Clutter away. I am interested too.

15 hours ago, frednork said:

I have been looking at that spectrogram for the last 10 minutes and going to myself I have no idea what he is looking at but then suddenly I saw it. Eureka! So indeed the red one follows that dirac shape less

Yes.

 

 

If you look at the zone centered around 25ms and 1khz (give or take) .... the SPL is (slightly) least on blue (quieter reflection from moving away from surface).

In general you want the energy to be inside your "diract envelope" .... or delayed by 25ms or so.     This is arguably more important than any overall "reverberation time" ..... and arguably, actaully what you are hearing when you broadly lower RT.

 

 

15 hours ago, frednork said:

So the big reveal, Red was behind the current LP by about 50 cm, The Green was the current LP and the Blue was about 50cm forward of it.

Cool.

 

14 hours ago, frankn said:

Can you clarify the comment “and more energy after 25msec “ ?  The Dirac pulse link doesn’t show an increase in energy after a significant delay - it shows a tapering of spl/time which would be expected (the only frequency content that increased after a significant delay appears to be a very narrow frequency range - e.g an artefact). 

The dirac pulse doesn't contain any reflections.

 

A measurement of a real room contains reflections... and you want these to not arrive too early.

 

14 hours ago, frankn said:

Also, can you expand on why are you focusing on single octave of 300-600 Hz response?

Because these are the reflections/cancellations that will be notably affected by the distance to the wall .... and are pressure which aren't (as) affected by absorbers and diffusers placed in the room.

 

ie. before "treamtment", pick the listening and speaker positions carefully to optimise the "SBIR"

 

14 hours ago, frankn said:

 The bandwidth an octave or two higher than this and the octave from 175-350 Hz are worse than the Red  in one of the other positions.  I’m trying to understand your preferences. 

Below a few hundred Hz, they are roughly equivalent .... becuase the modes are becoming sparse (the room is taking over), and there is little avoiding the reflections (and cancellations) as all the surfaces in the room are "close".At this point, it's just a matter of picking a LP which is not horrible .... and placing the sources where any cancellations and rining are avoided.... and then using EQ.

 

For that region especailly it would be helpful to look at a comparison of different LP locations (and source locations) that had been EQed as best they could ..... then what would remain is any absormailities in evergy vs time....  as some of what you can see in the chart now at LF is just "the frequency responses are not equal (ie. flat)".

2 hours ago, davewantsmoore said:

If you look at the zone centered around 25ms and 1khz (give or take)

TBC I mean a bpretty big zone.... like 20 to 40ms and 400 to 1200hz... or similar.

 

You can of course use pressure traps to target the narrow modes. Depends on what’s socially acceptable and practical in the design (size/placement) within the room. 
The worlds best DSP is of course in your head. Your brain will override the signals your senses provide. 

In hearing, the precedence effect when listening to music can be large and may well lessen the perceived effect of DSP. I know there will be a difference but will it be a large as expected based on the before/after measurements because your brain doesn’t process & interpret the sound in the same way as a microphone records it. 
I like the idea of using DSP with as light a touch as possible. 

@davewantsmoore in the picture you linked to the spurious amplitudes I mentioned are loosely circled. 
That doesn’t look like it is correct (assuming the sampling used ensures there is no aliasing) Maybe a coding issue?  

12 hours ago, frankn said:

You can of course use pressure traps to target the narrow modes.

 

They're quite difficult to design (lots of "build, measure, repeat" required), and need to cover a signficant portion of the surface in question.    Sure, they're basically (except for modifying the source properties) the only way to address the issue ... but optimising the placement of everything is the thing to agressively persue first.

 

12 hours ago, frankn said:

I like the idea of using DSP with as light a touch as possible. 

It's important to consider what errors are being corrected to define "light touch".    It's right to think that DSP isn't very appropriate to fix some errors...... but a very generalised "don't use much of it" tends to get people irrationally "afraid" of it.

 

11 hours ago, frankn said:

@davewantsmoore in the picture you linked to the spurious amplitudes I mentioned are loosely circled. 
That doesn’t look like it is correct (assuming the sampling used ensures there is no aliasing) Maybe a coding issue?  

Sorry, I possible wasn't very clear sorry.  The picture posted is of a real measurement.... that just happens to look a relatively close the shape of the spectrogram I was reffering to (as in all the energy would be concerntrated in that main envelope).... it was just the closest pic I happened to find in google images as I was away from my audio tools.

 

Yes, an actual "dirac pulse" would look "perfect".

eg. https://www.roomeqwizard.com/help/images/spectrogramidealwavelet.jpg

1 hour ago, davewantsmoore said:

The picture posted is of a real measurement....

Is that pic  anechoic?

 

if so any pics of a good non anechoic response.

12 minutes ago, frednork said:

Is that pic  anechoic?

I checked the source site, and it didn't say, but the inferrence was no.... but if you look at the scale, resolution, and windowing I would expect it's not the same as yours, so it is difficult to compare.... and we don't even know what distance the measurement was taken from.

1 hour ago, frednork said:

Is that pic  anechoic?

if so any pics of a good non anechoic response.

 

It depends on what you call good.... I think it is generally a mistake to just look at low delayed energy (RT) as good... but rather look at very specific (early) reflections, and very specific attention to the initial "direct" response and the speaker directivity (hard to evaluate in room).

 

Here's a members response that I measured with a view to helping out the speaker setup.    It's a highly! treated room, with ~200ms RT values .... but the wonks in the frequency response, and the fairly early strong reflections (there were surfaces closeish to the speakers) were fairly audible.    The time alignment was also borked (XOs at 2khz, 400hz, 60hz).    I would focus on specific things in the 5 to 25ms range, as oppposed to saying "look how all the energy is gone after 50ms, that's great".

 

 

5 hours ago, davewantsmoore said:

 

It's important to consider what errors are being corrected to define "light touch".  

……….

Thanks for your comments - perhaps a more accurate comment is informed (guided) application of wave-shaping DSP.  
The use of DSP to separate primary (direct) signal from noise (reflections, or multiples of the direct signal) has been around for decades beginning with telephony, radar, sonar, hydrography, astronomy/electromagnetic spectrum imaging, medicine (X-Ray, MRI), photography, optics,  seismology, audiology etc. and companies like Dirac are now using similar algorithms developed with learnings from the aforementioned combined with growing understanding of neuroscience in human processing of hearing. 
 

I like what Dirac say they are doing but there isn’t much detailed technical information on their website. 
 

My involvement in DSP began back to the 1990s as a Geophysist.  I have lead technical teams in the acquisition of seismic data, the signal processing, and modeling (acoustic, full-waveform etc.) and interpretation of seismic data to locate oil & gas drilling rigs. I have published on integrating modeling, recorded seismic and using seismic data derivatives for fluid prediction so I have more than a passing interest in DSP.  I retired a couple of years back and am now enjoying using my background to improve my future use of DSP in home audio. 

 

BTW - “Dirac pulse” is marketing a marketing slogan.

Edited March 2, 20233 yr by frankn

 

42 minutes ago, frankn said:

I like what Dirac say they are doing but there isn’t much detailed technical information on their website. 

Yes.   I suspect it is simpler (ie. less super secret magic) than many expect .... effectively optimising both the balance between linear and minimum phase (commonly termed "mixed phase") to fill in the gaps.... allowing much more effective "brute forcing" of the frequencyh response in the listening position(s).

 

42 minutes ago, frankn said:

BTW - “Dirac pulse” is marketing a marketing slogan.

https://en.wikipedia.org/wiki/Dirac_delta_function

?

 

42 minutes ago, frankn said:

My involvement in DSP began back to the 1990s as a Geophysist.

Nice.

My recent career change has me working with some basic DSP outside of audio too  .... spectral analysis of very very noisy time series'... eg. MESA (aka MESE?) and a lot of the stuff apprears to be first used in geology and millitary.

Edited March 2, 20233 yr by davewantsmoore

Delta-function - agreed. 
pulse - marketing. 
might be me being pedantic. 
From some comments I think that to reduce noise Dirac may be applying a auto-correlation & deconvolution (https://en.m.wikipedia.org/wiki/Deconvolution) which is a staple in time-series analysis and signal processing. 

38 minutes ago, frankn said:

pulse - marketing. 
might be me being pedantic. 

Yes, I think so.   

 

Quote

 

In mathematics, the Dirac delta distribution, also known as the unit impulse

 

 

@davewantsmoorei thought you were using a phrase from Dirac software blurb!  😀

51 minutes ago, frankn said:

@davewantsmoorei thought you were using a phrase from Dirac software blurb!  😀

LOL, right. No.

 

I just mean "what the spectrogram would look like if you made one on a (more) perfect input signal" (which in the extreme would be just a pulse)