Trinnov Altitude / JBL SDP-75

[Olli]

Phase alignement between speakers IS a BM feature.

sure. I was just confused about no subs.

 

[Dimifoot]

cool! will try instantly - where can you set the correction levels?

Optimizer settings-settings-excursion curve

 

[Olli]

Optimizer settings-settings-excursion curve

Thanks, just found it myself.

 

[Olli]

The adjustments look very promising, especially if you dial up cycles & octave as well. I have zero problems going up to 600 ms, does the Amethyst use a lower grade PC than the Altitude?

Inspired by Mitchos post on number of taps,and his and Olli's Trinnon/Audiolense/JBL/ FR plots i tried to check the effects of number of taps (or FIR msec length) in my Trinnov Amethyst.

1. First I checked the predicted(Trinnov) and actual results (REW) with the standard settings, and found a quite good match.=upper plots
2. Then I Increased number of FIR taps from 4800 to 14400 by increasing FIR length from 100 to 300ms ( Trinnov says 300 is OK, and that 600 will crash), and increased number of IIR filters from 10 to 30 and increased max Hz from 150 to 200. I also allowed more negative correction( -10 to -15db).
The results seem quite good, (bottom 2 quadrants) when I have the opportunity I will measure the results in REW, but for now I have to trust the predictions, sorry for the busy graph, but I like to make notes like this to keep track.

View attachment 62666

zoomed prediction
View attachment 62667

 

[DonH56]

In their webinars and literature Trinnov claims good agreement from predicted corrected response to real-world measurements; very nice to see that verified!

I'm curious the impact of such long FIR lengths for small rooms. Looked at one way, 300 ms is the equivalent of about 338 feet, or 1/0.3 s = 3.333 Hz. Seems like decay times in small rooms would not be so long, or is it that some are so live that reverberation times are long? It does not look like there is a large change in predicted "after" response form the longer length. @mitchco ?

Increasing the number of filters makes sense to me if they are needed or depending on how much you want to flatten the response. My room has some large nulls, mainly from the front subs, and some secondary reflections especially from the center that are likely the couch and perhaps the big TV screen behind it. I have never been able to do much about the nulls from the front subs despite moving them as much as possible; the solution was to add another pair in the rear.

I'd already set the global variable to allow an extra couple dB of attenuation but my main problem frequencies are room modes that create big nulls. I could up the boost, but all the EQ in the world won't fix those, and would lead to poor response everywhere else. My MLP is constrained by room geometry including a door; my original plans for nice primed dimensions were tossed (along with about 1/3 of my floor area) when we added a bedroom for my younger son (a worthwhile trade, but now that he's out of the house I sometimes wish I had all that extra volume back).

I have decided to not get too anal about it just yet. Too easy to spend all my free time (extremely limited) tweaking instead of enjoying it. I plan to add a couple more measurements today (instead of just the single point I've done so far) and live with it for a bit. I am working with my dealer and he is coordinating with Harman to have everybody look at what I've done so far. Sanity check plus hopefully make suggestions and answer a few lingering questions I have. Like the deal with early reflection compensation on the L/R mains; they suggest turning it off, but in my system the impulse response looks much worse than with it on.

 

[Balle Clorin]

Hi thanks. Not sure what you mean by cycles&octaves, is cycles the same as number of reflections ?
It was Trinnov support that told me to goto 200-300ms FIR length and 30# IIRs , but also said 600ms would cause a machine crash.

By the way , my room is an untreated living room with Revel F36 speakers placed 20% of room length from wall , mic/I am at 70%. No subs or sorround

 

[Olli]

Predicted results
Tweaked:

Basic

 

[Olli]

I'd already set the global variable to allow an extra couple dB of attenuation but my main problem frequencies are room modes that create big nulls. I could up the boost, but all the EQ in the world won't fix those, and would lead to poor response everywhere else.

That's what I did, my correction range is now from -18 to +20. What's the downside from that?

 

[mitchco]

In their webinars and literature Trinnov claims good agreement from predicted corrected response to real-world measurements; very nice to see that verified!

