Problem Graphing Waterfalls

[altenburgpj]

I found a bug in how REW displays Waterfall plots . I am using REW V4.11 Build 1086. What I discovered is an error in the scaling of the waterfall data as shown in the following pictures.

This first waterfall plot correctly shows a dip near 8.4kHz


This waterfall plot was generated by simply changing from Log to Linear scale. The X-axis is zoomed all the way out. As you can see the dip has moved to a point near 7.4kHz.


This waterfall plot was generated by zooming the X-axis in one time. As you can see the cursor now reads the dip to be close to the original position near 8.4kHz.


This final plot shows what happens when I slide the horizontal scroll bar to the left. The graph has changed significantly. Not to mention the reading on the cursor. No other changes were made except to slide the scroll bar.


There is some sort of bug in the graphing function.

[brucek]

Yeah, the LIN mode for waterfalls has a known issue with the waterfall plot where it's truncated when the frequency axis is linear. You should use the LOG mode instead.

As far as your frequency reading being incorrect, note that the cursor frequency and value readouts correspond to the frontmost slice - not the rear.

BTW, waterfalls, are really only useful up to about 200Hz.

brucek

[altenburgpj]

The Linear mode is very useful for locating harmonics, or modes, as both will show at exact multiples (equal spacing in Linear mode). It needs to work.

The frequency of the cursor is the same for all the slices. I was not looking at the levels.

Why, does the math behind the generation of time shifted windowed frequency response plots no longer apply at mid and high frequencies? Waterfall plots work at all frequencies as long as the window size and time shift are properly set to reveal the information.

[brucek]

It needs to work.

As I said, it's a bug in this version. It will be repaired in the next version.

The frequency of the cursor is the same for all the slices.

I don't believe that's correct. The cursor is placed on the front slice and the frequency is the same in LIN and LOG.

Why, does the math behind the generation of time shifted windowed frequency response plots no longer apply at mid and high frequencies? Waterfall plots work at all frequencies as long as the window size and time shift are properly set to reveal the information.

Waterfalls are just not useful or meaningful outside the modal region (~>250Hz). Basically, waterfall show what happens after the direct sound stops and so show the response of the room modes. Outside the modal bandwidth, Energy Time plots are used, or if you have a large enough room you can use RT60. REW offers both.

brucek

[altenburgpj]

So there is no way to use a waterfall to see the ringing modes in a tweeter diaphragm that would decay within the first 5 to 10 msec? Your assumption is that I am not using the software where there are no modes from the room, and that I am not windowing so tightly as to see only the high frequency modes I need to see. The software may have been designed for use in room correction, but that is not all it is capable of. ETCs are useful too, but only to know when the energy arrived at the mic and compare it to the initial pulse. Great for locating in time (distance) reflections, and time aligning drivers in a multiway system, but would not help much on the above listed tweeter testing. I did not see any mention of this graphing problem in the readme file. Is it discussed on the forum some place?

[JohnM]

Hello Philip,

As Bruce said, the waterfall linear display bug has been fixed for REW V5. It is not mentioned in the readme file because I didn't know the bug was there at the time! That's kind of the nature of bugs

The current 50ms min window width is too long to be of much use in examining decay of high frequency resonances, V5 allows window widths down to 1ms. It will be several months before V5 is released though.

[altenburgpj]

Thats great. I was going to ask about tightening the window in the waterfall plots.

I have a few suggestions for the new version. Would I post them, or send them to you?

[JohnM]

Either is fine.