Right.
Also we were assuming that the source impedance was low, say <500R.
If not, that will also have an unwanted effect.
Also we were assuming that the source impedance was low, say <500R.
If not, that will also have an unwanted effect.
Last edited:
It seems adding the series capacitors (shunting the 10k series resistors) reduces the effect of the 'cross-feeding', by quite a lot.
Perhaps I should try Meier's circuit as he described it, with the low values of R and C. They wouldn't be affected as much by the loading from the 50k volume control on the headphone amp.
With my current Rseries = 10k setup, using .15uF for the Cseries, I had to reduce Rshunt to 5.8k ohms to get the blending I wanted. That's in parallel with 100k, so that's an actual 100k*5.8k/105.8k = 5.2k ohms for Rshunt.
22k/5.8k = 3.793, so scaling Cseries up from 47nF for 22k, 47nF*3.79 = 180nF, or maybe that's close enough with 0.15uF? I can also try 0.22uF to see if that sounds better subjectively.
I guess the way to do this is to rig up a switch, with the Natural Crossfeed circuit set how I like it, to be switched in to tame these early '60s pingpong-stereo recordings, and forget about adjustability.
I think the circuit would be like this (see attached diagram)....
Perhaps I should try Meier's circuit as he described it, with the low values of R and C. They wouldn't be affected as much by the loading from the 50k volume control on the headphone amp.
With my current Rseries = 10k setup, using .15uF for the Cseries, I had to reduce Rshunt to 5.8k ohms to get the blending I wanted. That's in parallel with 100k, so that's an actual 100k*5.8k/105.8k = 5.2k ohms for Rshunt.
22k/5.8k = 3.793, so scaling Cseries up from 47nF for 22k, 47nF*3.79 = 180nF, or maybe that's close enough with 0.15uF? I can also try 0.22uF to see if that sounds better subjectively.
I guess the way to do this is to rig up a switch, with the Natural Crossfeed circuit set how I like it, to be switched in to tame these early '60s pingpong-stereo recordings, and forget about adjustability.
I think the circuit would be like this (see attached diagram)....
Attachments
First off, if I use the scaled-up values derived from the original Meier circuit, the stereo image is still very wide. Too wide for these 'pingpong stereo' recordings, at least for my liking. That's why I'm trying to find a different combination of Rseries and Rshunt, and Cseries to match.
With the circuit set up as in the diagram attached to the previous message (Rseries = 10k, Rshunt = 5.6k), I was able to switch Cseries between 47nF, 100nF, 147nF.
I found that increasing the value of Cseries widens the stereo image in the high frequencies, which makes the presentation sound 'brighter' and 'wider'.
This begs the question, is there a 'correct' value of Cseries to be found? Or is this a matter of taste? Is there anything 'wrong' with reducing the value of Cseries to 100nF if I like it better that way? Or maybe I can wire in 100nF and make adding 82pF in parallel switchable. Clip-leading caps in and out does not result in objectionable noises, so a decent switch shouldn't either.
I am not after the ultimate Natural Crossfeed to recreate the feeling of listening to loudspeakers in a room. I'm merely trying to tame the too-wide stereo effect of a certain class of stereo recordings (almost all of them mixed to stereo by Rudy Van Gelder, but there were other engineers making stereo mixes in this style at the time), without resorting to the sledgehammer solution of summing to mono (not that mono is bad or anything).
--
With the circuit set up as in the diagram attached to the previous message (Rseries = 10k, Rshunt = 5.6k), I was able to switch Cseries between 47nF, 100nF, 147nF.
I found that increasing the value of Cseries widens the stereo image in the high frequencies, which makes the presentation sound 'brighter' and 'wider'.
This begs the question, is there a 'correct' value of Cseries to be found? Or is this a matter of taste? Is there anything 'wrong' with reducing the value of Cseries to 100nF if I like it better that way? Or maybe I can wire in 100nF and make adding 82pF in parallel switchable. Clip-leading caps in and out does not result in objectionable noises, so a decent switch shouldn't either.
