AVS Forum banner

JBL Synthesis Calibration video

19K views 109 replies 15 participants last post by  andyc56 
#1 ·
Hello,


I have read much about the different types of equalization systems, and always end up reading that the Synthesis EQ is as good as it gets. I found this video showing a complete calibration: http://www.youtube.com/watch?v=PHaWWAKAGEM


I'm a little surprised though... It looks like, it is "just" a set of PEQ's for each channel? Trinnov and Audyssey always write about how they calculate the room in the time domain etc, but is the Synthesis EQ just some simple PEQ's to match a given target curve (with some really good processing equipment)?


If anyone has details about the Synthesis EQ-setup (Called Harman ARCOS i believe), then I'm very interested in hearing what exactly is being done
 
See less See more
1
#2 ·
This demo condenses an hours long process into a few minutes. It shows the highlights of the process and the software, but does not drag us through every test and calibration event. When my Synthesis® One Array system was calibrated two years ago (using the pre-HATS system called DACS) it took about seven hours from start to finish. The newer HATS systems automates some of the work, but it still requires real human interaction to complete the process, and that takes time.


One of the key and often overlooked advantages for JBL Synthesis® is that the calibration software comes with preloaded profiles for every Synthesis® component (and then some). So as it learns the identity of the specific speakers, processor, SDEC unit, etc. it already understands the performance characteristics of each as a single unit and the synergy of the group as a whole.


When the mics are calibrated to the room, they can supply additional room-related info to the HATS unit that can be used in the calibration of the individual channels. (That's why the mics must be re-calibrated for every room and system.)


This knowledge of the electronic components, of their placement in the room, and of the characteristics of the room itself makes for a very tight integration of room and equipment into a "whole" system. I think of it as a true, complete system in which the room's characteristics, the equipment's characteristics, the seating positions (as determined by mic placement), and the intended PEQ curves come together in a--wait for it, wait for it--synthesis.
(Cheesy, I know.)


Additionally, it's possible for the operator to make modifications on the spot, so the "house curve" is not dictatorial in any sense. Once the final calibration is done, it's then uploaded into the SDEC unit's memory. Thus the HATS CPU trains the SDEC, whereas in other systems the built-in algorithms may be more predictive and directive.


The JBL SDEC units are customized BSS Soundweb London BLU units. I have the SDEC4000, so there are two BLU units (linked together by Ethernet like a pair of servers) rather than the one pictured in the SDEC3000 YouTube video. That's a lot of freaking processing power, yet it all goes back to the HATS unit's software and calibration equipment and human controller to bring it all together. It's far, far more than just a set of PEQs per channel.
 
#3 ·
Filecat covers everything I would have said
.


We have two parallel systems in our theater , one of which uses the Synthesis, the other Audyssey Pro. The Synthesis systems blows away the Audyssey in so many dimensions. As filecat mentions, since Audyssey doesn't know the speaker system, it will try to do heroic things to even out the curve, only to create subjectively worse performance. We currently do not use our Audyssey and instead are in the process of using the Synthesis to do what it is supposed to do.


The other factor is the extensibility. Our primary JBL Systhesis drives a 20-speaker JBL package. But given the poor performance of Audyssey, we added another expansion module and now are performing 40 channels of EQ! All in three, 1RU units.


Harman's philosophy is that if you start with speakers with good directivity, you can then EQ their response in the room effectively and with it, also fix time domain issues. Once you achieve (near) flat in-room response, there should be no desire for anything else.


That said, Trinnov has other functionality such as virtually moving speaker locations and such. If that is of interest, then that product might serve you better. Otherwise, JBL Synthesis performs exceptionally well.
 
#4 ·
Both your replies makes very good sense!


Anirm: Its funny, I actually stumbled across your website yesterday, when googling about the Synthesis calibration. It looks very nice!


I completely agree on your thoughts about Audyssey. I was using Audyssey Pro in my room to begin with, and it did something to the sound, which I didn't like. Sure, it flattened out the response, but something was lacking. It sounded a bit diffused/unprecise and lacked punch.


