[sportflyer]

I applied a 6db house curve to REW. Graph vert axis 20-130 db, apply 1/3 octave smoothing . Target curve was 78.7 db

With the older REW version 4.11 , The smoothing was so good , that no filters are required !

With Version 5 REW, I ended up with 7 filters !

So what gives ?

Thanks

 

[Moonfly]

If your just talking about the smoothing, then it is applied a little differently in V5. There is apparently no set standard as to how smoothing should be applied, so it can differ between devices. I think John (REW author) mentioned that in V4 the smoothing was probably a little too smooth, and this had been addressed with V5, making it a little more accurate so to speak.

<edit>

I got it the wrong way around, John clarifies it better here:

REW V5 has some modifications to the smoothing functions (to be technical, it uses an approximation to a Gaussian kernel instead of a simple moving average) which make the graph look more smoothed. A 1/3 octave filter has a well defined frequency span over which it is applied, but there is no particular standard for the smoothing kernel which should be used. V4 was probably not as smooth as it should have been.

I hope that is the info your looking for :T

 

[sportflyer]

Are you saying that V5 is probably more accurate in the smoothing functions ie not over smoothing the curve vs ver4 so I should use the v5 filters vs the v4 . Tks

 

[Moonfly]

As I understand it yes, but I will ask John to comment as well, just to be sure :T

 

[JohnM]

Are you saying that V5 is probably more accurate in the smoothing functions ie not over smoothing the curve vs ver4 so I should use the v5 filters vs the v4 . Tks

Broadly speaking, yes, but some screenshots of the responses would help clarify.

 

[Moonfly]

Thanks for jumping in John :T

 

[sportflyer]

OK here are the curves. Both use 6db house curve with 1/3 octave smoothing.

This first chart uses REW ver 4.1 only one filter was used : 43HZ , -2db and bw 8


This second chart uses REW v 5 with 6 suggested filters


The filters were :

Freq Gain BW

28.9 7 16
43.5 -7 11
33.2 -9 8
37 5 11
53.9 5 10
61.7 -5 5

 

[salvasol]

I like the second graph better...:T

Are you trying to avoid the use of many filters??? or you're just comparing v4 vs v5 results???

 

[sportflyer]

I like the second graph better...:T

Are you trying to avoid the use of many filters??? or you're just comparing v4 vs v5 results???

Both.

I prefer to use a small number of filters. However I am also surprised that there is such a big difference in the REW's auto EQ algorithms between the 2 versions.

 

[laser188139]

If you look at the description of the changes in the REW v5 beta, John has made a lot of changes in the EQ algorithm. One thing I notice, comparing v4.11 and v5 against the same v4 sub measurement file is that v4 generated only cuts, v5 generates both cuts and boosts. The amount of boost allowed, and the desired flatness relative to the target are now under user control under the filter tasks. V5 also offers an automated Set target level function to center the level of the target before computing the equalization filters. So it makes sense that v5 would generate more filters and achieve a smoother response, because it has a greater range of choices available.

Bill

 

[Wayne A. Pflughaupt]

I applied a 6db house curve to REW. Graph vert axis 20-130 db, apply 1/3 octave smoothing . Target curve was 78.7 db

With the older REW version 4.11 , The smoothing was so good , that no filters are required !

With Version 5 REW, I ended up with 7 filters !

So what gives ?

Thanks

Can’t tell from your post if you’re manually equalizing, or letting REW do it.

Or both? Sounds like maybe you were going manual with V4 and automatic with V5.

In any event, looking at your two graphs, you can see that the second one doesn’t deviate from the Target much at all. It should be obvious that anytime response is virtually flat-line “clamped” to the Target, it’s going to require more filters.

BTW, the 20-130 axis/smoothing approach is really only for helping people not go overboard when manual equalizing. REW ignores smoothing when it performs equalization.

Regards,
Wayne

 

[sportflyer]

Can’t tell from your post if you’re manually equalizing, or letting REW do it.

Or both? Sounds like maybe you were going manual with V4 and automatic with V5.

In any event, looking at your two graphs, you can see that the second one doesn’t deviate from the Target much at all. It should be obvious that anytime response is virtually flat-line “clamped” to the Target, it’s going to require more filters.

BTW, the 20-130 axis/smoothing approach is really only for helping people not go overboard when manual equalizing. REW ignores smoothing when it performs equalization.

Regards,
Wayne

I was using auto EQ for both versions of REW . I experimented with setting the target curve with and without chart smoothing and my conclusion is that its easier to set the target curve to approx average between peaks and dips without the smoothing function. In any case the final results are pretty much the same with both methods .