Target Curves, should What You See be What You Get?

[Le Concombre]

Puzzled by the Harman Curve presented here : http://www.audioheritage.org/vbulletin/showthread.php?39134-Harman-Target-Curve and attached in a smoothed version here, I have aligned the Time Windowed Vector Average of a bunch of measurements one can make in REW to this 16 dB wide sinuous Harman Curve.

Turned out that when the convolution filters obtained this way are applied to Steady State measurements (non FDWed RMS Average or MMM) we get the classic B&K with a 6dB downward tilt.

I believe I have reverse engineered what Harman does (and maybe, under the hood, what dedicated software do). Why take this complicated route ? While from Pink Noise + RTA times B&K up to Harman's RR1 all much talked about Target Curves describe desired profile at Steady State, more advanced acoustic research (Toole, Olive, etc) suggests to rather correct early stage captured by time-windowed measurements, hence taking the VA with its peaks and dips not exactly the same as those at Steady State.

Aligning REW's VA to Harman Reference Room 10 dB target yields convolution filters that translate RMS Average into EBU/ITU.
Aligning REW's VA to Harman Reference Room 10 dB target up to 160 Hz and to Harman Curve above yields convolution filters that translate RMS Average into B&K.
Aligning REW's VA to Harman Reference Room 12 dB target up to 160 Hz and to Harman Curve above yields convolution filters that translate RMS Average into Harman's RR1.

Have fun with the attached REW ready targets. The graph illustrates that aligning REW's VA to Harman Reference Room 10 dB target up to 160 Hz and to Harman Curve above yields convolution filters that translate RMS Average into B&K. Of course, by nature and especially for higher frequencies, it's not a perfect B&K but it's nevertheless a +- 3dB B&K sounding better than if made directly on the steady state measurement for what has been corrected and not corrected is more relevant. Plus correcting Time and Amplitude domains in REW + RePhase combo on the same basis yields much neater, Impulse, steps, GD etc than if you correct Time on VA and Amplitude on a Steady State measurement. I advise you to manually finalize the Amplitude correction in RePhase because of discrepancies with REW in the steep raise region.

I suggest you make the 3 sets for what matters is the fit between creation and listening conditions and, unless you have excellent on fly tone controls, you need several operational room response options, in hope of recreating the conditions in which the Art was created.

Personal tastes, audition oddities, loudspeakers behaviour etc etc : can’t say there’s no argument for personal operational room response but I hereby suggest advanced ways to have at hand EBU/ITU standard, historical B&K and Harman’s RR1 of most recent fame, that all are likely to have been in use in good mastering facilities one time or another.

 

[Le Concombre]

Has anyone tried this recipe ?

I’m really happy with the set I created based on ERB smoothing. BTW, equivalent of the above attached graph is likely to look much messier in the steep raise region with VAR smoothing; so what?

I'm currently enjoying all target curve options (EBU/ITU standard, historical B&K and Harman’s RR1) one time or another.

Maybe because I have been much into classical and SACD these days and mainly in Mch DVDA otherwise, and maybe these are worlds were standards are more readily followed : I favor EBU these days

Please note that when I state that aligning REW's VA to Harman Reference Room 10 dB target yields convolution filters that translate RMS Average into EBU/ITU, I mean following a 6 dB downward slope from 2K to 20 K, not yielding a 20 Hz 20 K flat line +- 1 dB

I sure do appreciate RR1’s bass extension sometimes but I’m puzzled when I see it touted as the panacea in conjunction with SPL in the 77-83 dB range. I wonder if it’s not, when selected as unique and systematically used, then masking issues or acting as sort of loudness. The often-quoted experiment in which RR1 shone does not, among other criteria, specify the SPL of the test.

Don’t know what a SPL in the 77-83 dB C Slow range meets in the real world. If it’s meant to be a reference to Bob Katz’s system, it’s misleading since the K system suggests to set the 0 dB VU for each channel at 83, so 86 both. In real world that meets a piano peaking around 0 dB VU (so around 86 ) and a Beethoven’s symphony hitting + 6 dB VU (yes we can in the K system) at its around 92 dB peaks, values I effortlessly reproduce with my system.

 

[Le Concombre]

twimc :
I simplified my quiver to sets aligning the VA to HRR 10 (attached) and 12 (looks alike but 12 dB elevation).
Applying them to RMS averages yield the attached results consistent IMO with ITU-EBU/B&K for 10 and RR1 for 12 (yes, a 12 dB amplitude applied on the VA translates into a 9.5 on the RMS)
The sets are "perfectly" aligned to HRR, at least much more closely than the convolved RMS appear to match EBU or RR1, respectively ; those discrepancies are expected by design
Both sound very good, depending on material, and much better than perfectly aligned RMS

 

[BillG]

Since I own speakers manufactured under their umbrella, it's not an issue for me since they're already tuned to it. They're Infinity Reference 162s and 152s, by the way.

What I've been doing lately, though, is tuning my IEMs in loose proximity to their curve. The reason I write "in loose proximity" because I don't accentuate the bass and treble as much as their target. I'm using simple Graphic EQ to accomplish it.

 

[Le Concombre]

I have no experience with IEM and little with headphones. My AKG 701 did not sound very well with a bass boost attempt : transducers matter and limit corrections. You might want to give Sonarworks a try

My speakers have remarkable directivity behavior (see anechoic measurements attached) and I would not put money in expensive stuff with no anechoic measurement published (seen any for the Wilson that so many audio gurus have?). So do yours probably and they then probably follow the right in room response and look good in Harman's Spinorama. Nevertheless you should measure their actual responses in Your room and mind the matching of both channels.