I bought a DBX DriveRack 4800 and measurement gear and did a manual calibration only using PEQ's to a Synthesis-like target curve. This blew away the Audyssey calibration! (It took much longer to setup though
).


Anirm, are you using the SDEC's on a non-synthesis speaker system in your room then? I am using JBL Pro cinema speakers (active x-over) and it would be interesting if my system could be calibrated this way if I got the BSS/SDEC units.
 
#5 ·
Thank you for the kind words regarding our site and theater
.


We are about to go down that journey, calibrating our Wisdoms with the Synthesis. It will be a challenge because the Wisdoms are pretty unconventional with their use of ribbons and traditional driver arrays. I will report back on how well we do. Currently the Synthesis processor is driving the Wisdoms but with no EQ.


From connectivity it was a pleasure to set up as we are using a single cat-5 with Cobranet between the primary SDEC processor and the expansion box driving the Wisdoms.
 
#6 ·

Quote:
Originally Posted by amirm /forum/post/20753941


Thank you for the kind words regarding our site and theater
.


We are about to go down that journey, calibrating our Wisdoms with the Synthesis. It will be a challenge because the Wisdoms are pretty unconventional with their use of ribbons and traditional driver arrays. I will report back on how well we do. Currently the Synthesis processor is driving the Wisdoms but with no EQ.


From connectivity it was a pleasure to set up as we are using a single cat-5 with Cobranet between the primary SDEC processor and the expansion box driving the Wisdoms.

If you are willing to share, I would be very interested in hearing about the calibration procedure of the non-synthesis speakers when you get to it. Do you then use the same Harman measurement software?


I am planning on purchasing the BSS BLU800 and BLU32 with digital inputs/outputs and an external DA/AD converter in the future. This will still be a completely manual procedure for setting it up, but as you have also said, I just dont trust the automatic procedures. The current result with the DBX DriveRack 4800 and by using the REW software and filtergeneration is very good though, so I'm not in a hurry
 
#7 ·

Quote:
Originally Posted by amirm /forum/post/20753829


Harman's philosophy is that if you start with speakers with good directivity, you can then EQ their response in the room effectively and with it, also fix time domain issues. Once you achieve (near) flat in-room response, there should be no desirefor anything else.

so the 'total' specular response of the room is not something that is taken into consideration regarding calibration? just "flat frequency response"? or is it to be addressed with other means. thanks,
 
#8 ·

Quote:
Originally Posted by amirm /forum/post/20753829


Filecat covers everything I would have said
.


We have two parallel systems in our theater , one of which uses the Synthesis, the other Audyssey Pro. The Synthesis systems blows away the Audyssey in so many dimensions. As filecat mentions, since Audyssey doesn't know the speaker system, it will try to do heroic things to even out the curve, only to create subjectively worse performance. We currently do not use our Audyssey and instead are in the process of using the Synthesis to do what it is supposed to do.


The other factor is the extensibility. Our primary JBL Systhesis drives a 20-speaker JBL package. But given the poor performance of Audyssey, we added another expansion module and now are performing 40 channels of EQ! All in three, 1RU units.


Harman's philosophy is that if you start with speakers with good directivity, you can then EQ their response in the room effectively and with it, also fix time domain issues. Once you achieve (near) flat in-room response, there should be no desire for anything else.


That said, Trinnov has other functionality such as virtually moving speaker locations and such. If that is of interest, then that product might serve you better. Otherwise, JBL Synthesis performs exceptionally well.

amirm;

When I win the Lotto - I'll have your team and Dennis Erkstine group on my short list for doing my next HT.

As an Engineer, this is where it's at:
Quote:
Advanced Mathematics to the Rescue

The solution to this problem is the subject of a later white paper but for now, a quick teaser.


Keith uses fancy computer simulation software called Computational Fluid Dynamics (CFD). The entire room including its furnishings is then recreated using a 3-D model in the computer. Speakers are simulated using their physical characteristics and their sound waves allowed to “excite” the room. As the sound waves propagate there, we are able to see the precise frequency response at each location.


Using the above simulation then, we can run different trials, examining the various position and number of speakers as to both achieve a smooth low frequency response and least amount of seat to seat variations.