I'm curious the impact of such long FIR lengths for small rooms. Looked at one way, 300 ms is the equivalent of about 338 feet, or 1/0.3 s = 3.333 Hz. Seems like decay times in small rooms would not be so long, or is it that some are so live that reverberation times are long? It does not look like there is a large change in predicted "after" response form the longer length. @mitchco ?

Increasing the number of filters makes sense to me if they are needed or depending on how much you want to flatten the response. My room has some large nulls, mainly from the front subs, and some secondary reflections especially from the center that are likely the couch and perhaps the big TV screen behind it. I have never been able to do much about the nulls from the front subs despite moving them as much as possible; the solution was to add another pair in the rear.

I'd already set the global variable to allow an extra couple dB of attenuation but my main problem frequencies are room modes that create big nulls. I could up the boost, but all the EQ in the world won't fix those, and would lead to poor response everywhere else. My MLP is constrained by room geometry including a door; my original plans for nice primed dimensions were tossed (along with about 1/3 of my floor area) when we added a bedroom for my younger son (a worthwhile trade, but now that he's out of the house I sometimes wish I had all that extra volume back).

I have decided to not get too anal about it just yet. Too easy to spend all my free time (extremely limited) tweaking instead of enjoying it. I plan to add a couple more measurements today (instead of just the single point I've done so far) and live with it for a bit. I am working with my dealer and he is coordinating with Harman to have everybody look at what I've done so far. Sanity check plus hopefully make suggestions and answer a few lingering questions I have. Like the deal with early reflection compensation on the L/R mains; they suggest turning it off, but in my system the impulse response looks much worse than with it on.

Hey Don, @Olli's results look promising!

Yah, I have found that most top notch DSP packages the simulation is virtually identical to the measured response. I have verified that to death in Acourate, Audiolense and a few other DSP software. I would expect the same from the Trinnov.

Re: long FIR filters in small rooms. In my room which is 31ft x 16ft x 8ft, I use 6 or 7 cycles at 10 Hz which is 600 or 700ms of time domain correction. It uses a frequency dependant window so by 600 Hz or so we more or less working with the direct sound, if required, and letting the room reflections through. Otherwise it is a partial correction stopping at 600 Hz.

This allows one to control the time domain response over a period of time in the room. For example, this is one channel of my JBL 4722 Cinema speakers with dual Rythmik F18 subs using FIR filter that is 131,072 taps long and 6 cycles:

No smoothing and default window of 500ms which is letting the direct sound and all of the reflections in. As we can see in both both responses tracks the minimum phase target just about perfectly at low frequencies. And not just at one location either as I have shown in my book and articles.

Another way to look at it is a step response. Aside from the time alignment of the direct sound all arriving at the same time, we see the tail of the low frequency response follows the minimum phase target closely in both channels over 300ms:

It will be interesting to see some frequency response/time domain measurements and commentary on how it sounds once it is dialled in.

Looks like you are moving in the right direction that's great!

 

[Dimifoot]

That's what I did, my correction range is now from -18 to +20

That’s a lot of boost. I would suggest you don’t go over +8dbs.

 

[Dimifoot]

And one position measurement only? Why?

 

[DonH56]

Thanks @mitchco ! And duh, I wasn't thinking that of course it needs a few cycles to work on... My room is smaller at 13' 3" W x 17' 7" L x 8' 6" H. I will play around with more and longer filters.

@Olli -- Depending on how they implement it, attenuation is not usually a problem until you start to hit the noise floor (not usually a problem in practice). But, large amounts of boost can be problematic, in a couple of ways I can think of off-hand (Mitch and others will undoubtedly list more). First 10 dB is a factor of ten in power, and 20 dB a factor of 100. That means you are asking your power amp to deliver (and speakers to handle) 10 to 100 times the nominal power. And with a corresponding decrease in headroom.

Second, deep nulls tend to be related to room modes and or boundary interactions (SBIR), and occur because signals are cancelling. In that case, 1 - 1 = 0 and 100 - 100 still = 0 -- you can't really win. I have measured 20~40 dB nulls in rooms and it is just not reasonable to hope you can overcome that with brute force.

