I, like others before me, have noticed that it's annoying to listen in headphones to recordings with loud sounds panned hard to one side, left or right. For me, it's worst with midrange sounds like lead vocals, electric guitars, solo horns, drum sets play from one side only. It's like someone shouting at you into one ear.
I rigged up a Stereo-Mono switch, which was easy enough and didn't cause any degradation I can detect. I find myself using it on early stereo records from the 1960s, especially Blue Note and Prestige jazz records. (It can hurt when Elvin Jones takes a wild solo in the right channel!)
I'd like to experiment with a passive Blend control, preferably infinitely variable so I can dial in how much the stereo image is reduced towards mono.
Some old receivers and preamps had a Blend control that allowed you to switch from full-width stereo to constricted width stereo, to mono. Some had Mono-L or Mono-R settings, so you could hear just one channel in both speakers, which I don't think is necessary for everyday listening.
I'm surprised I found so few examples on the innerwebs. I've attached a scrawled set of schematics showing three simple examples.
Ex A: Lifted from a harman-kardon TA-5000X tube receiver. It uses a dual-gang 100k potentiometer. I assume linear taper.
Looking at the pots with connections labelled top (1), wiper (2), bottom (3):
I think this would only work if the top (1) connection of the left channel pot goes to the Left channel input, while the (3) connection of the right channel pot goes to the Right channel input.
The bottom (3) connector of the left channel pot and the top (1) connector of the right channel pot are shorted and connected to signal ground.
The two (2) wiper connections are shorted together.
This arrangement would put the two tracks connected opposite from one another, so that when you turn the knob clockwise, the left channel track moves clockwise while the right channel track moves counterclockwise.
When you move the knob all the way clockwise, the two wipers will be grounded, and the two channels will play in full stereo.
When you turn the knob all the way counterclockwise, the left and right channels will be shorted together, with each 100k track loading each channel to ground, so the two channels will be blended together into mono.
Ex B: I found this on an electronics hobbyist forum. Uses a single track 10k ohm pot connected between the channels, preceded by 1k series resistors, one for each channel.
Turn the pot full clockwise and the two channels are shorted together (mono). Turn the pot full counterclockwise and the wiper is shorted to ground, leaving the full pot resistance between the two channels (stereo). The pot is not connected to ground.
Looks nice, but will this inevitably leave some 'bleed' between channels, causing permanent reduction of channel separation? (That might not be a problem with headphones. How much separation do you really need?)
Ex C: Lifted from a cheaper harman-kardon tube receiver (TA-3000). It uses a single 2M pot with its top (1) connector left disconnected, wiper (2) connected to the Left channel signal, and bottom (3) connection to the Right channel signal.
It looks like this one attempts to keep the two channels from blending (bleeding) too much by using a high value of resistance. Does such a high resistance introduce enough capacitance to cause attenuation of high audio frequencies?
Are all of these doomed to failure for one reason or another, and that's why you never see this kind of control?
The source I'm using is a Raspberry Pi with an Allo Boss DAC (output Z probably well below 1k ohms). The load is a tube headphone amp with a 50k volume control pot, or an O2 amp (10k vol control). I'm going to try to wire a box up today and give these controls a try. I'll post results if I find anything noteworthy.
Thoughts?
--
I rigged up a Stereo-Mono switch, which was easy enough and didn't cause any degradation I can detect. I find myself using it on early stereo records from the 1960s, especially Blue Note and Prestige jazz records. (It can hurt when Elvin Jones takes a wild solo in the right channel!)
I'd like to experiment with a passive Blend control, preferably infinitely variable so I can dial in how much the stereo image is reduced towards mono.
Some old receivers and preamps had a Blend control that allowed you to switch from full-width stereo to constricted width stereo, to mono. Some had Mono-L or Mono-R settings, so you could hear just one channel in both speakers, which I don't think is necessary for everyday listening.
I'm surprised I found so few examples on the innerwebs. I've attached a scrawled set of schematics showing three simple examples.
Ex A: Lifted from a harman-kardon TA-5000X tube receiver. It uses a dual-gang 100k potentiometer. I assume linear taper.