I am not after the ultimate Natural Crossfeed to recreate the feeling of listening to loudspeakers in a room. I'm merely trying to tame the too-wide stereo effect of a certain class of stereo recordings (almost all of them mixed to stereo by Rudy Van Gelder, but there were other engineers making stereo mixes in this style at the time), without resorting to the sledgehammer solution of summing to mono (not that mono is bad or anything).
--
All of this is purely to personal taste, though there is some theory involved.
You could first try the circuit with no capacitors, 10k series Rs, and adjust the shunt R to taste.
Then experiment with the C values, keeping both equal.
You could first try the circuit with no capacitors, 10k series Rs, and adjust the shunt R to taste.
Then experiment with the C values, keeping both equal.
Thanks rayma, that's actually what I've been doing. Using clip leads. Totally unscientific. What I've found so far:
1) The reduced width effect is most pronounced without the series caps. The problem is that it also makes the sound rather dull (perceived lack of high frequencies) and there's some weirdness with bass frequencies having more stereo separation than treble frequencies, which is unnatural. But this is quite adjustable with a pot for Rshunt. It works well enough.
2) Adding the value of cap as suggested by Meier results in the stereo width being retained a lot more.
With Rseries = 10k and Rshunt = 7.5k and no capacitors, I found the stereo width to my liking for these pingpong recordings. Adding the scaled-up 0.15uF caps left the bass frequencies at the same narrower width, but the treble frequencies spread out almost all the way back to full-width stereo.
With the 0.15uF caps in place, reducing Rshunt to 5.6k brought the stereo width back to where I wanted it. But now the presentation is rather bright sounding. Reducing the cap value to 0.1uF reduces the width, but deadens the tonal quality. However, I can see this setup with 0.1uF in place might be just what's needed for an overly bright *and* overly wide pingpong stereo recording. Perhaps it would be best to be able to switch between Cseries = 0.1uF or 0.147uF, by switching a 0.047uF cap in/out of parallel with the 0.1uF cap.
3) Now the problem is how to defeat this mess!
I think a 3PST or 3PDT switch could simultaneously short out the Rseries||Cseries pairs while opening the connection to Rshunt.
A DPST or DPDT switch could switch in/out the parallel 0.047uF cap around Cseries (0.1uF).
So now there are two switches, Crossfeed In/Out and Crossfeed Narrow/Wide.
I may need to revisit parts values as I listen more, but I think this is a useful general plan for a simple little switchbox to experiment with.
What do you think?
____________________________
PS - I wonder if all this would be more manageable with a buffer added after it.
1) The reduced width effect is most pronounced without the series caps. The problem is that it also makes the sound rather dull (perceived lack of high frequencies) and there's some weirdness with bass frequencies having more stereo separation than treble frequencies, which is unnatural. But this is quite adjustable with a pot for Rshunt. It works well enough.
2) Adding the value of cap as suggested by Meier results in the stereo width being retained a lot more.
With Rseries = 10k and Rshunt = 7.5k and no capacitors, I found the stereo width to my liking for these pingpong recordings. Adding the scaled-up 0.15uF caps left the bass frequencies at the same narrower width, but the treble frequencies spread out almost all the way back to full-width stereo.
With the 0.15uF caps in place, reducing Rshunt to 5.6k brought the stereo width back to where I wanted it. But now the presentation is rather bright sounding. Reducing the cap value to 0.1uF reduces the width, but deadens the tonal quality. However, I can see this setup with 0.1uF in place might be just what's needed for an overly bright *and* overly wide pingpong stereo recording. Perhaps it would be best to be able to switch between Cseries = 0.1uF or 0.147uF, by switching a 0.047uF cap in/out of parallel with the 0.1uF cap.
3) Now the problem is how to defeat this mess!
I think a 3PST or 3PDT switch could simultaneously short out the Rseries||Cseries pairs while opening the connection to Rshunt.
