MiniDSP EARS measurement- REW auto EQ- Foobar Math Audio EQ playback

[sendler2112]

Tuning my HiFiMan Ananda Nano headphones. These new Superlight diaphragm planar headphones can be ultra revealing but tend to come with several break up modes which benefit from EQ. I'm using the excellent Math Audio Headphone EQ plugin for Parametric EQ and Headphone Crossfeed with Foobar as a player which also allows easy application of a WASAPI exclusive plugin to eliminate any involvement of the Windows Kernel Mixer and it's sonic degradation so I can pass HiRez PCM out the USB.

https://www.foobar2000.org/components/view/foo_headphone_eq

https://www.foobar2000.org/components/view/foo_out_wasapi

The MiniDSP EARS measurement jig is USB plug and play and measures with serial number specific cal files including compensation for the mics and its intrinsic 4.3kHz canal resonance, superimposed with the Harmon curve. A gated sine wave sweep is output via the excellent Room EQ Wizard software through your playback hardware and captured again by REW .

https://www.minidsp.com/products/acoustic-measurement/ears-headphone-jig

https://www.roomeqwizard.com/

Here is the resulting measurement of the Ananda Nano showing the corrected for my jig Harmon curve in the middle and the deviation of the Nanos from the curve at the top.

The measured frequency response deviation is shown above. The EARS HEQ curve in the middle is the target and the frequency response at the top would appear flat if there were no deviation from the curve which as you can see incorporates a Harmon curve with additional EARS response correction for the 4.3kHz Helmholtz resonance of the cylindrical "canal" of the unit leading in to the mics, and the serial number specific mic cals all superimposed.

The EQ section of Room EQ Wizard will then allow auto EQ of up to 20 filters, and display the predicted resulting frequency response deviations as referenced to the calibration curve that was used to create the preceding measurement.

And predicted phase deviations from the cal curve after EQ

I then type the filters into Math Audio. As for my particular Nanos and EARS

The composite correction looks like this. Disregard the 20kHz rise which I can manually chop down.

I always find the Harmon curve (and flat quasi anechoic loudspeakers) to be dim on top so to bring back the skin and cymbal strike of drum kit I then go through the graphic EQ that is built in to Foobar to fine tune for preference. I wish Math Audio had a tab to overlay an 8/16/32 band Graphic EQ screen right in that app to avoid having multiple things running.

There may be some contribution to my preference for some added highs due to the Sparkos Pro op amps that I have fitted to my Burson Conductor Reference SE which tend to be a bit polite on top compared to the Burson Vivid op amps or OPA1656. But the transparency and fine detail and power in the bass of the Sparkos Pros is unmatched. I compared them again to the stock Conductor 3XP with Vivids (I had a set of Sparkos Pro in that unit first but they don't fit under the lower case lid so stole them for the Reference which has a taller case and plan to make a new set for the 3XP without the adapter boards to get them to fit) and just as quickly unplugged from it.

The HiFiMan Ananda Nanos are a bargain at $600 and can be elevated to an entirely new league with some EQ where they then show the best of the impressive fine detail that an ultra light planar film can provide.

https://hifiman.com/products/detail/334

 

[sendler2112]

I noticed that the photo above of the cumulative filter set after transcribing it to Math Audio included some manual adjustments I had fooled around with in an attempt to avoid stacking the Foobar GEQ on top of it so I went back in and verified all of the values straight from the chart that was produced by REW in an attempt to prove the simpler procedure of just trusting the numbers as referenced to the cal file sounded good with easy fine tuning by ear in the GEQ . This is the actual curve as dictated by REW. It still intuitively looks a bit low at 18k to me compared to the initial measurement but sounds really good on all material with the final tweaks of the GEQ.