Looking at the pots with connections labelled top (1), wiper (2), bottom (3):
I think this would only work if the top (1) connection of the left channel pot goes to the Left channel input, while the (3) connection of the right channel pot goes to the Right channel input.
The bottom (3) connector of the left channel pot and the top (1) connector of the right channel pot are shorted and connected to signal ground.
The two (2) wiper connections are shorted together.
This arrangement would put the two tracks connected opposite from one another, so that when you turn the knob clockwise, the left channel track moves clockwise while the right channel track moves counterclockwise.
When you move the knob all the way clockwise, the two wipers will be grounded, and the two channels will play in full stereo.
When you turn the knob all the way counterclockwise, the left and right channels will be shorted together, with each 100k track loading each channel to ground, so the two channels will be blended together into mono.
Ex B: I found this on an electronics hobbyist forum. Uses a single track 10k ohm pot connected between the channels, preceded by 1k series resistors, one for each channel.
Turn the pot full clockwise and the two channels are shorted together (mono). Turn the pot full counterclockwise and the wiper is shorted to ground, leaving the full pot resistance between the two channels (stereo). The pot is not connected to ground.
Looks nice, but will this inevitably leave some 'bleed' between channels, causing permanent reduction of channel separation? (That might not be a problem with headphones. How much separation do you really need?)
Ex C: Lifted from a cheaper harman-kardon tube receiver (TA-3000). It uses a single 2M pot with its top (1) connector left disconnected, wiper (2) connected to the Left channel signal, and bottom (3) connection to the Right channel signal.
It looks like this one attempts to keep the two channels from blending (bleeding) too much by using a high value of resistance. Does such a high resistance introduce enough capacitance to cause attenuation of high audio frequencies?
Are all of these doomed to failure for one reason or another, and that's why you never see this kind of control?
The source I'm using is a Raspberry Pi with an Allo Boss DAC (output Z probably well below 1k ohms). The load is a tube headphone amp with a 50k volume control pot, or an O2 amp (10k vol control). I'm going to try to wire a box up today and give these controls a try. I'll post results if I find anything noteworthy.
Thoughts?
--
Attachments
Last edited:
Thanks.
Re: Ex B -- What is the "branch with the pot"? Put the switch between the left channel signal and the top (1) connection of the pot? I think that's what would work...
Also, would a higher value pot than 10k yield better results for Ex B? Or is there something about 10k that is important?
What is the problem with A? Bleed between channels?
If that's the problem, I have a dual-gang 1M linear pot I could throw at this. With the knob full counterclockwise, it would essentially be like having two 1M parallel resistors. I figure 1M won't load anything down at all.
Re: Ex B -- What is the "branch with the pot"? Put the switch between the left channel signal and the top (1) connection of the pot? I think that's what would work...
Also, would a higher value pot than 10k yield better results for Ex B? Or is there something about 10k that is important?
What is the problem with A? Bleed between channels?
If that's the problem, I have a dual-gang 1M linear pot I could throw at this. With the knob full counterclockwise, it would essentially be like having two 1M parallel resistors. I figure 1M won't load anything down at all.
You can add a SPST switch in series with the pot to open circuit the pot connection
between the channels. There would be complete channel separation when the switch is open.
The best values of the resistors, and the best value and taper of the pot, would depend on
the rest of the equipment. For solid state sources, try two 5k series resistors, and a 10k linear pot
With tube sources, try 20k resistors and 100k pot.
Both A and C can directly short together the channels, which could be damaging to the circuitry.
Series resistors would have to be added to prevent that. But version B should work well.
between the channels. There would be complete channel separation when the switch is open.
The best values of the resistors, and the best value and taper of the pot, would depend on
the rest of the equipment. For solid state sources, try two 5k series resistors, and a 10k linear pot
With tube sources, try 20k resistors and 100k pot.
Both A and C can directly short together the channels, which could be damaging to the circuitry.
Series resistors would have to be added to prevent that. But version B should work well.
Last edited:
Thanks again.
All my sources are solid state, and any tube sources will have a buffer on their outputs (usually a source follower, sometimes a cathode follower). Nothing will have an output impedance above 1k ohms.
Good point about not directly shorting the two source outputs together.