A DPST or DPDT switch could switch in/out the parallel 0.047uF cap around Cseries (0.1uF).
So now there are two switches, Crossfeed In/Out and Crossfeed Narrow/Wide.
I may need to revisit parts values as I listen more, but I think this is a useful general plan for a simple little switchbox to experiment with.
What do you think?
____________________________
PS - I wonder if all this would be more manageable with a buffer added after it.
Attachments
Last edited:
Looks good, it may take a while to settle on component values that you like the most.
The bypass switch will be good for comparisons, to avoid altering the tonal balance too much.
If the following circuit is 50k input Z, that's high enough that I wouldn't worry about
using an output buffer. Just keep the series Rs as 10k.
And experimenting is not unscientific. That's how Faraday discovered the electromagnetic field,
and how Fleming discovered penicillin. Among many other discoveries throughout history.
The bypass switch will be good for comparisons, to avoid altering the tonal balance too much.
If the following circuit is 50k input Z, that's high enough that I wouldn't worry about
using an output buffer. Just keep the series Rs as 10k.
And experimenting is not unscientific. That's how Faraday discovered the electromagnetic field,
and how Fleming discovered penicillin. Among many other discoveries throughout history.
Last edited:
Thanks again for your help.
I worked out a wiring layout to solder all the parts onto a couple of switches (one DPDT and one 3PDT) with no flying parts and no terminal strip or perfboard. Should be good.
I worked out a wiring layout to solder all the parts onto a couple of switches (one DPDT and one 3PDT) with no flying parts and no terminal strip or perfboard. Should be good.
This headphone 'cross-feed' or 'blend' thing is taking me down quite the rabbit hole. Thank you PRR for pointing me to the Headwize project archive. I've been reading about Jan Meier's and Chu Moy's cross-feed circuits, and the modifications people came up with. I've also downloaded Meier's and Chu's spreadsheets to get an idea of what's going on when you plug in different parts values. Interesting.
For now, I've settled on a high-Z version of Meier's circuit, but retained a potentiometer for the shunt resistance. I've attached a diagram of the circuit I'm listening to now.
The series Rs and Cs are straight out of Meier; 10k ohms and 47nF (scaled up from Meier's original 1k ohms and 470nF values).
I added a bypass switch, but I couldn't find the 3PDT switch I thought I had. I used a DPDT switch instead, with a 1k 'end stop' resistor in series with the shunt potentiometer to protect the source outputs from being fully shorted together. With the switch in the bypass position, the series R and C are shorted out, so the signal passes straight from input to output. The shunt potentiometer is always in circuit (unfortunately), which is why I figured the end stop resistor is necessary (in case I flip the switch to bypass with the pot set to 0 ohms).
The shunt resistance is supplied by a 25k log pot (with the 1k end stop resistor in series). There's a knob on the pot, with indicator numbers from 0 to 10.
A good recording to hear the effects of this circuit is Cannonball Adderley "Something Else", with Miles Davis as a guest artist. It's a classic jazz album and a perfect example of a Rudy Van Gelder 'ping-pong hyper-stereo' mixdown. The horns are panned hard left, the piano is dead center, and the bass and drums are both panned hard right. Reverb is panned to the center to soften the hard-panning effects, but that doesn't work for listening in headphones.
With the effect enabled:
With the pot set to 10, the pot's 25k ohms with the 1k end stop resistor in series combine to 26k ohms shunting the L and R channels. If I bypass the circuit in this position, I can hear a noticeable difference, but it's generally subtle. The horns still sound panned 270 degrees hard left and the bass and drums still sound panned 180 degrees hard right, but not as far as in the bypass position. There is definitely some softening.
The narrowing of stereo width proceeds very gradually as you turn down the control, only beginning to blend the higher frequency sounds toward the center at knob setting 3.
When you get down to 2, there's a noticeable shift of the hard-panned sounds towards the center. They start sounding like they're coming from about 335 degrees and 225 degrees (or you can think of it as 10 o'clock and 2 o'clock).
