Try my cross-feed network

Solderer · Jul 5, 2016

Here is the design of a simple passive circuit to improve listening to common stereo recordings with headphones. Naturally my own subjective response to using it is very positive, so I want to convince you to build one for yourself and confirm or dis-confirm listening improvement. My judgement is colored by the fact that the network is my original design. Of course you are welcome to comment on the design itself without having tried it.

My network is placed between the power amplifier and headphones (in cascade). The component values of my network correspond to the rated impedance of the headphones plugged into the output of the network. It produces only slight reduction of volume, so what you are using presently to drive your headphones will be fine driving both network and headphones.

The inductors of the network (two) can be commercial ones intended for speaker system crossovers. This will be easier to try with headphones of 32 Ohm rated impedance or less as L1 of the network corresponding to 32 Ohm equals 1 mH. The combined load of network and headphones seen by the amplifier at a minimum equals the impedance of the headphones.

There is a full explanation of how my design was derived here (my posted article):

http://www.zen22142.zen.co.uk/Design/Xfeed-Net.pdf

The schematic diagram of the network and equations for calculating component values is below.

-S

RBFC · Jul 5, 2016

Welcome to WBF! Please take a moment and tell us a bit about yourself and your audio journey.

Lee

Solderer · Jul 6, 2016

While not a professional, I've done quite a bit of self-study of audio reproduction, both with loudspeaker systems and headphones, and also electronics generally. My design of the cross-feed network for headphones is the result of a great deal of experimentation last year and my studies over many years past. My primary reference for the design of the network, the book Hearing by Gulick and others, I first looked at something like 20 years ago.

-Pete

dalethorn · Jul 6, 2016

I've found, after equalizing approx. 150 headphones over the past 5 years, that rather than smearing the sound with a mixing algorithm, it's better to identify the more important aspects of frequency response and timing (and resonances) to achieve a natural sound, as would be heard with live music. As far as I know, most recordings over the past couple of decades, and many over decades before that, have been mixed down for playback on headphones and speakers. Today, serious undistracted hi-fi listening on loudspeakers is a very marginal activity compared to headphone listening. To think that recordings would ignore that is untenable.

Solderer · Jul 7, 2016

Dalethorn,

In my article (see the link in my opening post), I argue that the recording mix must be very different for proper reproduction by either a pair of loudspeaker systems or headphones. I don't think that one type of recording would be correct for both simultaneously.

That is news to me that recording engineers now are attempting to record primarily for reproduction with headphones. Even if that is true, there is a vast library of music that has been recorded with stereo reproduction with speakers in mind.

Regards,
Pete

dalethorn · Jul 7, 2016

I wish I had the time to dig out the links for you, but I don't and anyway, if you're curious you can certainly find them yourself. Or you can take a logical shortcut and look around. It's a different world from the days when large room-filling speakers were the norm for audiophile listening. AS IF audiophile listening were EVER the object of recordings released by Capitol, Columbia et al. And in the final analysis if you don't look into my experience in reducing the resonances to achieve a more natural sound, then what you'll be doing is taking a bunch of vitamins while eating an atrocious diet. Make the best fix first.

Solderer · Jul 7, 2016

dalethorn said:
And in the final analysis if you don't look into my experience in reducing the resonances to achieve a more natural sound, then what you'll be doing is taking a bunch of vitamins while eating an atrocious diet. Make the best fix first.

Q of series resonance in the direct signal path of my network is a benign 0.65. For example, for the left channel, reactance of L1A at the resonant frequency of 5 kHz is equal to RL (rated headphone impedance). Resistor R3A equals about 0.5 times RL. Q equals XL1A w.r.t. 5 kHz divided by (R3 + RL), or RL divided by 1.5RL. That is, Q calculates to about equal to 2/3.

dalethorn · Jul 7, 2016

Solderer said:
Q of series resonance in the direct signal path of my network is a benign 0.65. For example, for the left channel, reactance of L1A at the resonant frequency of 5 kHz is equal to RL (rated headphone impedance). Resistor R3A equals about 0.5 times RL. Q equals XL1A w.r.t. 5 kHz divided by (R3 + RL), or RL divided by 1.5RL. That is, Q calculates to about equal to 2/3.

I can't guess what the acoustic peaks and dips are outside of the drivers, but the housings in the earcups have an enormous influence on the sound, as do the earpads, and 'influence' means non-neutral.

