Correction of tri-amplified horn loudspeakers

[hulkss]

Maybe a clue here. The two misbehaving bass rerouting filters have a delay about equal to the length of noise on the front end of the step response.

 

[hulkss]

When I go back to narrower XO filters those initial delays on the filter impluses get longer.

 

[hulkss]

Here's a comparison of a good (left) and bad (right) rerouting of bass. The right is delayed for some reason?

 

[hulkss]

I played a test signal through the AL convolver and looped back the convolved signal and a timing reference.

I tested the bass offloading filters for the left and center channels. These mix to a common output going to the same subwoofer. They need to be time aligned and they are not. Here are the two signals and the vector sum with a cancellation at 75 Hz.

Here's these filters as designed and followed by the convolved filter measurement. Wow, that seems to be working great. I do not understand why the delays are reversed?

 

[hulkss]

Actually The IB + Epiks combine OK. Here is a plot of the combined Epiks (dotted magenta), the IB (dotted green), and the Combination (red) in the main listening position with no correction or gain adjustment. I'll try a two driver sub, 4 Epiks as one driver, and the IB as the other. They will sum the best that way as you suggest since they will get separate corrections.

View attachment 39380

I tried the two driver sub with separate correction for IB & Epiks. It did not work well. A lot of gain was wasted trying to flatten the IB response separately. The response in red (in the view attachment link), where both subs are driven in parallel, is much easier to correct.

 

[hulkss]

I have a work-around for the timing problem in the bass offloading filters. If I force the center to use an offloading filter from the left channel I can get everything to sync. Here's separate offloading filters and then with the center the same as left. I had to raise the XO to 120 Hz to find a spot where the center was happy with the left filter. The noise before the main step is better now at the higher XO frequency.

Here's a measurement of the L + C combined offloading filters for the two cases. Unaligned causes the big dip to form.

I would guess that the timing issue is easy to fix. The suspect XO filter "noise" i don't know.

 

[hulkss]

I just happened to notice that the time delays have changed from editing the noise removal of the measured impulse responses.

 

[jjazdk]

I tried the two driver sub with separate correction for IB & Epiks. It did not work well. A lot of gain was wasted trying to flatten the IB response separately. The response in red (in the view attachment link), where both subs are driven in parallel, is much easier to correct.

My findings, when toying with the non-registered version, is that a decent gain matching between the channels is important. Otherwise it is easy to have a lot of wasted gain.

 

[hulkss]

My findings, when toying with the non-registered version, is that a decent gain matching between the channels is important. Otherwise it is easy to have a lot of wasted gain.

That's true. When I tried the Epiks as two groups of two, they were way out done by the IB and the multi-driver sub lost a lot of gain. When I did the four together it was much better, however, then the dips in the IB response were a problem to correct. The "just a bunch of subs" approach is working out best for me.

 

[jjazdk]

That's true. When I tried the Epiks as two groups of two, they were way out done by the IB and the multi-driver sub lost a lot of gain. When I did the four together it was much better, however, then the dips in the IB response were a problem to correct. The "just a bunch of subs" approach is working out best for me.

Another good example on the difficulties of finding a "one size fits all" approach :-)

 

[hulkss]

I gave my latest correction filters a listen.

Thanks Audiolense Working pretty good now with all the bass filters in sync. This is my Min Delay, Min Phase XO solution for streaming internet content.

Before moving on to TTD, I would like to find a proper solution to get time aligned bass offloading while using unique offloading filters for Left, Right and Center channels.

 

[hulkss]

Why am I looking at all this signal detail? We need to be very careful about timing and phase in bass management. Stereo and multichannel music is produced and mastered on Digital Audio Workstations (DAWS). The engineers use DSP plugins like "Monofilter" from Nugen Audio to control and align low frequencies to be phase coherent. We need to keep bass signals correlated when they are mixed into the subwoofer from Right, Left, Center, LFE, and surround channels using digital filters.

From the Nugen Audio "Monofilter" Plugin manual:

The NUGEN Audio Monofilter is designed for the control of stereo width and phase
correlation in the low frequencies of the audio spectrum. We have paid particular
attention to the practical application of the processes, allowing for optimisation at all
stages of the recording process, tracking, mixing and mastering.

Monofilter allows for rapid and intuitive low frequency correction without unwanted
artefacts, allowing the user to quickly centre the LF power distribution toward and below
localisation frequencies, leaving stereo perception intact whilst increasing the definition
and focus of the lower frequency ranges.

Low frequencies are generally considered to be ‘omni-directional’ i.e. The ear has
difficulty in telling which direction they come from. Stereo information in these
frequencies is largely redundant, and can often contribute to a lack of clarity and
focus in the mix. Below 80Hz it is more or less impossible to determine direction
using level difference. In practical listening environments, this value may well be
much higher.