The above process is a far more sophisticated than the standard method used by acoustic designers. Instead of modeling the physical room and speakers, they make gross simplifications that allows them to use simple calculators to compute the bass anomalies.


Problem is that real world doesn't work that way. Drywall for example, flexes ever so slightly as the lower tone sound waves hit it and thereby absorbs some of the bass energy (a good thing). It is not a rigid boundary as the simplified acoustic modeling assumes. Chairs are soft and have an effect in the room. All of this adds up to a lot of inaccuracies.


Keith’s CFD modeling takes all of the above factors into account, creating surprisingly accurate results that come very close to actual measurements in the room (+- one decibel in technical talk). The drawback is the high cost of tools, the sophisticated expertise required to properly model the room, and good set of ears to know when the computer is wrong -- all the things Keith and his team excel at.

The only thing I'll add - hopefully not too much OT this thread - is most people are like me, "avg joes" who are into this as a hobby.

We do need simplistic tools for our home theatre rooms. CFD is good, but its a very detailed/specialized field, and $$$.

Now, as CFD is done, hopefully those lessons learned can be turned into "best pratices", say getting you 80-85% "there, so "avg joe" HT enthausist can benefit.

I work @ Auto OE and see that daily. We have specialized dept of people who do that, and they turn their findings into design std's that for the next design are considered as baseline.
 
#9 ·

Quote:
Originally Posted by JonasHansen /forum/post/20754069


If you are willing to share, I would be very interested in hearing about the calibration procedure of the non-synthesis speakers when you get to it. Do you then use the same Harman measurement software?

Yes, we will be using the exact same kit, except that we will be telling it we have unknown speakers.

Quote:
I am planning on purchasing the BSS BLU800 and BLU32 with digital inputs/outputs and an external DA/AD converter in the future. This will still be a completely manual procedure for setting it up, but as you have also said, I just dont trust the automatic procedures. The current result with the DBX DriveRack 4800 and by using the REW software and filtergeneration is very good though, so I'm not in a hurry

Well, if you go that route, you won't have a choice but to use the manual system. The Auto EQ requires that you buy the JBL version of the BSS hardware.


We do a ton of systems with BSS using manual procedure so that is a fine option too. As you say, it requires time and understanding of what it does. BTW, even with automated process, we still hand tune as the video said.
 
#10 ·

Quote:
Originally Posted by mtbdudex /forum/post/20754463


amirm;

When I win the Lotto - I'll have your team and Dennis Erkstine group on my short list for doing my next HT.

I hear you
.

Quote:
The only thing I'll add - hopefully not too much OT this thread - is most people are like me, "avg joes" who are into this as a hobby.

We do need simplistic tools for our home theatre rooms. CFD is good, but its a very detailed/specialized field, and $$$.

I hear you
.


Fortunately, Keith has now packaged some of the CFD work as a service. For a fixed cost (less than $5K), Keith takes the drawings of your room and potential areas for subs, and will recommend using CFD simulations the optimal locations and number of subs for you. Since the impact of the room is in hugely in that area, that problem is solved mathematically and clearly. According to Keith, in-room variations are as low as just 1db vs the model! I am working on a write-up and video/pictures for this and will post on the site in a few weeks.


It is still costly of course but given that none of us have the tools and training to run the same, it is still a good tool for mid to high-end systems that you put together rather than a company like mine doing it.

Quote:
Now, as CFD is done, hopefully those lessons learned can be turned into "best pratices", say getting you 80-85% "there, so "avg joe" HT enthausist can benefit.


I work @ Auto OE and see that daily. We have specialized dept of people who do that, and they turn their findings into design std's that for the next design are considered as baseline.

And that has been the challenge to make this something everyone can use. Keith uses PhD students who have specialized in this area, together with training on the software as you say in your field to do this work. Even entering the room data can be time consumer.
 
