Stereo to Mono with resampling?

Hello,

This is my first post here. I don't know much on DSP but I think I
found something interesting. Suppose we have a 44.1Khz stereo wave
file. We can treat it like 88.2Khz mono since stereo pairs are
interleaved. If we downsample this to 44.1Khz (using a lowpass filter
first), the result is a mono version of the original file, but not the
same if we just added the two channels. It surprisingly sounds better,
especially in the high frequencies.  I also found that it works even if
one of the channels is inverted, it seems there is less cancelling due
to inverted phases. I searched but couldn't find any mention of this,
so is this new, what do you think?

Best regards.

tolga@tolgas.com wrote:
> 
> Hello,
> 
> This is my first post here. I don't know much on DSP but I think I
> found something interesting. Suppose we have a 44.1Khz stereo wave
> file. We can treat it like 88.2Khz mono since stereo pairs are
> interleaved. If we downsample this to 44.1Khz (using a lowpass filter
> first), the result is a mono version of the original file, but not the
> same if we just added the two channels. It surprisingly sounds better,
> especially in the high frequencies.  I also found that it works even if
> one of the channels is inverted, it seems there is less cancelling due
> to inverted phases. I searched but couldn't find any mention of this,
> so is this new, what do you think?

That certainly sounds very interesting. I'll have to try it out.

Cheers,
Erik
-- 
+-----------------------------------------------------------+
  Erik de Castro Lopo  nospam@mega-nerd.com (Yes it's valid)
+-----------------------------------------------------------+
"Every time microshaft's stock price drops again, I rejoice. I
want to see that bunch of criminals brought to their knees.
Preferably at the chopping block."
-- rixt in http://linuxtoday.com/stories/20659_flat.html

Thanks for your interest. However, I think I had made a mistake. The
result doesn't sound better, it sounds the same... Yet, it's curious
that downmixing can be achieved by downsampling. I wonder if downmixing
n channels would work too, I need to experiment more..

Best regards.

tolga@tolgas.com wrote:

> Thanks for your interest. However, I think I had made a mistake. The
> result doesn't sound better, it sounds the same... Yet, it's curious
> that downmixing can be achieved by downsampling. I wonder if downmixing
> n channels would work too, I need to experiment more..

As I understand it, early CD players used one DAC multiplexed
between the two channels.  The result, then, as a mono signal
(L+R) would be similar to what you have.

For lower frequencies it is pretty much the same as averaging
the left and right channels.  Even at the higher frequencies
of most audio sources the effect is pretty small.

-- glen

"glen herrmannsfeldt" <gah@ugcs.caltech.edu> wrote in message 
news:9eidnY4Jve-t_1PfRVn-tA@comcast.com...
> tolga@tolgas.com wrote:
>
>> Thanks for your interest. However, I think I had made a mistake. The
>> result doesn't sound better, it sounds the same... Yet, it's curious
>> that downmixing can be achieved by downsampling. I wonder if downmixing
>> n channels would work too, I need to experiment more..
>
> As I understand it, early CD players used one DAC multiplexed
> between the two channels.  The result, then, as a mono signal
> (L+R) would be similar to what you have.

I've heard that one DAC was used too, but my understanding is that it still 
produced discreet left and right channel outputs, but that they were one half 
sample out of phase with each other (e.g. left was delayed one half sample vs. 
right).  Maybe that's the same thing you are saying? 

tolga@tolgas.com wrote:
> Hello,
> 
> This is my first post here. I don't know much on DSP but I think I
> found something interesting. Suppose we have a 44.1Khz stereo wave
> file. We can treat it like 88.2Khz mono since stereo pairs are
> interleaved. If we downsample this to 44.1Khz (using a lowpass filter
> first), the result is a mono version of the original file, but not the
> same if we just added the two channels. It surprisingly sounds better,
> especially in the high frequencies.  I also found that it works even if
> one of the channels is inverted, it seems there is less cancelling due
> to inverted phases. I searched but couldn't find any mention of this,
> so is this new, what do you think?
> 
> Best regards.
> 

I have treated L/R interleaved stereo as mono before.  The way I 
justified it to myself was by thinking of it this way:

Start with two mono streams, left and right.  Insert zeroes between each 
left sample, so you get 88.2K samples/second.  This is exactly the same 
as the first step in an interpolation process.  Do the same to the right 
channel. Shift the right sequence by one sample, and add them together. 
  Voila - the two mono streams are L/R interleaved.

This means that the signal from 0 to 22.05 KHz is the sum of the left 
channel and a time-shifted version of the right channel.  The signal 
from 22.5 to 44.1 is the mirror image of that.

It works the same way if you interleave more channels.  You can analyze 
the resulting spectrum with multi-rate techniques (i.e., interpolation).

-- 
Jim Thomas            Principal Applications Engineer  Bittware, Inc
jthomas@bittware.com  http://www.bittware.com    (603) 226-0404 x536
Quidquid latine dictum sit, altum sonatur.
Whatever is said in Latin sounds profound.

Jon Harris wrote:

> "glen herrmannsfeldt" <gah@ugcs.caltech.edu> wrote in message 
> news:9eidnY4Jve-t_1PfRVn-tA@comcast.com...

(snip)

>>As I understand it, early CD players used one DAC multiplexed
>>between the two channels.  The result, then, as a mono signal
>>(L+R) would be similar to what you have.

> I've heard that one DAC was used too, but my understanding is that it still 
> produced discreet left and right channel outputs, but that they were one half 
> sample out of phase with each other (e.g. left was delayed one half sample vs. 
> right).  Maybe that's the same thing you are saying? 

Yes, but one could connect the two outputs together for a mono signal.

My amplifier has a mono/stereo/reverse switch.

-- glen

My guess is that you are using a half-band filter (or something very
close) to do the 88->44 conversion.

This will be equivalent to delaying one of the channels by a
half-sample, then adding it to the non-delayed other channel.

In other words, it is pretty much the same as just adding the two
channels together.

-- Mark Borgerding

"glen herrmannsfeldt" <gah@ugcs.caltech.edu> wrote in message 
news:P7Sdnb4F3Z3VUVPfRVn-qA@comcast.com...
> Jon Harris wrote:
>
>> "glen herrmannsfeldt" <gah@ugcs.caltech.edu> wrote in message 
>> news:9eidnY4Jve-t_1PfRVn-tA@comcast.com...
>
> (snip)
>
>>>As I understand it, early CD players used one DAC multiplexed
>>>between the two channels.  The result, then, as a mono signal
>>>(L+R) would be similar to what you have.
>
>> I've heard that one DAC was used too, but my understanding is that it still 
>> produced discreet left and right channel outputs, but that they were one half 
>> sample out of phase with each other (e.g. left was delayed one half sample 
>> vs. right).  Maybe that's the same thing you are saying?
>
> Yes, but one could connect the two outputs together for a mono signal.
>
> My amplifier has a mono/stereo/reverse switch.

Got it!