Audyssey Dynamic EQ Correction

[maxmercy]

One of the best less than Reference Level experiences was when I had a Denon AVR with Audyssey's Dynamic EQ.  It would alter the level of the channels and the EQ applied to each channel depending on listening level, to preserve low frequency and high frequency perception via psychoacoustic formulae and discussions with Hollywood sound mixers.

 

I now use the Denon DN500AV, which does not include a room correction algorithm, but uses all balanced connections.  I use a MiniDSP for EQ and bass integration, but missed the Dynamic EQ correction.  Before selling my old Denon AVR I measured what the Dynamic EQ algorithm did.  If you have a DSP and want to implement a version of the correction (MiniDSP does not have dynamic  EQ capability, but other DSPs do), here are the corrections at various listening levels:

 

 

-6dBRef:

Sub: Low Shelf 50Hz +2dB Q0.5, Low Shelf 20Hz +1dB Q0.5, Peak 70Hz +0.25 to +2dB (varying on soundtrack overall volume at the time) Q0.25

LCRS: Peak 70Hz +0.25 to +2dB (varying on soundtrack overall volume at the time) Q0.25


-10dBRef

Sub: Low Shelf 50Hz +3.5dB Q0.5, Low Shelf 15Hz +2dB Q0.5, Peak 70Hz +0.5 to +3dB (varying on soundtrack overall volume at the time) Q0.3

LCRS: Peak 70Hz +0.5 to +3dB (varying on soundtrack overall volume at the time) Q0.3


-20dBRef

Sub: LowShelf 40Hz +6dB Q0.5, LowShelf 30Hz +4dB Q0.5, Peak 75Hz +1 to +4dB (varying on soundtrack overall volume at the time) Q0.5

LCRS: Peak 75Hz +1 to +4dB (varying on soundtrack overall volume at the time) Q0.5

 

 

JSS

[SME]

Does Dynamic EQ really work differently in the mains vs. sub like your filters suggest?  If so, then that's just plain weird.  I'm about to migrate to doing bass management in my custom DSP and will be turning off the XO in my AVR.  I use Dynamic EQ with music and in "Audyssey L/R Bypass Mode" to turn off the filters in the left and right.  The sub filter that's in there is relatively benign, so I've lived with it in but will be glad to get it out after the migration.

 

Does it really boost up to +2 dB when the MV is at "-3"?  That doesn't seem to be a smooth transition given that it's supposed to do nothing at "0".

 

As a related point of interest my recent experiments have convinced me that flat in-room response is rarely the correct target response to EQ to.  It seems that a lot of people, myself included, got confused between whether we should be aiming to achieve a flat, on-axis in-room response or a flat, on-axis anechoic response from the speaker(s).  My recent experiments suggest that, as indicated by Harman (and probably others), the latter is closer to the correct approach.  I would add that the relevance of "anechoically flat" breaks down somewhere in the low mids or bass region where waves become long enough relative to boundaries that some reflections convolute the first arrival.  Hence, I think it makes since to talk about having a flat "first arrival", even though I'm still a bit fuzzy on what exactly first arrival is.  It depends on some psychoacoustics, but my recent experiences lead me to suspect that the brain can and will rely on the shortest time window it can get away with.

 

The take-away message is that your system may be more correct or "closer to reference" if your in-room frequency response is slanted and has a significant bass rise in particular.  I would expect that a balanced in-room frequency response will show a pretty significant gain around the speaker baffle step point because sound power increases so much there.  This is noteworthy, because the vast majority of people boost their bass to taste by just increasing the sub level.  Unfortunately, the tends to hollow out the crucial 100 Hz+ bass, which I believe is essential for good punch.  Honestly, I think Audyssey and any other room EQ that force the user to use a flat(-ish) in-room curve should be chucked.  For manual EQ or for those room EQ systems that allow target curve specification, experiments may be necessary to find the best target response.  Or if one's speakers sound balanced in-room without EQ, then model the target off of that.

 

Along similar lines, I'm thinking playback level calibration could be much more reliable if based on impulse response first arrival level instead of continuous pink noise measurement.  I'm thinking that the first arrival of the impulse response measured using a -20 dBFS sine sweep ought to measure at around 76 dB or a little higher for reference level playback.  I'm basing that on playback level recommendations for near-field monitoring, which vary from 76-79 dB, depending on the source.  I'm guessing that the 76 dB number is appropriate for true near-field, hence the first arrival should match this number.  That also suggests that in a large room where 85 dBC pink noise would be an appropriate calibration level (the normal way), the room reflections and reverb add about 9 dB SPL to the direct sound of the 500-2 kHz continuous pink noise.  That seems pretty reasonable to me.  See also the X-curve, which is a 3 dB/octave roll-off from 2 kHz on up that may account for a combination of narrowing speaker directivity and increased air absorption.