 

[sendler2112]

Here is the Harmon curve that was determined by that team through listening tests to be the prefered frequency response for headphones in order to best approximate what the listening panel expected the various instruments to sound like . You can see in comparison to the MiniDSP EARS Newer HEQ calibration target posted above that MiniDSP added some level at 4.5kHz to compensate for the canal resonance that arises from the cylindrical canal that leads through the rubber ear. I have been searching for more critique of this simple looking cal curve but get no replies from the MiniDSP admins on their own forum or by any hobbyists anywhere else. I am surprised that there is little to no interest in this affordable measurement system. The pros bashed it from day one compared to the $8,000 GRAS unit but it seems to me that with a dialed in cal file, it could really be very accurate in allowing home users to quickly build a great EQ for any headphone. For $300 plus some $50 programs.

Headphone.com uses a different target. The Diffuse Field target attempts to approximate the frequency response of loudspeakers in a room through your headphones. Here is their measurement of the Ananda Nano with a mic cal only so where as my measurement already included the HEQ curve with mic cal and would ideally be a flat line, theirs should ideally follow the curve. Not sure what their reasoning is to have a desired bump at 8kHz. As measured on the Bruel and Kjaer 4128 system.

https://www.bksv.com/media/doc/bp0521.pdf

Again, this is what I came up with on the MiniDSP EARS which was quite repeatable on different days and not so careful placement matching as referenced to the calibration curve where only the deviation to that curve is shown as non flat..

Here is what I get using just the raw mic cals. Pictured are 3 runs overlaid and then repositioned the headphones a little higher (the pads were resting on the ears) and did 3 more runs. Compare this to the Headphones.com measurement to see the contribution of the cylindrical canal resonance at 4.3kHz.

The Nano version of the Ananda is too new to find many other posted measurements from the usual sources to compare.

 

[sendler2112]

Auto EQ will only be as good as the measurement and calibrations. So this brings up the question of the different measurement systems and theories of human hearing. The MiniDSP EARS measuring jig was pretty much summarily rejected by the ruling headphone measurement gurus when it came out as a cute toy but I am finding it to be very consistent. And, consistency of its own repeatability is the most important attribute of any measurement system since each type of jig seems to have its own major idiosyncrasies.

Headphones.com posted a chart of the Arya Organic with side by side comparison of the same headphone on two different measurement systems which each apparently even recommend different target curves. The B&K rig on the left is shown with a "Diffuse Field" target and seems to be the one I usually see Headphones.com using. And the GRAS system on the right is shown against the "Harmon" target. Which one of these kilobuck measuring systems and targets will provide an auto EQ that sounds the most pleasing with the least amount of manual tweaking? Is one or the other of these systems an inaccurate toy compared to the other? Or would a dialed in target curve, taking into consideration the repeatable idiosyncrasies of the different systems make each one always provide essentially the same subjective listening result compared to the other? And therefore, wouldn't the $300 MiniDSP EARS measurement system be just as good for creating auto EQs, despite its 4.3kHz Helmholm resonance, if it had its own dialed in calibration curve?

Here are the Headphones.com measurements of the Arya Organic. As you can see, each of these expensive measurement systems shows very different results above 5kHz but are nearly identical up to 3k aside from a slight additional rise in the B&K chart from 200 to 1kHz..

I don't have an Organic and there are very few posted measurements of the Anada Nano yet (although at $600, this is going to turn out to eventually be a big seller). Headphones.com did post a measurement for the Nano on the B&K and then here is my measurement of 6 runs with repositioning of my Nano via the EARS.

Looking at all four charts, It seems that the B&K is adding some extra peak at 4k, 6.5k and 8.5K that the EARS and the GRAS are not showing. The EARS will always tend to obscure any fine differences around 4.3kHz but otherwise looks very similar to what a GRAS like measurement of the Nano would be and I'm sure someone will eventually post a measurement of the Nano on the GRAS.

MiniDSP did produce a newer target calibration, HEQ which made a bigger correction for the 4.3k resonance and added more downslope, after first selling the unit with the earlier HPN curve.