#11 ·

Quote:
Originally Posted by localhost127 /forum/post/20754294


so the 'total' specular response of the room is not something that is taken into consideration regarding calibration? just "flat frequency response"? or is it to be addressed with other means. thanks,

Oh, the whole thing is about in-room response. The measurements are with the speakers in the room. What I meant is that the software has both in the room and out of room measurements. So when it sees something that matches between the two, it won't try to compensate and push the speaker into distortion. I am oversimplifying here as they may do some correction but most of the time the system is heavily correcting in-room response.


So the flat response will be the measured in-room version as the room almost entirely defines your low frequency response. You can see this in the video although it is not the most clear explanation. See the post it notes that show up on the video, indicating what is the corrected in-room response vs original that is shown in gray.
 
#12 ·

Quote:
Originally Posted by JonasHansen /forum/post/20754069



I am planning on purchasing the BSS BLU800 and BLU32 with digital inputs/outputs and an external DA/AD converter in the future. This will still be a completely manual procedure for setting it up, but as you have also said, I just dont trust the automatic procedures. The current result with the DBX DriveRack 4800 and by using the REW software and filtergeneration is very good though, so I'm not in a hurry

Just for grins, I bought a pair of BSS 366T Omnidrive Compact plus Speaker Management Systems (from a former JBL Synthesis® employee at Northridge, BTW) to use in my Frankenstein JBL L250 quad-amped stereo pair with JBL inverted dome Ti drivers replacing the 8" and 4" units and with a new 1.5" Ti dome in the top. Using the most current London Soundweb software, FuzzMeasure, a MacBook Pro, and a wheelbarrow full of calibrated mics and equipment my buddy brought from JPL (not JBL) we made a stunningly wonderful, stupidly expensive (for a DIY project) pair of speakers.


Of course it took the better part of three days spread over three months to test all the drivers outside both in and out of the cabinets, assemble everything, cable the four stereo amps and four leads to each speaker, connect the pre and BSS units together then the BSS units via eight XLR cables to the amps, then calibrate it all, but Wow!


For people who need things sooner and with less futzing around, it makes no sense, but it does show that you can do just about anything with quality electronics, patience, and care.
 
#13 ·

Quote:
Originally Posted by JonasHansen /forum/post/20752223


I'm a little surprised though... It looks like, it is "just" a set of PEQ's for each channel? Trinnov and Audyssey always write about how they calculate the room in the time domain etc, but is the Synthesis EQ just some simple PEQ's to match a given target curve (with some really good processing equipment)?

I've been wondering about this very issue for a long time. I had a few different EQ systems in my room and tried to see if all the "phase and filter type(IIR, FIR), time-domain, adjustment points, etc" discussion among Audyssey and Dirac made some audible difference.


The plot shows 3 different EQs. Top is MultEQ XT, middle is Dirac Live, bottom is PEQ. Measure electrically using REW, no smoothing.




Same curves, expanded for a better look at HF EQ detail.




In order of sound quality, it's 3d, 2nd, 1st. PEQ is best, for me, my room, my age, my music, my ears. It's not really anything to do with spectral balance, although they were not all identical (I modeled the PEQ after Dirac, so they were very close). It was more about "do I want to listen to this for a long time or not?" The AP20 sounded fine, and I did the A/B comparisons by feeding the SSP into its A/D - D/A chain so I could easily switch Dirac on/off. Only after I removed it from the chain and let the SSP once again drive the amps did I feel the "lifting of the veils" thing. That may be down to the extra A/D, different D/A and output stages, psychology, etc.


At a different time, the Onkyo's line outs were driving the amps, with the SSP flowing through the 7.1 analog bypass. Yet even so, the SSP sounded much more real, especially with vocals.


At the risk of misrepresenting the true nature of how these EQ systems synthesize corrective processing, the Dirac curves look to apply something like 1/10th octave corrections, uniformly across the spectrum. MultEQ varies by a factor of 50:1. It has 50 times more granularity at high frequencies than the bass. Not sure if that is a design choice based on human perception, or that's just how FFTs work.


Having seen such a wide range of EQ resolutions, with no obvious correlation to sound quality (other than maybe inverse correlation), it begs the question, how much EQ is enough, and when does more of a good thing become a bad thing?
 
