearly visibile.
I attach a screenshot with fft averaging disabled in both versions, the issue is noticeable in the same way.+
I'm using Win7 64 bit with discrete Nvidia GeForce210.

Since your screenshot shows "normal" waterfall speed, I have tested setting speed to "normal" in both versions instead of "fast", and the waterfall are much more similar, very very close. But using fast speed situation looks different.

I attach also a couple of screenshots taken with normal and fast speed setting, where the different appearance of the waterfall could be more easy to visualize.

73 de Phil, iz1fks


Dave
 

There are some small differences in waterfall processing between 4.1.14 and the current alpha 4.1.18.40.  The changes were made specifically to improve CPU performance and to make the waterfall appearance independent of modem sampling rate.  You can expect that the waterfall appearance at FAST drop down will not be as distinct as NORMAL, and NORMAL not distinct as SLOW.  This is a result of latency reduction.  The waterfall is a time/amplitude representation of the audio spectrum computed using a Fast Fourier Transform.  The sample set size for the SLOW, NORMAL and FAST drop rates vary by 4x, 2x and 1x respectively and the computation cycle changes by 1x, 2x, 4x.  The latency is selected to give a quality view in the NORMAL drop rate.



SLOW drop rate
WF expanded to 4x


NORMAL drop rate
WF expanded to 4x


FAST drop rate
WF expanded to 4x

Image is an Rx capture of a generated transmit wav file, CQ CQ CQ DE W1HKJ K, psk31 mode.

The purpose of the waterfall is to provide s visual assist to signal mode identification and acquisition.  It is not an analytical tool.  fldigi does have a signal analysis tool available, menu item "View/Spectrum scope"



Shown above for the same signal.

I can assist you with where in the code to make changes If you are interested in testing your display and cpu usage for various amount of FFT latency.  You will have to compile fldigi for your testing.  The fldigi wiki provide information on how to build from code on various operating systems and computer platforms.

73, David, W1HKJ