Solderer · Jul 7, 2016

dalethorn said:
I can't guess what the acoustic peaks and dips are outside of the drivers, but the housings in the earcups have an enormous influence on the sound, as do the earpads, and 'influence' means non-neutral.

Do you find "an enormous influence on the sound" by means of listening tests, or are you measuring such as frequency response?

You own 150 pairs of headphones?

dalethorn · Jul 7, 2016

Solderer said:
Do you find "an enormous influence on the sound" by means of listening tests, or are you measuring such as frequency response? You own 150 pairs of headphones?

The best of headphones have acoustic resonances that produce peaks and dips in the response. The Senn HD800 for example is very smooth comparatively, but still a bit uneven. Rather than measure or depend on a fixed schema, I start by running a combination of frequency tones and listening for big changes from one freq. to the next, randomizing the samples so as not to build a "memory". Part of that analysis is using 2 or 3 other reference headphones at the same time to net out any hearing variances. That's an iterative process, cumbersome, but works well. Once the worst peaks and dips are identified, I experiment with the parametric eq to partly correct them and zero in on a more natural sound. Getting a more natural sound is less affected by individual hearing differences, because you're not eq'ing for a flat response in the end, you're eq'ing for a natural sound, and whatever you hear in live music, if your hearing is more-or-less normal (undamaged), *is* natural. I try to leave the headphone's basic flavor as is, i.e. warm, detailed, analytical etc., but if there's a significant midrange coloration then I correct for that.

Once you get into this, if your goal is to eq-out the unnatural parts of a headphone's sound - not completely or perfectly necessarily, but to a significant extent, then you will get a more natural sound and an open, pleasant soundstage that you can never achieve with mixers and cross-feeds. The key is to understand what's possible and how to proceed to get there, rather than follow a path to technical neutrality. It really works, and I think most of my published curves could serve as a good starting point for someone who has a matching headphone and a parametric equalizer.

One caveat: If you check Innerfidelity's published curves, you'll see some pretty amazing differences in different samples of the same headphone model. Not always, but sometimes, and some of those are expensive.

dalethorn · Jul 7, 2016

Solderer said:
Do you find "an enormous influence on the sound" by means of listening tests, or are you measuring such as frequency response? You own 150 pairs of headphones?

I have owned more than 150 since 2011, but I keep only about 20 at a time. Most of the time I need only 3 or 4 for reference when EQ'ing a new headphone.

Solderer · Jul 9, 2016

dalethorn said:
Getting a more natural sound is less affected by individual hearing differences, because you're not eq'ing for a flat response in the end, you're eq'ing for a natural sound, and whatever you hear in live music, if your hearing is more-or-less normal (undamaged), *is* natural.

With the parametric equalizer, try boosting response at 5 kHz by +9 dB in both channels. If you have adjustable Q, that should be set to about 2.5. Let me know if this produces a natural sound. The boost at 5 kHz is a very important part of the processing of my network in addition to cross-feed. By the way, I'm not the first to propose such signal processing.

dalethorn · Jul 9, 2016

Solderer said:
With the parametric equalizer, try boosting response at 5 kHz by +9 dB in both channels. If you have adjustable Q, that should be set to about 2.5. Let me know if this produces a natural sound. The boost at 5 kHz is a very important part of the processing of my network in addition to cross-feed. By the way, I'm not the first to propose such signal processing.

Here's a quick sampling from my website /Photos/Audioforge, of EQ's that I've done on some flagship headphones. Headphone response is all over the place and none are very much alike. The real fix is to address each headphone's peculiar responses rather than try a "one fix for all", which never works.

Solderer · Jul 9, 2016

dalethorn said:
Here's a quick sampling from my website /Photos/Audioforge, of EQ's that I've done on some flagship headphones. Headphone response is all over the place and none are very much alike. The real fix is to address each headphone's peculiar responses rather than try a "one fix for all", which never works.

Your graphs look great, but at least for me I can't say what they show. By way of criticism to be helpful, if you want to communicate to others what you are doing, you need to provide more details IMHO. For example units of frequency and decibels of the X and Y axes aren't given.

dalethorn · Jul 9, 2016

Solderer said:
Your graphs look great, but at least for me I can't say what they show. By way of criticism to be helpful, if you want to communicate to others what you are doing, you need to provide more details IMHO. For example units of frequency and decibels of the X and Y axes aren't given.

I told you about my website. Did you look? Every question including settings is there in detail. Tutorials as well.

Photos are under photos, hifi misc has the eq settings and tutorial. And there's tons more.