 

If you are playing at your room appropriate playback level and your system is calibrated as per above (with a 76 dB first arrival and a slanted frequency response target curve), you may not need Dynamic EQ at all.

[3ll3d00d]

http://www.aes-media.org/sections/pnw/ppt/jj/room_correction.pptprovides some insight into that (first arrival at HF) approach to EQ

https://mitpress.mit.edu/books/auditory-neurosciencegoes into quite a bit of detail on how it is thought to work but, in relation to the above, it typically just means a fairly short FDW

 

I don't think any eq system out there has targeted a flat in room response has it? possibly you could argue ARC does but that has a HF limit on it instead. Target curves going back to the B&K one in the 70s have involved a rolloff towards the high end.

[maxmercy]

Sorry, made a mistake there.  the 3dB boost is at the -6dBRef listening level.  Corrected above.  at -3dB, the correction is only 1.3dB.

 

JSS

[SME]

http://www.aes-media.org/sections/pnw/ppt/jj/room_correction.pptprovides some insight into that (first arrival at HF) approach to EQ

https://mitpress.mit.edu/books/auditory-neurosciencegoes into quite a bit of detail on how it is thought to work but, in relation to the above, it typically just means a fairly short FDW

 

I don't think any eq system out there has targeted a flat in room response has it? possibly you could argue ARC does but that has a HF limit on it instead. Target curves going back to the B&K one in the 70s have involved a rolloff towards the high end.

 

I think most room EQ systems correct the response in-room.  Or they use a window that's much too long to adequately decouple the response of the speaker from the room.  Dirac Live suggests a target curve with +/- 3 dB slope across the range and allows the user to modify it.  If it was the anechoic or first arrival response being matched to the target curve, the recommended default would probably be way too dark and short of detail.  Even if you corrected using an FDW, it has to be very short to avoid capturing some early reflections in typical small rooms.

 

Anyway, I just did a quick and dirty cal to flatten the hell out of my response with a 2.2 cycle FDW and I am *very* pleased with the result so far.  I have quite a lot of rise in the low mid / upper bass part of my steady state in-room frequency response, but it sounds very flat.  Vocal intelligibility with music is excellent, not just at the MLP, but throughout the house.  (I wouldn't expect such great results from a single measurement with speakers that don't have the kind of stellar off-axis performance that these do.)

[andy497]

I'm kinda surprised that audyssey dynamic eq does not do any time-based dynamic compression.  In fact, when listening to music with dynamic eq on, I could swear I've heard the pumping effect from the attack/decay triggers getting tripped regularly with the beat.  Perhaps it was in the source material, but it seemed to go away when dynamic eq was switched off.  That was an unusual occurrence however.  Most of the time I like it.

[SME]

I'm kinda surprised that audyssey dynamic eq does not do any time-based dynamic compression.  In fact, when listening to music with dynamic eq on, I could swear I've heard the pumping effect from the attack/decay triggers getting tripped regularly with the beat.  Perhaps it was in the source material, but it seemed to go away when dynamic eq was switched off.  That was an unusual occurrence however.  Most of the time I like it.

 

Dynamic EQ does have a non-linear "dynamic" component.  The level of boost it applies depends on the relative level of the signal being boosted in addition to the master volume setting.

 

I find it interesting that you report hearing pumping with Dynamic EQ.  I believe I've seen others report the same, but I can't say I've noticed it.  I do notice lots of pumping due to excessive compression in actual content.  I listen to a fair amount of bass heavy electronic music, and it's all too common to have a big bass hit come in and briefly squelch everything else that's playing.  I try not to let it annoy me, but it's kind of hard to un-hear. 

[Kyle]

Dynamic EQ does have a non-linear "dynamic" component.  The level of boost it applies depends on the relative level of the signal being boosted in addition to the master volume setting.

 

I find it interesting that you report hearing pumping with Dynamic EQ.  I believe I've seen others report the same, but I can't say I've noticed it.  I do notice lots of pumping due to excessive compression in actual content.  I listen to a fair amount of bass heavy electronic music, and it's all too common to have a big bass hit come in and briefly squelch everything else that's playing.  I try not to let it annoy me, but it's kind of hard to un-hear. 

 

That sounds like the source being compressed right? Or maybe the 110v circuit running out of juice

[SME]

That sounds like the source being compressed right? Or maybe the 110v circuit running out of juice

 

Yeah.  I meant I hear it in the source material (i.e. content) all the time.  And this is true even at low-to-moderate listening levels, mind you.    This is just loudness war collateral damage and is unfortunately a widespread problem in what is out there.

[maxmercy]

IIRC, DynEQ has a one frame look-ahead time, so 1/24th of a second.

 

JSS