At 0 the sound gets very close to mono, but not completely, because of the 1k end stop resistor. That's OK, though. It's close enough that it's not bothersome, and besides, I find straight-up mono in headphones sounds unnatural, like the sound is right in the middle of your forehead. The slightest hint of stereo separation actually works for me.
Another thing I noticed is that with the cross-feed engaged, bass 'weight' definitely suffers. The bass becomes a bit light and more diffuse. It works for jazz recordings, but I'd miss hearing tight, punchy bass on rock or electric jazz recordings. It also might be a noticeable problem on recordings of large orchestras.
With the effect bypassed and the pot set to 10:
The bass and drumset playing in my right ear is painful at moderately loud volume. However, the bass response is definitely tighter, and it sounds like it goes lower in frequency. This is a definite trade-off. I can see why Chu Moy added a bass boost option to his circuit.
The piano (panned dead-center) sounds more full and 'tight' with the effect bypassed. Enabling the effect makes the piano sound farther away and a bit more diffuse. It's not an unpleasant effect.
The horns in the left channel sound weightier, but again, it's painful to have them playing close-up and from just behind my left ear. This is where the effect really helps.
With the effect engaged and set to 6 to 10 on the knob:
There's not much difference between 6 and 10, probably because of the log taper. Perhaps this is an application where a linear taper pot would work better. I don't have a 25k linear pot to try, so I can't say for sure.
Comparing to the bypassed sound, the effect is subtle at these higher settings. The mix sounds a bit farther away, which can be good. Old stereo 'ping pong' recordings are definitely improved, although the stereo spread remains very wide. The effect does move the hard-panned sounds a bit farther from your ears, and maybe a little to the center. More modern stereo recordings sound hardly changed at all.
With the effect engaged and set to about 2 on the knob:
I listened to B.B. King "Live at the Regal", which has the horn section and B.B.'s electric guitar panned hard right, the piano panned hard left, and the bass, drums and B.B.'s voice panned center. There's lots of room sound to soften the blow (it's a great sounding mix on loudspeakers), but the cross-feed definitely keeps the lead guitar from making your right ear go numb when B.B. takes a solo. The cross-feed does reduce the ambient spaciousness of the recording, but one can dial in the desired compromise with the pot. The electric bass and B.B.'s voice panned dead-center don't change in tonality at all.
I checked a couple more recordings with vocals panned in the center, and they all come through fine. The only problems happen when the bass is panned hard left or hard right, as in The Beatles "Taxman" on "Revolver". McCartney's bass guitar is panned hard left but moves to dead center with the cross-feed engaged. However, the bass loses some 'punch' in the process. On the other hand, when the guitar solo started, panned hard right, I was grateful for the effect. That midrange-heavy and dry-sounding electric guitar really hurts playing straight into my right ear.
All in all, I think having a cross-feed circuit to switch in is useful. I'll need to get a 3PDT switch so I can do a properly full bypass.
Thanks again to rayma and PRR.
--
For now, I've settled on a high-Z version of Meier's circuit, but retained a potentiometer for the shunt resistance. I've attached a diagram of the circuit I'm listening to now.
The series Rs and Cs are straight out of Meier; 10k ohms and 47nF (scaled up from Meier's original 1k ohms and 470nF values).
I added a bypass switch, but I couldn't find the 3PDT switch I thought I had. I used a DPDT switch instead, with a 1k 'end stop' resistor in series with the shunt potentiometer to protect the source outputs from being fully shorted together. With the switch in the bypass position, the series R and C are shorted out, so the signal passes straight from input to output. The shunt potentiometer is always in circuit (unfortunately), which is why I figured the end stop resistor is necessary (in case I flip the switch to bypass with the pot set to 0 ohms).
The shunt resistance is supplied by a 25k log pot (with the 1k end stop resistor in series). There's a knob on the pot, with indicator numbers from 0 to 10.