Solderer · Jul 11, 2016

dalethorn said:
Here's a quick sampling from my website /Photos/Audioforge, of EQ's that I've done on some flagship headphones. Headphone response is all over the place and none are very much alike. The real fix is to address each headphone's peculiar responses rather than try a "one fix for all", which never works.

My suggestion is to firstly "address each headphone's peculiar responses". Do you mean by that to obtain flat frequency response as much as possible? Then, secondly, after flattening response, if that is what you mean, make the manipulated headphone response to some extent match what would occur if the listener were listening to reproduction by a pair of stereo loudspeaker systems at a distance. To me, this seems like a much more reasonable approach, rather than in hit or miss fashion with an equalizer trying to obtain natural reproduction.

-Pete

dalethorn · Jul 11, 2016

Solderer said:
My suggestion is to firstly "address each headphone's peculiar responses". Do you mean by that to obtain flat frequency response as much as possible? Then, secondly, after flattening response, if that is what you mean, make the manipulated headphone response to some extent match what would occur if the listener were listening to reproduction by a pair of stereo loudspeaker systems at a distance. To me, this seems like a much more reasonable approach, rather than in hit or miss fashion with an equalizer trying to obtain natural reproduction. -Pete

Flat is similar to the goal, but not exactly. And hit-or-miss is definitely NOT what you want. Headphones have issues, such as resonances or other limitations in the cups that produce uneven sound. If you believe that flat response, whether programmed by a machine or by listening with test tones, will produce a natural sound, no - it likely will not. To achieve results like what I've gotten, which reduce peaks and recesses and greatly improve the natural soundstage, you need to do basically two things at the same time, or side-by-side. One is to use a set of discrete test tones, in a more-or-less random manner (to prevent tester's memory) to detect large peaks and recesses, and then adjust for those, partially or mostly, but don't overdo it. The second thing is to listen to music that 1) you're extremely familiar with, and 2) that contains good reference tone in all ranges from upper bass to lower treble, to detect colorations, and then work on adjusting those. When you make an adjustment, you save the new settings under a name such as 'Test1', 'Test2' etc., so that you can toggle between those while listening to hear which is better. I also keep other reference headphones handy so I can make sure I'm heading in the right direction.

Now, the dream of most headphone-sound tinkerers is to be able to get a more natural sound more easily, with less work. You can buy into that idea, but you'll lose out that way. It's not easy until you get a lot of experience, and even then some headphones are very difficult to EQ. On the other hand, since I publish more than 120 sets of EQ's, people who have those headphones can use my data, at least as a good starting point for what they will end up with.

Edit: I have EQ'd some desktop speakers, but not room-size speakers as yet. However, the principle is exactly the same, and the end result will be the same - the most natural sound. Someday I hope to sit with someone's speaker system and room for a couple of days as a test. Confidence is 100 percent.

Brucemck2 · Jul 12, 2016

dalethorn said:
On the other hand, since I publish more than 120 sets of EQ's, people who have those headphones can use my data, at least as a good starting point for what they will end up with.
.

Looks like Fab Filter Pro in the graphs?

Where would I find your recommendations for (older) Sennheiser 800s or HiFiMan he1000s?

dalethorn · Jul 12, 2016

Brucemck2 said:
Looks like Fab Filter Pro in the graphs? Where would I find your recommendations for (older) Sennheiser 800s or HiFiMan he1000s?

I use Audioforge for iOS, but any parametric equalizer should do. These graphs are missing 300 and 3000 hz vertical lines, but all the others are there. The horizontal lines represent 6 db plus or minus on the vertical axis. The actual settings with 'Q' values are in the 'HiFi Misc.' page on my website.

I haven't heard the HE1000, but I gave up on planars as all the ones I owned or borrowed had treble harmonics problems. I owned the HD800 from June '09 to March '14, before I started with the equalizer. I would like to work with the HD800 sometime, and based on my tests and reviews, I'd probably boost the low-mid bass about 3 db, and cut somewhere in the treble by about the same amount. But getting the best sound means the midrange has to be right, although the HD800 (and probably the HE1000) need little or no adjustments to the mids.

Try my cross-feed network

New Member

WBF Founding Member

New Member

Headphone user

New Member

Headphone user

New Member

Headphone user

New Member

Headphone user

Headphone user

New Member

Headphone user

New Member

Headphone user

New Member

Headphone user

Member Sponsor

Headphone user

Similar threads