BBV's Room Measurement Saga Continues...

Just a quick up-date. (Ha, I said that before posting all the graphs!) Went to Nicks today to help him now he has the new soundcard and stuff. He was panting when I met him so I figured he needed help of some kind!

Did some measurements for adjusting the crossover. Seems the controls are simply level of each frequency band. There are no slope controls. So we got the response as flat as possible. Because the system was meant to be an MB2 & XBD setup though, the closest to flat was simply mid and treble right down and bass right up.

Interestingly, taking a measurement from the listening position, of the mid and low frequency drivers separately, showed that the slopes were not really matched that well. While the mid dome rolled-off at about 24dB per octave, the bass driver was more like 16dB per octave. I think this is a main cause of the bump around 400-500Hz. This is also why I said adjusting slopes can be useful. If you have a system that is not going to move or change (including listening position) then it could be said that the room is simply an extension of the loudspeaker system. Thus the drivers should be made to cross well taking this into account. Different positions and dispersion patters of drivers in a room can cause them to act differently even over the same frequency range. As the manufacturer can not know the room it will be in then they can but cross the drivers in a non-environment and the closer you get to this, the more accurate it will be.

Right, back on to the subjective stuff rather than my late night hi-fi waffle!

We took a measurement of the response like this but I'm afraid I have lost it. I do have one after the speaker matching which will be pretty similar though.

We then took a separate measurement for the left and right speakers. Red is the right-hand speaker.


As you can see there are some discrepancies between them. So, we EQ’d them to match as closely as possible without using too much EQ. (will change too much depending on seating position)



We then took a measurement of both together.



And EQ’d that.



We then fine-tuned a tad by ear (the tad was out of tune, you see) but it wasn’t much of a change as it sounded good right off! I don’t have any measurements for this.

I also took some readings with ETF.

Non-EQ’d impulse response



The EQ’d impulse response.



As you can see the main reflection at about 3.5ms has been reduced. I think this is because the right speaker’s mid-treble level was reduced. I believe it was closer to the wall than the other so that’s why it needed it.

And the NON-EQ’d waterfall



And the EQ’d waterfall.



The EQ'd waterfall might look worse but that is just because when the response was flattened out ETF raised the highest peak to the top of the graph and thus the floor was raised.

I think the next step is to tighten up the bass and clear up the treble/midrange with some room treatments. The LF decay is not bad, but not brilliant. Most studios are happy if it averages around 400ms.

Here is an RT60 plot of the EQ’d system, just for interests sake.



The energy time and cumulative delay plots showed that the EQ’d system had a smoother and slightly faster mid and high frequency decay as well.

It really does sound pretty good as it is now though I must say. Nearly all harshness is gone, despite what I said on my last visit. What was there I believe was the DEQ’s DAC and the bare walls/ceiling.

That should be enough graphs to keep even the most measurement hungry hippos happy!

None of this really gives much of an idea about how it sounds. However it does nicely show where corrective EQ can help a system and where it can’t, I feel. Which of the things here produce the biggest improvement in sound is probably dependant on what sound aspects you give importance to.

 

Thats very interesting......

I had a demo of a tact unit, although the room curves were matched and averaged the sound via the tact was very well "digital" is the best I can describe it. Which was a shame as it was in a system contex of about £50k.....Once the unit was switched out the sound relaxed and everything seemed to flow again.

 

Thanks for all your help Tenson! I met tenson after the gym hence the panting - it wasnt just because I was happy to see him

How much did you lift 100-200 hz after the final plot?

I will probably remove the diffraction grill on the tweeeters - which should raise it up to 20 khz a bit.

There is a little residual harshness but this seems to be due to the deq2496 when the output level gets quite high with the output boost on the analog stages starts to go into distortion. On lower level signals it seems to dissapear. The return of my usual dac should cure that.

 

Try reducing that peak around 1khz.

 

Sounds like a good idea - I will put up the eq settings too. Reducing 10khz by 2db might also be a good idea.

Tenson put up 100-400 a couple of db I think compared with the last plot. It might need a little more.