If I were to add the series resistors to Ex A and Ex C so that neither can directly short the two channels together, would A and C now work equally as well as Ex B? (see attached updated examples)
Or, will there be appreciable bleed between channels with all three examples, so all would require the defeat switch?
I'd like to be able to do this without an additional switch, if possible. But really, it's just an experiment. I'm trying to understand the upsides and downsides of each of these example circuits. It seems to me one would want to use as high a value of resistance as possible for the pot, in order to reduce bleed and not load down the source, but you also don't want to introduce extra shunt capacitance that reduces bandwidth too much. Another balancing act...
All my sources are solid state, and any tube sources will have a buffer on their outputs (usually a source follower, sometimes a cathode follower). Nothing will have an output impedance above 1k ohms.
Good point about not directly shorting the two source outputs together.
If I were to add the series resistors to Ex A and Ex C so that neither can directly short the two channels together, would A and C now work equally as well as Ex B? (see attached updated examples)
Or, will there be appreciable bleed between channels with all three examples, so all would require the defeat switch?
I'd like to be able to do this without an additional switch, if possible. But really, it's just an experiment. I'm trying to understand the upsides and downsides of each of these example circuits. It seems to me one would want to use as high a value of resistance as possible for the pot, in order to reduce bleed and not load down the source, but you also don't want to introduce extra shunt capacitance that reduces bandwidth too much. Another balancing act...
Attachments
Last edited:
Most all solid state sources and preamps are designed to drive 10k, including their output coupling capacitors.
So 10k series resistors are ok. If no switch, the pot should be large enough to get to around 30dB (about x30)
channel separation. Then the pot should be roughly 10k x 30, so use a 250k pot. But then you'd probably be
best off with a log taper pot, so it would be easier to set the pot for smaller values of resistance (more blending).
You want the pot to be no more than x30 the series resistors, since it would be hard to adjust for smaller separations,
even with a log pot. So circuit C is bad. Not sure about circuit A.
So 10k series resistors are ok. If no switch, the pot should be large enough to get to around 30dB (about x30)
channel separation. Then the pot should be roughly 10k x 30, so use a 250k pot. But then you'd probably be
best off with a log taper pot, so it would be easier to set the pot for smaller values of resistance (more blending).
You want the pot to be no more than x30 the series resistors, since it would be hard to adjust for smaller separations,
even with a log pot. So circuit C is bad. Not sure about circuit A.
Last edited:
Ah, the balancing act.
If:
- the sources can all drive 10k ohms, and
- the two channels are shorted together through the two 10k ohm series resistors, then
- the two 10k resistors are now in parallel, correct? That would make 5k ohms in series, with the 50k volume control in parallel (shunt to ground).
5k ohms in series between source and load would introduce a 1:10 voltage divider, would it not?
Assuming the output impedance (Zout) of the source component is 1k ohms, adding 5k in series to that makes the Zout look like 6k ohms, correct?
Wouldn't we want to minimize that increase in Zout from the source?
When I wired up my stereo/mono switch, I first used 4.7k resistors in series, but noticed a pronounced insertion loss when I switched to mono. I settled on 1k series resistors, which gave me just enough insertion loss to match levels between stereo and mono settings.
Finally, if the problem with Ex. C is that it shorts the two sources together, would adding an 'end stopper' resistor between the left channel and the pot's wiper fix that problem? (see attached diagram)
Would 2k ohms between the two channels provide enough protection for a source with Zout = 1k ohms? Perhaps a 4.7k ohm resistor could be used there? Or would that not allow the two channels to fully sum to mono?
--
If:
- the sources can all drive 10k ohms, and
- the two channels are shorted together through the two 10k ohm series resistors, then
- the two 10k resistors are now in parallel, correct? That would make 5k ohms in series, with the 50k volume control in parallel (shunt to ground).
5k ohms in series between source and load would introduce a 1:10 voltage divider, would it not?
Code:
(MONO)
L ch-----10k----+-----------------------+-----(L in)
| |
| (shorted) |
| 50k (vol ctrl)
R ch-----10k----+----+---(R in) |
| |
| gnd
50k (vol ctrl)
|
|
gndAssuming the output impedance (Zout) of the source component is 1k ohms, adding 5k in series to that makes the Zout look like 6k ohms, correct?