A good recording to hear the effects of this circuit is Cannonball Adderley "Something Else", with Miles Davis as a guest artist. It's a classic jazz album and a perfect example of a Rudy Van Gelder 'ping-pong hyper-stereo' mixdown. The horns are panned hard left, the piano is dead center, and the bass and drums are both panned hard right. Reverb is panned to the center to soften the hard-panning effects, but that doesn't work for listening in headphones.
With the effect enabled:
With the pot set to 10, the pot's 25k ohms with the 1k end stop resistor in series combine to 26k ohms shunting the L and R channels. If I bypass the circuit in this position, I can hear a noticeable difference, but it's generally subtle. The horns still sound panned 270 degrees hard left and the bass and drums still sound panned 180 degrees hard right, but not as far as in the bypass position. There is definitely some softening.
The narrowing of stereo width proceeds very gradually as you turn down the control, only beginning to blend the higher frequency sounds toward the center at knob setting 3.
When you get down to 2, there's a noticeable shift of the hard-panned sounds towards the center. They start sounding like they're coming from about 335 degrees and 225 degrees (or you can think of it as 10 o'clock and 2 o'clock).
At 0 the sound gets very close to mono, but not completely, because of the 1k end stop resistor. That's OK, though. It's close enough that it's not bothersome, and besides, I find straight-up mono in headphones sounds unnatural, like the sound is right in the middle of your forehead. The slightest hint of stereo separation actually works for me.
Another thing I noticed is that with the cross-feed engaged, bass 'weight' definitely suffers. The bass becomes a bit light and more diffuse. It works for jazz recordings, but I'd miss hearing tight, punchy bass on rock or electric jazz recordings. It also might be a noticeable problem on recordings of large orchestras.
With the effect bypassed and the pot set to 10:
The bass and drumset playing in my right ear is painful at moderately loud volume. However, the bass response is definitely tighter, and it sounds like it goes lower in frequency. This is a definite trade-off. I can see why Chu Moy added a bass boost option to his circuit.
The piano (panned dead-center) sounds more full and 'tight' with the effect bypassed. Enabling the effect makes the piano sound farther away and a bit more diffuse. It's not an unpleasant effect.
The horns in the left channel sound weightier, but again, it's painful to have them playing close-up and from just behind my left ear. This is where the effect really helps.
With the effect engaged and set to 6 to 10 on the knob:
There's not much difference between 6 and 10, probably because of the log taper. Perhaps this is an application where a linear taper pot would work better. I don't have a 25k linear pot to try, so I can't say for sure.
Comparing to the bypassed sound, the effect is subtle at these higher settings. The mix sounds a bit farther away, which can be good. Old stereo 'ping pong' recordings are definitely improved, although the stereo spread remains very wide. The effect does move the hard-panned sounds a bit farther from your ears, and maybe a little to the center. More modern stereo recordings sound hardly changed at all.
With the effect engaged and set to about 2 on the knob:
I listened to B.B. King "Live at the Regal", which has the horn section and B.B.'s electric guitar panned hard right, the piano panned hard left, and the bass, drums and B.B.'s voice panned center. There's lots of room sound to soften the blow (it's a great sounding mix on loudspeakers), but the cross-feed definitely keeps the lead guitar from making your right ear go numb when B.B. takes a solo. The cross-feed does reduce the ambient spaciousness of the recording, but one can dial in the desired compromise with the pot. The electric bass and B.B.'s voice panned dead-center don't change in tonality at all.
I checked a couple more recordings with vocals panned in the center, and they all come through fine. The only problems happen when the bass is panned hard left or hard right, as in The Beatles "Taxman" on "Revolver". McCartney's bass guitar is panned hard left but moves to dead center with the cross-feed engaged. However, the bass loses some 'punch' in the process. On the other hand, when the guitar solo started, panned hard right, I was grateful for the effect. That midrange-heavy and dry-sounding electric guitar really hurts playing straight into my right ear.
All in all, I think having a cross-feed circuit to switch in is useful. I'll need to get a 3PDT switch so I can do a properly full bypass.
Thanks again to rayma and PRR.
--