 

I'd shelve it down above 800hz. You will be able to listen louder for longer with no loss of resolution. Depth will increase and you will lose any last trace of harshness. But hey, that's just me.

 

Some of the harshness is the berhinger - the output stage isnt fantastic. As the recording goes into the red it can sound a bit much.

I will step down 1-2 khz first and take it from there... we did try dropping the crossover treble a notch but it seemed to make it sound worse.

 

Hi,

That RT60 is really way high in the midrange and treble.

Switching to dipoles or other controlled directivity speakers would help a lot to reduce the room effects.

Otherwise it looks to me like the only other solution is to apply some major heavy room treatment for both midrange/treble and for the LF peak around 42Hz.

Something that would help in the midrange and treble are BBC Style felt masks or my little "felt dot" tweak from my "Freakzoid Tweekaloid" article on enjoythemusic.com.

I would say stacking two felt dots for the area around the tweeter and four for the area around the midrange should probably show a significant improvement, you would need around five stacks around the tweeter and a lot of them around the midrange, large packs of black self adheasive felt dots around 1" Diameter are available at your local B&Q to be applied to chair/table legs on laminate floors.

For the bass very little can be done except large plate absorbers.

I suspect a handfull of "Audio-Smile" absorbers IN ADDITION to the felt masks/dots will also do nicely.

Ciao T

 

I think thats very sensible 3d - obviously room treatment and directionality control will reduce the need for eq and improve the waterfall plots. When I settle in a permanant residence I plan to do exactly that. In the mean time finishing stocking all my bookshelves might help a little!

Sadly even a thick rug can only do so much against concrete walls and laminated floors

 

Hi,

First, if the meter goes into the red, you MUST reduce the overall (digital) gain of the Behringer DEQ.

Second, get Tenson to perform the "analog stage bypass tweek" I taught him, next time he is around. You may find you prefer the DEQ "au naturelle" over attaching another DAC to it's digital output.

Honestly though, given Tensons comments on X-Over slopes etc, your best choice would be a Behringer DCX2496 and to get a custom made 6-Channel MF Audio Passive Preamp while applying the "bypass tweek" to the DCX Outputs and fitting a "transformer only tweek" based input stage to the analog inputs for use with your turntable.

With the DCX you can make sure that the crossover actually meets the essential criterias to work correctly.

I have recently spend a lot of time on researching this back from the first principles (as I needed to develop a relatively "normal" speaker for my new commercial venture - Abbington Music Research - www.amr-audio.co.uk - currently no website posted yet, in a few weeks).

The bottom line is that Siegfried Linkwitz was actually ABSOLUTELY correct in asserting that the 4th order butterworth crossover provides corrct integration and a slow and even rate of turn of phase which leads to minimal waveform distortion which retains a reasonable transient performance (if not a perfect one). However, this requires that the ACOUSTIC OUTPUT from the driver follows the 4th order butterworth curve, NOT the crossover itself, plus the drivers need to be timealigned.

Sounds to me just like the job for a digital crossover. In fact, my current research has given theoretical validation to that "ages ago" activated PMC studio setup I worked on, as empirically we ended up with 4th order acoustical LR crossover slopes, with the drivers time aligned via delays. I now know why we did.

BTW, a perfect transient performance from a multiway speaker system is practiclally impossible or at least extremely difficult to attain, as more than four octaves of overlap is needed between drivers around each crossover point, together with true first order filters, which in turn require midrange and treble units that are in effect capable of producing serious excursion well into the mid/low bass.

For reference, if we choose a 1KHz first order crossover with timealigned drivers our woofer needs a flat response up to AT LEAST 4KHz (better 8KHz) with minimal response non-flatness and a smooth rollout above it's cutoff, while the treble unit would need to be flat to 250Hz (better 125Hz) with minimal response non-flatness and a smooth rolloff below this, PLUS the treble unit would need to be able to handle a lot of excursion.

Lastly, for a passive X-Over to work like that without major lobing errors would require a co-incident driver.