Wouldn't we want to minimize that increase in Zout from the source?
When I wired up my stereo/mono switch, I first used 4.7k resistors in series, but noticed a pronounced insertion loss when I switched to mono. I settled on 1k series resistors, which gave me just enough insertion loss to match levels between stereo and mono settings.
Finally, if the problem with Ex. C is that it shorts the two sources together, would adding an 'end stopper' resistor between the left channel and the pot's wiper fix that problem? (see attached diagram)
Would 2k ohms between the two channels provide enough protection for a source with Zout = 1k ohms? Perhaps a 4.7k ohm resistor could be used there? Or would that not allow the two channels to fully sum to mono?
--
Attachments
As Rayma says, all these cross-connections need resistance in front, or the channels will fight. Yeah, 10k for chips, 50-100k for tubes.
The old Headwize forum had several builds and threads for headphone cross-connect. Too many for my fingers to link--- look halfway down this page. Also 'crossfeed'.
DIY Projects – HeadWize Memorial
This one is simple/quick and chip premps can drive the ~~1k impedances.
A DIY Headphone Amplifier With Natural Crossfeed. – HeadWize Memorial
More elaborate crossfeed:
The Lindesberg Portable Headphone Amplifier With Crossfeed. – HeadWize Memorial
A fine plan from Chu Moy himself:
An Acoustic Simulator For Headphone Amplifiers. – HeadWize Memorial
The old Headwize forum had several builds and threads for headphone cross-connect. Too many for my fingers to link--- look halfway down this page. Also 'crossfeed'.
DIY Projects – HeadWize Memorial
This one is simple/quick and chip premps can drive the ~~1k impedances.
A DIY Headphone Amplifier With Natural Crossfeed. – HeadWize Memorial
More elaborate crossfeed:
The Lindesberg Portable Headphone Amplifier With Crossfeed. – HeadWize Memorial
A fine plan from Chu Moy himself:
An Acoustic Simulator For Headphone Amplifiers. – HeadWize Memorial
Oh.... a resistor-only crossfeed looks good but can sound funny. In open air, off-center sound waves bigger than your head reach both ears; smaller waves reach one ear more than the other. Depends on frequency. There's much bull-plop about HTF. In most Hi-Fi cases, any approximate curve takes most of the curse off of a hard-panned recording, and fine-trimming to suit the acoustics (if any) and your particular head is not really worthwhile.
Ouch.
Well, thanks for the fleshed out info. That's deeper than I wanted to go with this. I'm not worried about realizing a more naturally externalized presentation from headphones; I'm just trying to make hard-panned recordings from the earliest days of stereo less annoying to listen to in headphones. Modern stereo recordings all sound good enough to me as they are.
Perhaps what I need is just a touch of crossfeed (blend) to bring the sides in towards the center just enough to relieve the discomfort, but not enough to noticeably mess up phase or create noticeable comb filtering issues.
If I find that one setting of the pot works just enough for those old stereo recordings then perhaps I can measure the resistances and use what I find to make a 3-way Stereo-Blend-Mono switch.
For now, I have a decent 100k log pot, so I'll use 10k series resistors and give it a try. I expect I'll have to mess with values a bit to get the compromise that works for me. Then probably wire in a defeat switch, as rayma suggested.
Well, thanks for the fleshed out info. That's deeper than I wanted to go with this. I'm not worried about realizing a more naturally externalized presentation from headphones; I'm just trying to make hard-panned recordings from the earliest days of stereo less annoying to listen to in headphones. Modern stereo recordings all sound good enough to me as they are.
Perhaps what I need is just a touch of crossfeed (blend) to bring the sides in towards the center just enough to relieve the discomfort, but not enough to noticeably mess up phase or create noticeable comb filtering issues.
If I find that one setting of the pot works just enough for those old stereo recordings then perhaps I can measure the resistances and use what I find to make a 3-way Stereo-Blend-Mono switch.
