Crossfeed

home
back to Learn

post separation

Limitations of crossfeed

A lot has been written about crossfeed. Some people like it and swear by it and others don’t like it at all.

Crossfeed basically is a frequency dependent ‘mixer’. Low frequencies are mixed to mono. The higher the frequency is the more ‘stereo’ it becomes. The highest frequencies will never be ‘fully stereo’ though, just less ‘mono’.

Crossfeed will never widen the stereo image. When the stereo image is wider than this effect is not crossfeed but some kind of stereo expander.
Crossfeed will ‘narrow’ the Left-Right effect and make the image smaller.

Crossfeed is an attempt to ’emulate’ speaker stereo somewhat where the stereo image itself is in front of you and not to the left and right.

Most music is recorded so it sounds good on speakers. On headphones the stereo image will be much wider.

Don’t expect crossfeed to give you a ‘speaker-like’ experience where the sound is coming from in front of you. It won’t … it will be between the ears.

Crossfeed is an electrical ’emulation’ of the stereo width which has some resemblance to how speakers ‘mix’ the left and right signal.
Do realise that mixing soundwaves and electrical signals is VERY different.
For example: when you mix a left and right signal where for instance the left and right signal of a certain (narrow) frequency band happens to be in counter-phase they will electrically ‘cancel’ each other out.
Because stereo is all about phase differences for higher frequencies this WILL happen in a lot of recordings.
With speakers, however, the left and right speaker will BOTH emit sound be them in counter phase. Due to the way sound propagates and ‘mixes’ there won’t be sharp ‘nulls’ as when mixed electrically.
The treble, due to reflections and the listener not being exactly in the middle etc. will thus differ when listening to speakers and to crossfeed.
Crossfeed thus will always be a flawed ’emulation’ of speakers.

Crossfeed solutions from different designers will sound different because of the way the signals are mixed, frequency bands could differ, the amount of mixing L and R differs.
Crossfeed can be made according to theoretical calculations based but can also be made in a less ‘mono-fied’ way. You could call this a ‘wider’ version or more subtle crossfeed.

Personally I prefer crossfeed with some headphones and also not always and with all music.
Depends a  bit on my mood as well.
I can listen longer to music played though (subtle) crossfeed. It can prevent listening fatigue on some headphones.

post separation

Crossfeed module for Kameleon amplifier.

xfeed PCB

Most music is recorded so a good ‘stereo’ image is achieved while playing the stereo signal on speakers.
The problem with headphones is that the same stereo signals will be exaggerated by quite some margin.

Instruments that are panned hard to the right (or left) will only be present on the right (or left) channel

The stereo image thus is kind of ‘wide’, where mono signals. such as bass and most vocals. are heard in the middle of the head (mono).

When using speakers stereo is heard quite different. Signals that are played on the left channel only are heard with both the left AND right ear. Where on headphones, that signal is only heard in the left ear and not on the right.
Our ears(or more exactly brains) derive sound location from the higher part of the frequency range.
This is also the reason why most systems with a subwoofer have only one of those. The ears can’t easily detect where low frequencies come from.

For higher frequencies it is easier to hear where they come from. This is for 2 reasons. The shape of the head and ears and something called time-delay. The higher a frequency is the shorter the wavelength of the frequency is.
When a sound source is from the left speaker in front of us the soundwaves will ‘hit’ the left ear fractionally earlier than the right ear.
For lower frequencies the differences in time between the left and right ear are impossible to detect because the time difference is much too short in relation to the wavelength. The brain cannot resolve these minute differences.
The higher the frequencies become the closer the wavelength will come to the time difference and the easier it gets for the brain to ‘measure’ the time difference = localization.

Because of the shape of the ears and head, higher frequencies will also be attenuated differently for sounds coming from one side or the other side. Simply because there is an object in its path (head). The way the earlobes are shaped also makes them ‘directional’ and even more sensitive in certain parts of the frequency range.

A low frequency coming from the left will be heard equally loud in the left ear as in the right ear.
The higher the frequency becomes the lower the amplitude (loudness) will be on the right ear where the left ear will hear everything equally loud.
So for higher frequencies the frequency and amplitude response will differ depending on WHERE the sound comes from where for low frequencies there will be no audible differences.