Finally, if you put a large boost in around the MLP (remembering bass waves are very long so a foot or two around the MLP may not matter much), then it is going to be very loud elsewhere in the room. If you move around, or have more people listening, y'all will get blasted by that much boost.

Oops, guess that was three reasons... I would limit the boost to 10 dB or so and, if there are deep nulls, look for the reason and try to work around them by moving speakers, subs, and/or the MLP. If they are narrow, especially in the deep bass, it is probably better to let them go as they are likely inaudible.

My solution was to add subs; by driving from different spots you can "cancel" the nulls, at least to some extent, without having to use excessive boost.

FWIWFM - Don

 

[DonH56]

Here are my current graphs. These use the Harman settings with a couple of changes: I extended the IIR to 10 Hz, and enabled early reflection compensation for the L/R speakers (they disable it by default). Using the Harman target curve.

 

[Olli]

Thanks @DonH56 and @mitchco! +12 dB still looks pretty good since below 80 or so it's all bass managed by the MSO optimised subs. @mitchco, attached all available graphs from the optimizer (exept for Amp. (direct)):

 

[Olli]

And one position measurement only? Why?

I have 2 different setups for 2 different listening positions ("Beamer" vs "TV").

The "TV"/Sofa listening setup is suboptimal anyways (that is the reason why I went for the Altitude instead of the DSP 75, despite using mainly Harman group speakers - I wanted to have the remapping feature), and usually it is just me and my wife watching news or stuff where perfect sound isn't so important. So at some point I might take multiseat measurements there with the Trinnov, but for now I am trying to opimize the single seat MLP in the "Beamer" setup. All measurements btw show the "Beamer" MLP where there is only one seat.

Subs are optimised for 4 seating positions anyways, since for reasons I still don't understand 100% I got the best results with MSO optimising for multiseat.

 

[Balle Clorin]

In notice that my Trinnov Amethyst is set up with. Level alignment to dbA, not dBc ,and room smoothing s to "SQUARE modulus", not "modulus". I guess square takes more aggressive action to deviations.

Another observation is the directs sound calculated by Excel file found on internet, and Trinnov "direct amplitude" is quite similar,
So far I can only eliminate the big valley is by setting calculator speaker very close/touching to side walls. Can be interesting to try in practice.

Green unorrected, blue Trinnov active,

Ettects on direct sound(actually (direct+first reflections according to Trinnov) and full sound (flat default precicion target to be consistent with my other posting, I prefer a 3db rising slope from 500Hz and down).
standard settings FIR=100 ms, 10 IIRs. DIRECT sound

+Correction cannot do anyhing with the dips in initial direct+first reflection sound,at 250/350Hz,,,, but the overall sound is quite even, but the 350Hz correction is futile.

 

[mitchco]

Hi @DonH56 and @Olli

Don looks good, I will be curious to see an REW measurement of both the corrected frequency and timing/phase response. It looks like the timing has tightened up in the impulse response. But IR displays are weighted towards high frequencies and obscures the bass frequencies which are about 1/100th the amplitude of the treble. So a step response, which is more evenly weighted across frequency, is more appropriate and what we can see in an REW measurement.

Olli, thanks for the graphs. The phase and group displays seem to indicate not too much timing correction down low... It will be interesting to see some REW measures.

So guys, how does it sound?

 

[Dimifoot]

Just added 2 subs (now 4) to my Trinnov

LFE measurement

 

[mitchco]

@Dimifoot looks good! Would love to see it with REW default vertical scale range of 45 to 105 dB SPL in 5 dB SPL increments. Any chance to show phase or group delay or a step response as well? Measured at the MLP?

 

[Dimifoot]

@Dimifoot looks good! Would love to see it with REW default vertical scale range of 45 to 105 dB SPL in 5 dB SPL increments. Any chance to show phase or group delay or a step response as well? Measured at the MLP?

I am calibrating right now .
Lots of work ahead.

I can give you your first request (always MMM for the MLP and 60-70 cm either way), as close as you have requested it, for some reason I can't get it to 45-105...