Phase response and imaging / resolution

[Branwell]

Hello,

Recently I got some BESL 5s with Phil’s active alignments for the Behringer DCX crossover to use as a reference to my DIY efforts.

I have created my own alignments for those same speakers and crossover, and while I can make alignments that are flatter in FR in my room, I have not been able to make alignments that have more detail, better imaging, depth, width or focus, or a more neutral sound. In fact, all my attempts fall short by significant margins.

The question is why ?.

FR is FR, Crossover slopes are crossover slopes and if they follow acoustical targets accurately, they should be good, right ?. Will it seems not.

The area I think I am falling down on is phase. I don’t mean having the tweeter reversed or not and I don’t mean time alignment, I mean shape of the phase response for several octaves around the crossover point.

Phil and my crossovers alignments both meet acoustical FR targets with about the same accuracy, but his phase responses differ from how mine look around the crossover frequency, and unless I program in his alignments into the DCX exactly as he does, I don’t get his phase response.

The bottom line here is that I think there must be a particular phase response shape that plays a large part in having a speaker sound its best and I don’t understand it…….

Can any one suggest an online site that has information that might help me solve this dilemma ?.

Thanks.

Branwell

[phase_accurate]

The bottom line here is that I think there must be a particular phase response shape that plays a large part in having a speaker sound its best and I don’t understand it…….

The best phase-response would be flat flat-phase throughout the speaker's reproduction bandwidth. Such a behaviour would be called phase_accurate.

But that:

1) is not achieved easily
2) comes at a price (less flat amplitude response for instance).

Regards

Charles

[wimms]

http://www.aip.org/pt/nov99/locsound.html

You don't hear phase beyond 3-4KHz. But you do hear start up transient response. See attach. You'd hear the sound coming from left despite that phase difference is zero. And phase difference of 5KHz signal does not change that much.

http://www.linkwitzlab.com/speakers.htm
http://www.linkwitzlab.com/frontiers.htm

[DSP_Geek]

Originally posted by phase_accurate


The best phase-response would be flat flat-phase throughout the speaker's reproduction bandwidth. Such a behaviour would be called phase_accurate.

But that:

1) is not achieved easily
2) comes at a price (less flat amplitude response for instance).

Regards

Charles

That's true if you restrict yourself to the analog domain. If you can mess about with your signal digitally then FIRs become reasonable at the price of a moderate delay, and it's a five-finger exercise to make those absolutely linear phase.


Francois.

[phase_accurate]

By using DSP one can do fantastic things, just have a look at the measurements of a K+H studio monitor:

http://www.kleinhummel.de/produkte/o...ven/O500-f.jpg

There is even a setting that allows for completely constant group-delay even compensating for the phase-response of the reflex-tuning of the woofer ! Cumulative decay is as good as practically possible, it looks almost like a wall !

But this can't be implemented cheaply.
Analog subtractive crossovers OTOH allow quite good (i.e. still not perfect) response in time- and frequency- domain without the increased complexity of DSP.

Regards

Charles

[wigginjs]

phase:

What sort of DSP are you using to achieve that result?

[phase_accurate]

This is NOT mine unfortunately ! It is an active studio monitor by a German manufacturer (Klein + Hummel). They don't say what DSP they are using though.
But the result is less depending on the DSP itself (apart from computing power of course) than on the developer's skills IMO.

Regards

Charles

[DSP_Geek]

Originally posted by phase_accurate
This is NOT mine unfortunately ! It is an active studio monitor by a German manufacturer (Klein + Hummel). They don't say what DSP they are using though.
But the result is less depending on the DSP itself (apart from computing power of course) than on the developer's skills IMO.

Regards

Charles

You _can_ do some pretty surprising things with DSP, even past response curves - JAES has some interesting papers on low frequency distortion reduction using Volterra filters - but even with lots of computrons you can't polish a t*rd. The distortion reduction is really only good for 2nd harmonic, 3rd if you want to really crank on the CPU, but past that you're on your own. Response linearisation is only good for one spot in the room, so if your speaker has off-axis funnies they will show up in the reverberant field.

In other words, DSP for speakers is useful, but not a panacea. The designer still needs to take care and do a good job before applying the finishing touch. It's just like any other form of EQ.

Francois.

[phase_accurate]

In other words, DSP for speakers is useful, but not a panacea. The designer still needs to take care and do a good job before applying the finishing touch. It's just like any other form of EQ.

This statement correlates quite well with the comparison of the different K+H monitor models. The "digital" ones are just improved incarnations of already excellent products.

It was one of their older models by the way that convinced me of going active (a quarter of a century ago!): While I was still an apprentice we visited a large audio exhibition, covering everything from lowest- to absolute high- end. A radio station was doing live broadcasting from the site. They used an older model (new back then of course) of a K+H monitor for moderate sound reinforcement. It was the best-sounding speaker that I heard that day !!!!

Regards

Charles

[Coolin]

Originally posted by Branwell
Hello,

Recently I got some BESL 5s with Phil’s active alignments for the Behringer DCX crossover to use as a reference to my DIY efforts.

I have created my own alignments for those same speakers and crossover, and while I can make alignments that are flatter in FR in my room, I have not been able to make alignments that have more detail, better imaging, depth, width or focus, or a more neutral sound. In fact, all my attempts fall short by significant margins.

The question is why ?.

FR is FR, Crossover slopes are crossover slopes and if they follow acoustical targets accurately, they should be good, right ?. Will it seems not.

The area I think I am falling down on is phase. I don’t mean having the tweeter reversed or not and I don’t mean time alignment, I mean shape of the phase response for several octaves around the crossover point.

Phil and my crossovers alignments both meet acoustical FR targets with about the same accuracy, but his phase responses differ from how mine look around the crossover frequency, and unless I program in his alignments into the DCX exactly as he does, I don’t get his phase response.

The bottom line here is that I think there must be a particular phase response shape that plays a large part in having a speaker sound its best and I don’t understand it…….

Can any one suggest an online site that has information that might help me solve this dilemma ?.

Thanks.

Branwell

I have noticed something simular when correcting for close range measurements.
I'll have to do it over again but now differently.
We have to remember room reflections also have ALOT to do with tonality. when doing a sine freq. sweep this might not show up in the measurements like we would like.
The pshyocoacoustical influence is probably different than the sine measurement.