Because speakers are not located on the far left and right of us but usually are positioned more or less in front of us the effect ‘blocking’ effect will be smaller than when speakers on the far left and right are used.
There will thus be mostly temporal (timing) cues for higher frequencies and some amplitude cues as well to derive a stereo image. Bass from the left speaker will be equally loud in the left and right ear where high frequencies will be louder in the left ear than in the right ear AND will have some temporal differences as well … when listening to speakers.

Headphones do not ‘mix’ the left and right signal. This makes speaker listening in a lot of cases less ‘tiring’ compared to headphone listening. The ‘extra wide’ spatial cues can cause listening fatigue for some people. Others may not be bothered by the L/R separation.
The values below are for the ‘theoretical’ crossfeed. That is without C2 and R7.

 

X-feed schematic

Crossfeed addresses this issue. The result will thus be a less wide stereo-image INSIDE our head than when no crossfeed Crossfeed only makes the stereo image smaller, less hard panned left and right with a part in the ‘middle’ of the head. Everything will be positioned more ‘in’ the head. For some people this will make some stereo recordings less ‘tiring’ to listen to for longer time periods.
The crossfeed module thus ‘mixes’ the left and right signals for the lower frequencies to the be the same level.
All stereo signals below 800Hz will thus become ‘
mono‘ essentially.
The higher the frequency will be the less ‘mono’ and the more ‘stereo’ it becomes. For lower frequencies L = R.

Above 1kHz, increasing with the frequency, less and less of the left channel signal will be ‘mixed’ with the right channel and less and less of the right channel will be mixed into the left channel.
At the
highest frequencies the ratio will have become 4:1 where for the lowest frequencies the ratio will be 1:1.
This mono-fying the sound ‘dulls’ the clarity a bit as well, similar to listening to a radio in stereo and then hit the ‘mono’ button. The ‘sparkle’ also is removed when the signal becomes to ‘mono’.

To counter this a little ‘twist’ was added to this module. The highest frequencies (>10kHz) are also mixed in as a normal stereo signal. This counters the effect of the lost ‘air’ and treble extension which is a downside of most crossfeed circuits.

Whether one chooses to add this little twist (which basically un-crossfeeds the signal for the highest frequencies only) can be determined when the filter is built. See it as an ‘option’ which you can always add or remove later on as well.

What crossfeed does to the sound is a create a more narrow ‘stereo image’ inside the head. Less extreme left and right.

This makes listening to most ‘studio recordings’ less fatiguing.

As this crossfeed is a filter module it can’t be combined with EQ filter modules. Thus the crossfeed can only be used with headphones that do not need a lot of EQ.

post separation

Here’s a simple and fun project / circuit in case you would like to try crossfeed. It’s based on a Linkwitz crossfeed.

There is a big difference in crosstalk and crossfeed (a.k.a. X-feed).
Crosstalk is a technical term for bad channel (left and right) separation.
Like the sort you get with vinyl and sometimes tape recordings where the recording head and playback head had a different height position on the tape and tuners.
a small part of the left signal enters the right channel in this case and vice versa.
If there is 100% crosstalk the signal is mono.
It’s mostly frequency independent so bass mids and treble have an equal amount of crosstalk AND there is no phase shift/delay in the signal.

In crossfeed only the lower frequencies are a slightly ‘monofied’ (a certain percentage) and the mid to high frequencies are affected increasingly less but do get a slight phase shift when transferred to the other channel. This causes the highs to drop a bit in level for 2 reasons, namely the adding of the lows and lowering altering of highs due to them becoming mono. This effect is also easily heard on FM radio when we switch from stereo to mono we hear less highs (treble reduces).

On higher frequencies, when a speaker is located on the right, the right ear ‘hears’ the speaker a fraction of a second earlier and the left ear hears the sound slightly delayed because of the bigger distance of the right loudspeaker to the ear.  The speed of sound waves in air (approx. 330m/s) is responsible for this. Also the highest frequencies are attenuated a bit because the left ear is slightly shielded by the head and the ear is listening ‘outwards’ so there is also a smaller HF part compared to the LF part which is another clue for our brain to pin point where a sound is coming from..
Delay/phase shift for the higher frequencies is mimicked by the capacitors/resistors in the circuit.
For the left speaker the same principle applies but for the other ear of course.