For now, I have a decent 100k log pot, so I'll use 10k series resistors and give it a try. I expect I'll have to mess with values a bit to get the compromise that works for me. Then probably wire in a defeat switch, as rayma suggested.
The 100k pot will allow about 20dB of separation at max, which should be ok with a switch.
The only time I've liked using headphones was with binaural recordings, and no electrical crossfeed.
The only time I've liked using headphones was with binaural recordings, and no electrical crossfeed.
OK, that went together easily enough.
I did the recommended Ex. B, with 10k series resistors and a 250k log pot I had left over from repairing Fender guitar amps.
I do notice some attenuation compared to without the control in place. I'll probably end up installing a defeat switch at some point. Maybe this evening.
The pot is wired so that when it's turned all the way 'down' (counterclockwise) the playback is in mono. When the pot is turned up to 10 it's in full stereo.
I have a Fender amp knob on there, which has numbers from 1 to 10.
Starting from 10 (all the way up) and slowly turning down the pot, I don't notice a reduction of stereo width until way down around 4. The progression from stereo to mono is smooth from 4 down to 1. It's fiddly, but I can dial in a nice spot that retains enough stereo separation to sound natural but doesn't have that hole in the middle (or worse yet, that sound where all the drums are way over on the right except for the bass drum bleeding into the piano mic panned hard left, so the bass drum pounds in your left ear while the entire drum set is coming from the right).
Now the full mono setting sounds unnatural to me. So I think this experiment is a success.
Thanks rayma and PRR for helping me get this to work well on the first try. Much appreciated!
--
I have to listen to headphones at night because I live in an apartment complex and my downstairs neighbor is home all the time because of an injury. I want to be nice to them. Maybe I'll need to try Chu Moy's 'natural crossfeed' network. It doesn't look too intimidating.
I did the recommended Ex. B, with 10k series resistors and a 250k log pot I had left over from repairing Fender guitar amps.
I do notice some attenuation compared to without the control in place. I'll probably end up installing a defeat switch at some point. Maybe this evening.
The pot is wired so that when it's turned all the way 'down' (counterclockwise) the playback is in mono. When the pot is turned up to 10 it's in full stereo.
I have a Fender amp knob on there, which has numbers from 1 to 10.
Starting from 10 (all the way up) and slowly turning down the pot, I don't notice a reduction of stereo width until way down around 4. The progression from stereo to mono is smooth from 4 down to 1. It's fiddly, but I can dial in a nice spot that retains enough stereo separation to sound natural but doesn't have that hole in the middle (or worse yet, that sound where all the drums are way over on the right except for the bass drum bleeding into the piano mic panned hard left, so the bass drum pounds in your left ear while the entire drum set is coming from the right).
Now the full mono setting sounds unnatural to me. So I think this experiment is a success.
Thanks rayma and PRR for helping me get this to work well on the first try. Much appreciated!
--
I have to listen to headphones at night because I live in an apartment complex and my downstairs neighbor is home all the time because of an injury. I want to be nice to them. Maybe I'll need to try Chu Moy's 'natural crossfeed' network. It doesn't look too intimidating.
Last edited:
It may be tough to smoothly segue from infinite separation to <20dB sep. Yes, a bypass switch may be a good way to go. Most of your recordings do not need cross-fade. When you do have a ping-pong disk you probably know it before or quickly, and can cut-in the X-fade.
This is a diagram of what I wired up (attached).
For me, those old 60s stereo recordings are much easier to listen to with the 'blend' control set at somewhere between '3' and '2' on the volume knob, which measures 7k ohms.
With the 10k series resistors, a 25k or 10k pot with a defeat switch would be the better setup, for sure.
I'm thinking headphones don't really need 100% full stereo separation for my situation. With the control on '10', I believe I hear a slight pulling in at the sides of the stereo mix, but I hear that as a slight benefit in headphones. Besides, 25dB separation is considered very good for a phono cartridge, and nobody complained about that back in the day. I doubt being limited to 25dB separation is a real issue for casual listening on a stereo loudspeaker setup. A system for critical listening would be a different story, of course.
--
Attachments
After listening to a few recordings, I'm finding that the mid- and high frequencies are more blended than the bass frequencies, which is backwards from what we expect to hear in real life. I can understand the need for additional blending of the bass frequencies to match the mids and highs. It doesn't bother me, though.