In effect this kind of brings us back to the old Olson/RCA LC1 coaxial driver or to devices such as the german Eckmiller. A modern implementation might use a large Cone driver with good response at higher frequencies and semi-coaxial small fullrange driver (which has good HF behaviour) using a first order electrical passive crossover as the best first approximation, JBL used to make such a device....:MILD:

Ciao T

 

Yes I agree - the dcx also with bypass mods would be the best solution.

Most dac chips can drive well enough. As soon as I get my dac repaired Tenson is going to bring over his modded deq to compare. We also plan to try it in the analog chain when I have a working phono.

I did turn down the gain of the deq by an amount just over the highest eq gain to remove the possibility of digital distortion. What we were hearing was more like an analog stage being run close to its limit. I have now turned off the extra pro analog gain to mitigate this.

Good luck with your new venture! I may have some news of my own soon

 

Hi,

The Felt-dot tweaks do not need to wait. Pop into B&Q next time you are there and get a few buld packs of the dots. They are easy to try, place and remove....

Here a picture of the same self felt dots on the pair of Epos ES-11's AMR is still using as reference monitors (we have a few other speakers sitting in Beijing, including ATC's and Opera, non match the old Epos's):

A thick rug on the floor becomes more effective if given a backing of thick and heavy carpet felt, a heavy wall rug becomes more effective if it is spaced away from the wall by maybe 2" and if "eggcrate" acoustic foam is inserted behind it. I especially siggest this for the place behind the listening position.

Lastly, while you get black felt dot's why not pick up a bulkpack of white ones and place them as per my ETM Article on walls and ceiling?

Ciao T

 

OK I will try the directionality tweak, thats a really good idea - will the white dots take the paint off the walls and ceiling? If so the landlord will object!

I will check the rug more closely tonight and look into using some additional backing...

Behind the listening position is the kitchen on 2/3 and the hallway on the other 3rd, its not possible to put foam behind sadly

 

The tweeters have a diffraction grill in place currently. Removing that should also improve directionality.

 

Hi,

It depends on both paint and the backing of the dots.

You might just want to try one in a spot that is not very obvious and leave it there for a week or two and see what it does. Alternatively, white-tack (the white version of blu-tack) could be used instead, this will definitly come off without pulling paper of the wall or leaving stains.

Carpet felt (natural) from any well sorted carpet shop. Simply get them to cut it a little smaller than the rug, put dwn the felt, then the rug on top, use some staples if neccesary to hold together.

Well, you can always try to do what you can get away with, anything else you have to live with. My wife vetoed my old wall-rug and we have not agreed on a new one yet....

Ciao T

 

I left a peak around 1KHz on purpose, I find it gives more presence. If you look at the equal loudness curves you can see why. The same goes for the slight dip around 100-400Hz.

Nick I raised 125Hz-300Hz by 1.5dB

Remember that with adding the felt dots you will need to adjust the EQ again. The matching between the speakers will also need another look.

Personally I wouldn't bother with the dots on the walls, but I would be interested to hear what you find from doing it.

Is the RT60 that high? It averages around 550ms. 350-400ms, if well controlled, would be about perfect I think so it is not hugely high. Just high. You can't really trust the RT60 plots below about 200Hz anyway (at least in ‘small’ rooms) so may as well ignore the 41Hz peak. Waterfall plots tell much more about this range.

 

Will do...

I can do whatever I like bar the landlords possible objections

 

I find it makes systems forward and fatiguing, not IMO ideal in Bryston/PMC setup.

It will bring certain things artificially forward in the mix, giving the illusion of greater detail. I prefer a more natural perspective, hence the shelving I tend to gravitate towards. More of a BBC perspective.

Not tried sticking furry dots all over my room. Not likely to either.

 

Fair enough, I like to slightly follow the equal loudness curves. I don't find my system forward and I don't think Nicks was either. I do like a lively sound compared to something that is mellow and laid back though.

Its up to Nick to fine tune it by ear really.

 

I am inclined to drop 10khz by 3db
drop 1250 - 1750 Hz by 2db
raise 125 - 275 Hz by another 1.5 db

Where did you put the peq fill in? If that compunds with the geq to raise over the 0db digital gain point then we might still get some distortion so I will check that too.

I wil put all the peq and geq settings on the thread tonight.