As crossfeed is only an approximation of this phenomenon the effect is not exactly the same. For some people these circuits can create a more realistic ‘soundstage’ but it won’t change the ‘in the head’ sound experience to an ‘out of the head’ or an ‘in front of the head’ listening experience for most. It can create a more natural and less fatiguing listening experience for some.

It cannot be used for binaural recordings made with a dummy head or a Jecklin disc for instance as these recordings already have real crossfeed ’embedded’ in the recording.

On (studio) mixed music where every instrument has gotten it’s ‘place’ in the stereo image, by the pan adjustment on the mixing console, a positive effect can be heard depending on the recording.

Whether one likes the effect or not is something of a more personal nature and has to be tried personally, don’t take anyone’s word for it… try it to see if it suits your listening experience

How I got to this circuit.

I had build some ‘standard’ crossfeeds before (the medium and high version) but found the effect too ‘obvious’ for my taste.
So came up with the idea to replace the resistors with pots and twiddle around with various settings to find out what my optimal setting was.

In this picture you can see part of the schematics and the actual build.

variablecrossfeed

This is the complete schematic of the amplifier it was built into (for those that are interested).

SSX-feedHPamp

After weeks of listening, switching, and trying different settings I came to a setting I liked on most music so I could leave it on all the time. I measured the settings of the potmeters and the noted the closest values.  These values are given as the ‘subtle’ crossfeed in the scheamtic below.
The effect is very subtle indeed but since HP listening is a very personal experience I recommend playing around with some of the resistor values.

Xfeed

Put this circuit between a source (PC, MP3 player, DAC, CD e.t.c. and the input of your headphone amp. The headphone amp needs to have an input resistance of 10kOhm or higher (most are close enough to this).
It can NOT be used between the output of an amplifier and the headphone itself.

3 different ‘levels’ of crossfeed have been calculated and it is up to the builder to see which you prefer. If you like less bass then omit CAP5.

The capacitors and resistors are pretty standard values (resistors 0.25W to 0.5W)

This is how you connect it between a cut RCA cable.

connectXfeed

For those that want to bypass is without having to take it out of line here is how to connect a bypass switch to the board.

connectXfeedbypass

Use a DPDT (Dual Pole Dual Throw) switch and connect it like described above.
This is the easiest but not most elegant solution.

There is a loss in amplitude level when using this filter and is inherent to it’s workings. This should be taken into account when bypassing the circuit.
In the amplifier a small potmeter was placed in the bypass path working has an attenuator. It is set to have the same attenuation as the crossfeed circuit to prevent the quite noticeable level difference when bypassing the crossfeed.

Here is how this can be implemented.

connectXfeedbypassdeluxe

Pictured is a single (Bourns or other manufacture) 10kOhm ceramic Pot. Size doesn’t matter. Other models and/or types can be used too of course as long as it is 10 kOhm linear. The switch is a 4 Pole dual throw switch (4PDT).

The boards are all sold out.
post separation

home
back to Learn

Comments
  1. Anonymous says:

    Incorrect. Digital crossfeed filters do separate the highest frequencies in full stereo. At the lowest frequencies, they never are full mono. The signal is the opposite channel is typically -3, -4.5, -6, or -9 dB in intensity.

  2. Solderdude says:

    There may well be digital crossfeed filters that adhere to different signal separation/mixing or even expand stereo width images and call it crossfeed.
    This all depends on the philosophy/preference of the designer.
    What’s discussed above is analog crossfeed, not digital.

    2 types of analog crossfeed are discussed which both have different results due to different types of signal mixing.

    Not many people prefer crossfeed by the way as the sound is more ‘in the head’. Some people hope to find a more ‘speaker-alike’ presentation and hope it throws the sound from in front of you.

    Those looking for that effect will have to go the digital route and have extensive digital processing done to the original signal… a big no-no for bitperfect fanatics or purists.

    There is nothing ‘incorrect’ about it, just different points of view in this matter.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s