#14 ·

Quote:
Originally Posted by amirm /forum/post/20754741


Yes, we will be using the exact same kit, except that we will be telling it we have unknown speakers.



Well, if you go that route, you won't have a choice but to use the manual system. The Auto EQ requires that you buy the JBL version of the BSS hardware.


We do a ton of systems with BSS using manual procedure so that is a fine option too. As you say, it requires time and understanding of what it does. BTW, even with automated process, we still hand tune as the video said.

What measurement gear are you using for the manual calibrations with the BSS systems? As I mentioned, I use REW because it creates the filters to match a desired target curve. These filters are a very good start and can be manually adjusted afterwards for fine tuning. But REW actually does a very good job of finding the minimum set of filters.


I thought about using Smaart v7, where I can connect multiple mics and see their individual responses in real time, but I dont think Smaart can generate filters for me.
 
#15 ·

Quote:
Originally Posted by Roger Dressler /forum/post/20757199


I've been wondering about this very issue for a long time. I had a few different EQ systems in my room and tried to see if all the "phase and filter type(IIR, FIR), time-domain, adjustment points, etc" discussion among Audyssey and Dirac made some audible difference.


The plot shows 3 different EQs. Top is MultEQ XT, middle is Dirac Live, bottom is PEQ. Measure electrically using REW, no smoothing.




Same curves, expanded for a better look at HF EQ detail.




In order of sound quality, it's 3d, 2nd, 1st. PEQ is best, for me, my room, my age, my music, my ears. It's not really anything to do with spectral balance, although they were not all identical (I modeled the PEQ after Dirac, so they were very close). It was more about "do I want to listen to this for a long time or not?" The AP20 sounded fine, and I did the A/B comparisons by feeding the SSP into its A/D - D/A chain so I could easily switch Dirac on/off. Only after I removed it from the chain and let the SSP once again drive the amps did I feel the "lifting of the veils" thing. That may be down to the extra A/D, different D/A and output stages, psychology, etc.


At a different time, the Onkyo's line outs were driving the amps, with the SSP flowing through the 7.1 analog bypass. Yet even so, the SSP sounded much more real, especially with vocals.


At the risk of misrepresenting the true nature of how these EQ systems synthesize corrective processing, the Dirac curves look to apply something like 1/10th octave corrections, uniformly across the spectrum. MultEQ varies by a factor of 50:1. It has 50 times more granularity at high frequencies than the bass. Not sure if that is a design choice based on human perception, or that's just how FFTs work.


Having seen such a wide range of EQ resolutions, with no obvious correlation to sound quality (other than maybe inverse correlation), it begs the question, how much EQ is enough, and when does more of a good thing become a bad thing?

Harman's view on FIR filters is that they don't provide any improvement over IIR types and that they are more complex and costly to implement,according to their research.


Ironically, the Synthesis system costs more than the others,at around $25,000 and has added complexity to set-up
I'm not knocking it, Harman Knows more about rooms than probably anyone else
,just that it costs more and adds it's own complexity.


The new JBL ARCOS system has resolution up to 1/100th of an octave.That seems like some pretty granular correction as well.
 
#16 ·

Quote:
Originally Posted by amirm /forum/post/20754777


Oh, the whole thing is about in-room response. The measurements are with the speakers in the room. What I meant is that the software has both in the room and out of room measurements. So when it sees something that matches between the two, it won't try to compensate and push the speaker into distortion. I am oversimplifying here as they may do some correction but most of the time the system is heavily correcting in-room response.


So the flat response will be the measured in-room version as the room almost entirely defines your low frequency response. You can see this in the video although it is not the most clear explanation. See the post it notes that show up on the video, indicating what is the corrected in-room response vs original that is shown in gray.

amirm,


One thing I've wanted to know about the Synthesis system,does it create different room curve/tilt for the mains vs. surrounds? Can it take into account the power response of the speakers,seating distance/critical distance and room acoustics to create a tailored room curve?
 
#17 ·

Quote:
Originally Posted by StevenLansing /forum/post/20759966


Ironically, the Synthesis system costs more than the others,at around $25,000 and has added complexity to set-up
I'm not knocking it, Harman Knows more about rooms than probably anyone else
,just that it costs more and adds it's own complexity.

