DSP XO > Minimum Phase?

[amo]

Hi- Can soomeone tell me what the following passage from SL's website means? I understand everything besides the Minimum Phase part. Also, does this hold true even for the new high end units, such as the ones sold on e-speakers.com (DEQX)?

"I doubt that off-the-shelf DSP based crossovers have enough flexibility to duplicate the three outputs of the ORION ASP in magnitude, phase, distortion and noise level. It certainly is not just a matter of adjusting for similar magnitude response since the ASP is not minimum-phase"

[Davey]

DSP crossovers may or may not be linear-phase depending upon how they're implemented. The ASP (analog signal processor) used for the Orion system is definitely not linear phase, and neither are the vast majority of analog constructed crossover networks (passive or active.)

The real question is whether or not these non-linear phase characteristics are audible in typical usage for crossover networks. Conceptually, and marketing-wise "linear-phase" certainly sounds good though.

There are some excellent online references for filter theory and DSP crossover implementation if you want to read further.

Cheers,

Davey.

[catapult]

Davey, I believe in SL's quote, he said "minimum phase", not "linear phase." His analog implementation isn't minimum phase because of the allpass delay networks. But then the inexpensive digital boxes, while they use minimum phase filters, have the ability to directly add time delays.

I still don't understand SL's point except as it relates to S/N ratio. If you stack digital filters to build a measured LR4 acoustic response from the drivers, and set the delays for the deepest reverse nulls, the phase stuff should take care of itself. The S/N concerns are valid though if you have to do too much boosting or cutting with the filters.

[Davey]

Well, I may be wrong but I thought "minimum-phase" and "linear-phase" were synonymous when used in the context of an acoustic system. (A constant group delay characteristic.)

SL's analog crossover networks would not be considered "minimum phase" even if they didn't use the all-pass delay networks because the LR filters themselves don't exhibit this characteristic.

Back to the original question. If you were to try and program a DSP-based (FIR) commercial crossover with the appropriate curves and delays to mimic the Orion crossover it still wouldn't exhibit the same phase change characteristic as the Orion ASP. That's not to say you couldn't achieve an acceptable outcome, but you'd have to approach it in a different way than SL did.

However, as you mentioned the S/N concerns definitely remain. The large boost required for the woofer network effectively shifts the dynamic range of the unit downward by a like amount. This is really not the optimal way to use a DSP crossover.

Cheers,

Davey.

[catapult]

SL's analog crossover networks would not be considered "minimum phase" even if they didn't use the all-pass delay networks because the LR filters themselves don't exhibit this characteristic.

Are you sure about that? I believe a conventional LR filter's phase response can be predicted mathematically from the amplitude response, implying that it is minimum phase.

If you were to try and program a DSP-based (FIR) commercial crossover

I think the inexpensive units use IIR filters which copy the phase and group delay characteristics of analog filters. FIR filters with many taps, like on a TacT, can be made linear phase - no phase shift as you change the amplitude response - but you won't find that kind of computer horsepower in the inexpensive boxes.

Screenshot from a QSC DSP-4 showing amplitude, phase and group delay of an LR4 lowpass.

[amo]

Originally posted by Davey
Well, I may be wrong but I thought "minimum-phase" and "linear-phase" were synonymous when used in the context of an acoustic system. (A constant group delay characteristic.)

SL's analog crossover networks would not be considered "minimum phase" even if they didn't use the all-pass delay networks because the LR filters themselves don't exhibit this characteristic.

Back to the original question. If you were to try and program a DSP-based (FIR) commercial crossover with the appropriate curves and delays to mimic the Orion crossover it still wouldn't exhibit the same phase change characteristic as the Orion ASP. That's not to say you couldn't achieve an acceptable outcome, but you'd have to approach it in a different way than SL did.

However, as you mentioned the S/N concerns definitely remain. The large boost required for the woofer network effectively shifts the dynamic range of the unit downward by a like amount. This is really not the optimal way to use a DSP crossover.

Cheers,

Davey.

So are you saying that a dipole speaker should not be implemented with a DSP crossover, no matter how beefy or high end, becuase it will simply introduce too much noise (during the 6 dB/oct dipole boost)? Are you saying that it is far cleaner to construct the speaker, measue it extensively, and add networks in front of the amplifyer and refine the system using iteration? I was hoping to get a high end DSP crossover or dedicate a PC and adjust phase, frequency and timing using the software. I was planning to do all this prior to the initial D/A conversion, and use a stand alone DAC for this job. Any ideas?

Thanks!

[Davey]

Catapult,

Run a simulation and sum the outputs and check the group delay on a 4th-order LR crossover and on a 1st-order crossover and note the differences.
A filters phase response can be predicted from the amplitude response, but this doesn't imply linear or minimum phase.

Amo,

No, I'm not saying that, but the Linkwitz Orion represents an extreme case because of the amount of dipole/Q correction required for the woofers. Using a commercial DSP crossover with a two channel volume control upstream of the unit means the nominal signal levels will be close to the lower end of the dynamic range. A multi-channel volume control downstream of the unit allows higher internal levels but risks clipping because of the large EQ boosts programmed. Neither is acceptable in my opinion.

Your project requirements might be less stringent and it could work pretty well.

Cheers,

Davey.

[catapult]

Linear phase 1

Linear phase 2

Minimum phase

[mhelin]

Minimum phase ~ minimum delay