In general, natural sounding audio is phase coherent in the lower frequencies,
and as a result, we tend to perceive phase inconsistencies as ‘weak’, ‘hollow’ or
‘flat’. As soon as recorded audio is in the studio however, the chances of
introducing such artefacts are high. Adding stereo FX processing and using
multiple mic. set-ups immediately bring the issue into play.

Bass frequencies are difficult, high-energy sounds for speakers to produce; this
energy is best shared between the speakers. Sharing the load between speakers
and amplifiers allows the system to work within optimal ranges, providing clearer
audio reproduction.

 

[menlobob]

Why am I looking at all this signal detail? We need to be very careful about timing and phase in bass management. Stereo and multichannel music is produced and mastered on Digital Audio Workstations (DAWS). The engineers use DSP plugins like "Monofilter" from Nugen Audio to control and align low frequencies to be phase coherent. We need to keep bass signals correlated when they are mixed into the subwoofer from Right, Left, Center, LFE, and surround channels using digital filters.

I'm about to undertake crossovers for a 2.5 way system, where 4 drivers share the bass duties in a sealed enclosure. This insight has been very helpful.

 

[hulkss]

What would happen if my set-up had perfect speakers? Let's find out.

I looped back all the Audiolense test signals through Plogue Bidue and mixed them all into the measurement microphone channel. I inserted some adjustable delays that I can add later to simulate what happens in a more "real" system.

Here's the 14 channel set-up with 3 triamped mains, a sub, and four surrounds:

Here's the filtered measured frequency responses:

That's pretty much what I would expect to see. Now let's do a Minimum Delay correction with unique bass offloading filters for each channel.
Here's the frequency response:

Not bad, a little wiggle above the upper crossover that's at 1 kHz.

And the step responses:

All dead nuts on top of each other. Only that XO induced "noise" before the main step.

Now let's get more real-world and add delay to the three main bass channels.

Here's how Audiolense sees the delays:

Now let's do another Minimum Delay Correction and look at the correction filters. I can tell you they were all the same before the time delays.

Now the filters are "disturbed". Remember the frequency response for the drivers has not changed at all.

Now the step responses with unique bass rerouting for each speaker:

We are stepping out of sync now and the center channel is oddly "different" than the others. These out of sync signals cause cancellations when they mix as I demonstrated in earlier posts.

Let's force them all to the left channel rerouting filter.

Maybe that's better for time-aligned mixing to the subwoofer, however, the individual channels now have frequency response issues as shown below (right channel using left channel rerouting filter).

What to do? Well....I would think Audiolense could do a time alignment on those impulses. Then get on with correction as if the delays were never there.

Actually, as a work around solution, I could use the loopback set-up to time-align my speakers when Audiolense measures them. I could get all the delays equal in the measurements and I would also need to use those delays during playback. I'm not sure if it will yield a significant improvement, but I'll give it a try.

Keep in mind, corrections using TTD may not have any of these issues.

 

[hulkss]

Well... I aligned all the speakers in time within .02 ms on the first try.

The manual time alignment did not have any beneficial effect that I could find. Just because the speakers are aligned in time does not mean that they are in phase at the crossover points (like my fake perfect signals were). I suppose Audiolense has got to deal with that issue too.

 

[hulkss]

I've been working with a minimum delay correction procedure while searching for a solution where Left, Right, Center, and LFE channels play back time aligned, in phase, and with the desired frequency response. I finally have a solution!!

I'm using a setup of three four-way full-range speakers. They actually share a common subwoofer array as the bass driver (I mix the bass outputs together in my MOTU 16a). The LFE channel is mixed with the Center channel using the channel routing matrix as discussed in my subwoofer array thread.

Here is the set-up screen:

The solution I have here was basically found with trial and error. Most of the challenge was caused by the fact that there is one odd driver in the mix. The center channel "lower midrange" is different from the Left & Right (La Scala vs Khorn). Do use identical L,R,C speakers if at all possible.

By iterating XO frequencies and filter widths I found an answer that got all three speakers working together. Now this is really important at 100 Hz and below as the bass signals tend to be highly correlated between channels. Here are the step responses that really match up well now.

As absolute proof this is working, I again used REW to measure in the main listening position. Here are R, L, C individually, then R+L (perfect alignment adds 6 dB), and R + L + C where perfect alignment adds about 9 db.

I'm not sure why I could not find a solution this good in a more typical set-up with three small three-way speakers using bass rerouting to the subwoofer.
Again, TTD correction probably makes this all easy - if you can live with the latency