Yes, but based on the video, it's the simplest yet most thorough approach.

Quote:
The new JBL ARCOS system has resolution up to 1/100th of an octave.That seems like some pretty granular correction as well.

Having that capability, such as for nulling a mode, is reasonable. Using it rampantly >3 kHz, not so much.
 
#18 ·

Quote:
Originally Posted by Roger Dressler /forum/post/20760321


Yes, but based on the video, it's the simplest yet most thorough approach.


Having that capability, such as for nulling a mode, is reasonable. Using it rampantly >3 kHz, not so much.

Hey Roger...

Did Sanjay tell you that he and I will be at Harman monday morning for 3days of ARCOS training?



RayJr
 
#19 ·
The ARCOS system, as I have been told, has a resolution adjustment just for the purposes of reaching a target curve. This resolution setting is not the same as the underlying HATS measurement resolution. This means that a finer( not finer than the fundamental measurement of course) or coarser resolution may be utilized in obtaining the target.


The ARCOS system is ultimately a grouping of macros used to operate the underlying measurement system. As such the steps involved, and there are many, are constrained. Although certain flexibility is afforded the ARCOS based calibrator, direct control of the measurement system will allow the greatest flexibility in calibration. And the most ways to get into trouble
 
#20 ·

Quote:
Originally Posted by StevenLansing /forum/post/20759997


amirm,


One thing I've wanted to know about the Synthesis system,does it create different room curve/tilt for the mains vs. surrounds? Can it take into account the power response of the speakers,seating distance/critical distance and room acoustics to create a tailored room curve?

Assuming I won't get shot for posting this, here's a composite look at all the process in my 7.1 "Two Jims Theatre" where the subs, mains and surrounds/rears are in the "integration" process.


Sorry about the sucky, jittery photo, but it's what I got.


You can see the surrounds and rears are being treated differently (bottom curve). On one similarly weak photo, the system tells us to check one of the mics due to unacceptably low input.


This is the old JBL DACS. The HATS is more modern and capable.
 
#23 ·
Quote:
Originally Posted by jamin
The ARCOS system, as I have been told, has a resolution adjustment just for the purposes of reaching a target curve. This resolution setting is not the same as the underlying HATS measurement resolution. This means that a finer( not finer than the fundamental measurement of course) or coarser resolution may be utilized in obtaining the target.
Thanks,


that explains the resolution capability better.
 
#24 ·
Quote:
Originally Posted by filecat13
Assuming I won't get shot for posting this, here's a composite look at all the process in my 7.1 "Two Jims Theatre" where the subs, mains and surrounds/rears are in the "integration" process.


Sorry about the sucky, jittery photo, but it's what I got.


You can see the surrounds and rears are being treated differently (bottom curve). On one similarly weak photo, the system tells us to check one of the mics due to unacceptably low input.


This is the old JBL DACS. The HATS is more modern and capable.
The picture is a little fuzzy,but I see what you are talking about.Out of curiosity, what are the distances between the seating area and the main vs. surround speakers?It looks like the lower curve for the surrounds is pretty flat,which I'm guessing would mean that the surrounds are located fairly close to the seats.
 
#25 ·
All the surrounds are JBL S4Ai speakers. Are you familiar with them?


They have three sets of drivers. The ones projecting at 45 degrees off vertical axis (one set left, one set right) have a Ti tweeter (in a waveguide) and 4" cone driver; the set going direct into the room has a Ti tweeter (in a waveguide) and an 8" Al inverted dome woofer. This last pair is situated so the tweeter is below the woofer near the bottom of the baffle, and the tweeter is set at ear height or slightly above. There's potential for a lot of MF and HF energy with twelve tweeters, eight mids, and four woofers. The right amount of absorption and diffusion is a big part of the design before cutting the first stud.


So for my installation, a line from that tweeter to the middle of the center seats is about 6.75 to 7.25 feet. The rears are essentially the same distance.


The L/R mains are about 11 ft, and the center (behind the screen) is 10.5, give or take.
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top