The new HEQ calibration curve in the middle:

And the earlier HPN target:

But both curves look oversimplified for what would be ideal and to incorporate the well known Harmon curve. And they seem to have largely given up to the negative hype and lost interest in spending time to come up with something better that could really dial in the EARS cal curve and make it compete at a world class. After trying it, I feel that computer EQ of these modern, very revealing but peaky headphones, is a big improvement. And also believe the the EARS measurement system could be quite adequate with a little more hobbyist development of the calibration curve.

 

[sendler2112]

I ran the Ananda Nano REW measurement calibrated to the MiniDSP HPN early curve again this morning. Here is the measurement.

Here are the results of the REW auto filter. I set the matching target to 1db to keep the number of filters simpler than what happens if you try to make the filtered response perfect down to 0.5db where the program fights to eliminate every wiggle.

I zeroed the first 4 filters that were reducing below 200hz because I new I was going to want to actually increase this frequency range after listening.

Here is what the correction looks like in Foobar Math Audio EQ after I added a bass low shelf at 72Hz that was tuned in with the sliders while listening to music. I can probably also dial in less boost for the big suck out at 11.5kHz to avoid throwing away so much level. I'm not sure why REW is doing so much with 12 and 13?

And finally, here is the Foobar Graphic EQ override. Keep in mind the Sparkos Pro op amps I have In my Burson Performance SE tend to sound polite above 10kHz. I would now say that I like the subjective sound as corrected by the earlier HPN curve that MiniDSP first came up with to correct the EARS ' 4.3kHz canal resonance back toward what would be a Harmon curve without so much decrease in the Highs. The way I like it so that drum kit can cut through the mix. It looks from examining this override that without the Bass shelf I added in the Math Audio Parametric below 150Hz, the HPN calibration might actually be almost dead on with just some simple additional broad scoop in the common area that we often prefer to dial in between 200-800Hz.

I looked at the text file for the HPN but they list the filter levels in frequency/ level/ "PHASE". Instead of "Q". So I would have to figure out how to write the scoop in those parameters to start dialing it in. At which point I believe I would have a Calibration curve that along with the individual unique mic cals that MiniDSP provides, would allow a completely no fuss auto EQ of any headphone in just a couple minutes.

 

[sendler2112]

I tamed the big peaking filters down to get the average level up so I wasn't throwing away so much. And adjusted the bass shelf up and low-mid scoop down a little more and got rid of the stacked GEQ. Now it looks like this. I made it a conservative EQ in the 5kHz and up so it sounds more accurate than more Better from artificial highs. Going back and forth to the bypass, I'm very happy. Both with the EQ and with these headphones. The EQ allows the Ananda Nano to shine through with amazing clarity of soundstage and fine detail.

The disappointment with the Math Audio app is that the entry fields for the filters do not populate with the values. So once you tune something by ear with the sliders, which are a very handy addition that other programs do not have, there is no easy way for me to notate or print my settings for someone else to try. And this is great for playing HiRez files through WASAPI Exclusive with USB out to the Burson but I now also need a way to get EQ for streaming audio which does not go through Foobar in order to pick up the EQ.

 

[sendler2112]

So, this is what happens when you discover that the gated sweep signal that Room EQ Wizard measures from can be played from a file in your player and through the EQ. And have 99 filters available in the Math Audio EQ plugin for Foobar. You sit there adding filters and running measurements, trying to come up with the perfect measured response. I have decided that the HPN calibration file that MiniDSP originally came up with for the EARS measurement jig is actually pretty close to my actual listening preference from the midbass up to 10k. I did add some shelf to the bass below 150Hz and decreased a shelf above 10kHz compared to the HPN cal curve but have found that an auto EQ to this curve is a quick way to start. Here is the measurement through the EQ in Foobar after some listening tweaks as correlated to the cal curve which corrects for the 4.3kHz resonance of the molded ear canal of the jig, and adds the 3-4kHz dome and HF roll off that is know from the Harmon curve.

Here is the raw measurement with only the mic cals.

And here is the EQ.