I wonder, can Meier's basic 'Natural Crossfeed' be scaled up for higher impedances? This is the published version, using 1k series resistors each shunted with 470nF capacitors, with a 2.2k parallel resistor:
Since I now know I like the circuit with 10k series resistors and an approx. 7.5k parallel resistor, do you think the Meier circuit would perform similarly using those 10X higher resistance values with the shunt caps scaled down 0.1X to 47nF? (see attached diagram)
I wonder, can Meier's basic 'Natural Crossfeed' be scaled up for higher impedances? This is the published version, using 1k series resistors each shunted with 470nF capacitors, with a 2.2k parallel resistor:
Since I now know I like the circuit with 10k series resistors and an approx. 7.5k parallel resistor, do you think the Meier circuit would perform similarly using those 10X higher resistance values with the shunt caps scaled down 0.1X to 47nF? (see attached diagram)
Attachments
To have the same performance, scale all of his R up and all of the C down, by the same factor.
The shunt R would have to include any loading. Try a variable resistor for the shunt R,
a 50k-100k log pot should work. If the effective shunt is different from the scaled-up 22k,
the frequency response will be changed from the original design.
The shunt R would have to include any loading. Try a variable resistor for the shunt R,
a 50k-100k log pot should work. If the effective shunt is different from the scaled-up 22k,
the frequency response will be changed from the original design.
Last edited:
OK, thanks.
I can try the experiment with what I have if I set the 250k pot to 22k (just under '3' on the knob), and tack in 47nF caps in parallel with the 10k series resistors.
If I find that I need more blending towards mono and the parallel R (Rshunt) really does have to be 7.5k ('2' on the knob), and since 22k/7.5k = 2.93, then I'd have to scale up the parallel caps (Cseries) to:
0.05uF * 2.9333 = 0.147uF
I figure Cseries = 0.15uF will be close enough for Rseries = 10k and Rshunt = 7.5k.
Correct?
I can try the experiment with what I have if I set the 250k pot to 22k (just under '3' on the knob), and tack in 47nF caps in parallel with the 10k series resistors.
If I find that I need more blending towards mono and the parallel R (Rshunt) really does have to be 7.5k ('2' on the knob), and since 22k/7.5k = 2.93, then I'd have to scale up the parallel caps (Cseries) to:
0.05uF * 2.9333 = 0.147uF
I figure Cseries = 0.15uF will be close enough for Rseries = 10k and Rshunt = 7.5k.
Correct?
Last edited:
Any loading at the outputs will need a higher value for the shunt R.
For example, if each output is loaded by 22k, then adjust the shunt R
value upward so the 44k in parallel with the Rshunt gives the right value.
Not sure if the circuit will work as intended if the amount of designed-in
blending is changed. All the parts values will interact.
For example, if each output is loaded by 22k, then adjust the shunt R
value upward so the 44k in parallel with the Rshunt gives the right value.
Not sure if the circuit will work as intended if the amount of designed-in
blending is changed. All the parts values will interact.
Last edited:
Yes, and interact they do.
By 'loading' do you mean for instance the volume control at the input of the headphone amp? That is 50k ohms, so then that would mean 100k in parallel with the Rshunt in the crossfeed circuit -- correct?
If so, then I need to take into account that there's always 100k in parallel with whatever value Rshunt is. Therefore, if I want to hear the effect Rshunt = 22k has on this setup, I need to make 100k||x = 22k. That would be Rshunt = 29k on the pot (and *not* 22k).
Correct?
I'll try it and see what that sounds like, to get a reference.
By 'loading' do you mean for instance the volume control at the input of the headphone amp? That is 50k ohms, so then that would mean 100k in parallel with the Rshunt in the crossfeed circuit -- correct?
If so, then I need to take into account that there's always 100k in parallel with whatever value Rshunt is. Therefore, if I want to hear the effect Rshunt = 22k has on this setup, I need to make 100k||x = 22k. That would be Rshunt = 29k on the pot (and *not* 22k).
Correct?
I'll try it and see what that sounds like, to get a reference.
Last edited: