New to my system this year was a DSP and building a set of speakers that would be active XO'd. I do not have a space to tune in free space/anechoic so room effects are very real when running sweeps. The LF can be measured relatively decently near field, but the HF measurements dont seem to paint a clear picture of what's going on.
After many hours of both learning curve with tuning coaxial drivers with CDs I have gone between doing sweeps in the LP as well as tuning based on rta and white noise. It seems that tuning by sweeps in the LP and even when doing a couple positions near the LP room effects start throwing you down the rabbit hole.
In more recent tuning/refinement I have found that running RTA with white noise is quite effective as you can "wave" the mic around the seating position and fill in nulls or squash peaks that would have been very present in the static sweep measurement. This seems very hackish and non scientific but seems quite effective in getting good sound and not taking a long time measuring in many locations for each adjustment. Previous to this I had some shrill highs which was caused by eq overcompensation from 3 position sweeps. Another area I found great value in this method is fixing time alignments. Initially time aligning with fixed frequencies only aligns that frequency, watching the spectrum change around the x-over freq gets an overall better match across the overlaps.
Does anyone else use this as one of their tools/methods (not saying this replaces sweeps) for fine tuning in their XO and EQ?
After many hours of both learning curve with tuning coaxial drivers with CDs I have gone between doing sweeps in the LP as well as tuning based on rta and white noise. It seems that tuning by sweeps in the LP and even when doing a couple positions near the LP room effects start throwing you down the rabbit hole.
In more recent tuning/refinement I have found that running RTA with white noise is quite effective as you can "wave" the mic around the seating position and fill in nulls or squash peaks that would have been very present in the static sweep measurement. This seems very hackish and non scientific but seems quite effective in getting good sound and not taking a long time measuring in many locations for each adjustment. Previous to this I had some shrill highs which was caused by eq overcompensation from 3 position sweeps. Another area I found great value in this method is fixing time alignments. Initially time aligning with fixed frequencies only aligns that frequency, watching the spectrum change around the x-over freq gets an overall better match across the overlaps.
Does anyone else use this as one of their tools/methods (not saying this replaces sweeps) for fine tuning in their XO and EQ?
Hi,
Yes interpretation of measurements is not easy... i incorporated an averaging method in my measurement 'routine', like you did:
https://www.diyaudio.com/community/...and-blog-of-jean-luc-ohl.370843/#post-6620174
Yes interpretation of measurements is not easy... i incorporated an averaging method in my measurement 'routine', like you did:
https://www.diyaudio.com/community/...and-blog-of-jean-luc-ohl.370843/#post-6620174
That definately reassuring I wasnt out to left field. The static and averaged plots are exactly characteristic of what I've seen.
Curious as to why they prescribe using pink noise and comparing against its measured FR instead of white noise that is supposed to be even across the board?
http://www.ohl.to/audio/downloads/MMM-moving-mic-measurement.pdf
Curious as to why they prescribe using pink noise and comparing against its measured FR instead of white noise that is supposed to be even across the board?
http://www.ohl.to/audio/downloads/MMM-moving-mic-measurement.pdf
Last edited:
I don't know what this means without reading the paper. Speakers aren't necessarily best with a flat response.
Comparing the in room to the anechoic behaviour of the speaker would allow some information on the room contribution. Taking this at different positions allows the effects to be averaged for a better representation of power. It is still not enough because there are other aspects of the room contributions (besides their level) that cause a problem.
Spatial averaging is often recommended for lower frequency work. Designing a speaker for reduced room interaction is often recommended for higher frequencies.
Hi Allen,
I agree, it's not a cure for all disease. But i find helpful to have another pov than single static measurements. Sure it doesn't take away early reflections and other nastys from room but i found it better than nothing.
Garboui have you tried FDW on your single position mic measurements? It usually gives good results. And some members here got spectacular results using single position measurements ( verified through multi measurements on an area...).
I don't get your comment on white noise and fr ( please apologize, my english is a bit rusted as i've not visited this place regularly for the previous 3 months).
Could you clarify and gives reference to the doc you linked please?
I agree, it's not a cure for all disease. But i find helpful to have another pov than single static measurements. Sure it doesn't take away early reflections and other nastys from room but i found it better than nothing.
Garboui have you tried FDW on your single position mic measurements? It usually gives good results. And some members here got spectacular results using single position measurements ( verified through multi measurements on an area...).
I don't get your comment on white noise and fr ( please apologize, my english is a bit rusted as i've not visited this place regularly for the previous 3 months).
Could you clarify and gives reference to the doc you linked please?
I have played with FDW but its result at LP is not much more useful/ representative (also congruent with its ref in the paper). At near field or measurements at a few feet FDW does improve the measurement data.
I corrected in the post with the noise used for measurement. FR = frequency response.
I use LP measurements too, for many reasons. Sometimes just as a figure of merit for a design, positioning and other things, sometimes as a way of managing the room.. but preferrably in the context of other measurements. I understand that some people are in the position that LP measurements need to serve as more than just that.
Ok, makes more sense to me!
Pink noise is closer to what our brain is 'looking for' than white noise. In fact when mixing you could make a first try by just comparing sounds to pink noise and make them equal loudness ( pink and sound!). It'll comes out right. ( i told a dirty trîck mixers use... i'll be banned of the community )It's one of the reason you'll see 'target curve' or 'house curves' with a downward tilt in hi freq ( B&K curves for example).
Pink noise is closer to what our brain is 'looking for' than white noise. In fact when mixing you could make a first try by just comparing sounds to pink noise and make them equal loudness ( pink and sound!). It'll comes out right. ( i told a dirty trîck mixers use... i'll be banned of the community )It's one of the reason you'll see 'target curve' or 'house curves' with a downward tilt in hi freq ( B&K curves for example).
Last edited:
I already assumed that much 🙂 I wasn't sure what you meant about comparing it to the response of the speaker or about it needing to be flat.
Oh. I meant that you would take the reference pink noise curve and compare that to the measured speaker response. Ignorant to possible other technical reasons that white noise would be easier as you are inspecting for a "flat" line.
Flat line on a RTA. Flat line on loudspeaker is ( for me) an intermediate ( but mandatory) step: our brains are not RTA, and most of us will prefer a downward tilt... 😉
I've been convinced of it's need as a former audioengineer, as i could easily compare (real) instruments to their recordings ( the ones i've made).
I've been convinced of it's need as a former audioengineer, as i could easily compare (real) instruments to their recordings ( the ones i've made).
Last edited:
Agree that tilt is where my preference lies, backed by measuring other speakers I like and where things end up. Maybe straight or linear response is a better term.
White and pink are both straight lines (on a logarithmic scale), but the ideal curve may not even be straight (my findings are that it isn't straight). I'm suggesting that either could be used as a kind of design reference but neither could be used as a final target. This makes the comparison moot in a sense.
Some prefer without this compensation of high end ( i call 'Tilt' cause historically it was a tool (eq) used at mastering (on lathe iirc), i think some 70's gear implemented it too as tone controls ( Quad maybe i can't remember? Anyway makes more sense to me that bass/high...).
I agree it can be a linear attenuation ( in fact i implement a linear tilt from1khz and up to -6db at something like 20 or so... courtesy of B.Katz. )
Pink noise is typically used because a standard RTA is an octave-based measurement with limited data points on the display, which means the energy per band would increase with each octave higher if the source signal was flat (like white noise), since the bands have more frequencies in them as they get higher. Pink noise uses the appropriate spectral tilt so that a flat-measurement on an impulse-based system also measures flat on the typical 1/3 octave (or other octave-based bandwidth) RTA.
You're basically compensating for the particular type of display normally used for an RTA by including the proper spectral weighting in the signal.
https://en.wikipedia.org/wiki/Pink_noise
"Pink noise or 1⁄f noise is a signal or process with a frequency spectrum such that the power spectral density (power per frequency interval) is inversely proportional to the frequency of the signal. In pink noise, each octave interval (halving or doubling in frequency) carries an equal amount of noise energy."
https://www.softdb.com/blog/what-is-white-noise/
https://en.wikipedia.org/wiki/Colors_of_noise
"Simulated power spectral densities as a function of frequency for various colors of noise (violet, blue, white, pink, brown/red). The power spectral densities are arbitrarily normalized such that the value of the spectra are approximately equivalent near 1 kHz. Note the slope of the power spectral density for each spectrum provides the context for the respective electromagnetic/color analogy."
You're basically compensating for the particular type of display normally used for an RTA by including the proper spectral weighting in the signal.
https://en.wikipedia.org/wiki/Pink_noise
"Pink noise or 1⁄f noise is a signal or process with a frequency spectrum such that the power spectral density (power per frequency interval) is inversely proportional to the frequency of the signal. In pink noise, each octave interval (halving or doubling in frequency) carries an equal amount of noise energy."
https://www.softdb.com/blog/what-is-white-noise/
https://en.wikipedia.org/wiki/Colors_of_noise
"Simulated power spectral densities as a function of frequency for various colors of noise (violet, blue, white, pink, brown/red). The power spectral densities are arbitrarily normalized such that the value of the spectra are approximately equivalent near 1 kHz. Note the slope of the power spectral density for each spectrum provides the context for the respective electromagnetic/color analogy."
Last edited:
Allen i think we agree. But if you have to choose between white and pink as a reference better use pink in my view, it's closer to what 'sounds right' in my findings.
But it's an opinion amongst others.
Allen, at which spl reference have you made your experiments?
I hoover between 75dbc and 85dbc for the last twenty years @ home ( differs from rooms to rooms ). Right now i'm at 82dbc ( lowered the level 6 month ago from 85).
My own find were at 83dbc.
But it's an opinion amongst others.
Allen, at which spl reference have you made your experiments?
I hoover between 75dbc and 85dbc for the last twenty years @ home ( differs from rooms to rooms ). Right now i'm at 82dbc ( lowered the level 6 month ago from 85).
My own find were at 83dbc.
Last edited:
Yes, this is how i understand it.
Mattstat is plain right.
Please apologize if i induced ambiguity in my bad wording.
Mattstat is plain right.
Please apologize if i induced ambiguity in my bad wording.