FD Magnitude sum

Post Reply
Lesage
Posts: 22
Joined: 07 Nov 2020 11:17

FD Magnitude sum

Post by Lesage » 03 Jun 2022 09:38

I have tried the Frequency domain function "Magnitude sum" to look at the way the left and right pulses are summing in my room, and was surprised by the result, at a first glance I thought that the vertical scale was linear, not logarithmic (in dB SPL)

At a second tought, I wondered wether there could be something bizarre, so I cheched each curve's level at the same frequency and confirmed this : the sum should be +6dB higher, not approximately the double level higher, as the curve's level should be converted in linear scale, summed, then converted again in logarithmic

Others relative results, ie phase and time curves, are impacted too
Attachments
2022-06-03 Magnitude Sum.jpg
2022-06-03 Magnitude Sum.jpg (164.6 KiB) Viewed 447 times
User avatar
UliBru
Posts: 468
Joined: 19 Oct 2019 13:58

Re: FD Magnitude sum

Post by UliBru » 03 Jun 2022 11:10

The magnitude as result of a Fourier-Transform does not contain time information anymore. Thus the sum of the magnitudes will clearly change phase and pulse response. So you can select between a linearphase or minimunphase result.
Any yes, the magnitude (= not signal amplitude!) is shown with logarithmic axis. The sum of two logarithms is equal to the multiplication of the non-logarithmic values.

The Amplitude Sum function can be very helpful for the convolution of either linearphase or minimumphase filters. A convolution in time domain corrsponds to a multiplication in frequency domain and corresponds to an addition in log frequency domain. Thus combining two minphase filters is like adding the curves in the frequency chart. This simplifies life.
Stay well tuned
Uli

Moderator
Acourate system: JRiver/Roon -> AcourateConvolver -> miniDSP U-DIO8 -> TacT M/S2150 amps -> DIY horn speakers
Lesage
Posts: 22
Joined: 07 Nov 2020 11:17

Re: L & R Magnitude sum

Post by Lesage » 05 Jun 2022 06:11

My bad, I forgot that FD Sum is purposed for chained processes only

Here are the properly summed (TD addition) L & R channels. The result is interesting (slighly L-shaped room and listening position, no ICPA)

Now I an trying to generate filters relative to the summed (uncorrected) response as to prioritize the stereo image center while sacrificing stereo accuracy

I have found how to generate corrections and filter in mono mode, how could I convert these Cor filter to stereo (ie : dual-mono) ?

Just copy and rename CorM filters to CorL and CorR ?
Attachments
20220604 Sommation G+D Out Boom.jpg
20220604 Sommation G+D Out Boom.jpg (165.44 KiB) Viewed 406 times
User avatar
UliBru
Posts: 468
Joined: 19 Oct 2019 13:58

Re: FD Magnitude sum

Post by UliBru » 05 Jun 2022 10:14

Imagine both channels are properly playing. Thus the time domain addition will lead to amplitudes increased by factor 2 = 6 dB.
Now imagine that one channel has a phase rotation of 90°at a given frequency, e.g. 200 Hz (see you curves). This means that the amplitude sum increases just by 3 dB.
Now when you create a correction by the mono sum you will lift both channels by 3 dB gain despite they are playing correctly. The proper correction would be to correct the phase.
This is a simple example that correcting the mono sum is not a good approach.

Indeed you should check if ICPA can do a better job in this case.
Stay well tuned
Uli

Moderator
Acourate system: JRiver/Roon -> AcourateConvolver -> miniDSP U-DIO8 -> TacT M/S2150 amps -> DIY horn speakers
Post Reply