stereo piano sounds bad in mono

[nielsdolieslager]

One thing to check at least - I had some issues with recording guitar two ways - one lined and one through analog amp.

I got a boxy sound trying to blend these.

In this case there are no timing differences at all between the two sources, just on the resulting recording - so I could just delay one side to make them line up - an boxyness disappeared.

Reason for it all was different propagation delay through A/D-converters. I have a RME card with two analog built in - and then extended with a RME addon card. This was enough to cause different timing two ways. Just 12 samples or something.

Anyway - check that you use the same equipment for both channels - or you can try forever and not get it right.

It seems I can't line them up perfectly because the distance to each mic was different per key. Still trying to find the best setting for sample delay with the least phase cancelling. If you want to try, be my guest, you can download a track with that piano after following the above link

[egbert]

the Voxengo plug has a very funny correlometer, shows correlation per freq band. The correlation problems differ per key, lol, and the problems don't correspond with what my ears tell me.

That is exactly what you would expect from the geometry. Each note is either a single string in the deep bass or two or three thinner ones close together as you move up the range. The distance from those strings to each of the mics changes as you move up to the next note. Not only that, because the frequencies of the strings are also changing, then the characteristic comb filter pattern (where you get a series of amplitude notches right across the frequency spectrum due to cancellation) will be different for every string. For a given pair of mics locations, you might have a path difference which allows for constructive reinforcement on the fundamental of one note and cancellation on the fundamental or the next note.

[nielsdolieslager]

Yeah, makes sense. So before this piano works best in mono I need to line up the samples per key. Will work on my plan a bit today.

New idea: make 7 tracks with one octave of the piano each and set sample delay settings per octave

[nielsdolieslager]

Little update, didn't take the time to set sample delays per octave yet. Learned two new things though:
Kore 2 has a "mic conditioner" that allows you to delay a l/r channel with down to 1/100 of a ms so when I'm done fine tuning I can load 7 or 8 instances of Kontakt in Kore, set the delays and save it as one instrument instead of messing around with sample delay and lots of channels in Logic. Could even use it in standalone for performance without having to worry about DAW latency. Kore keeps surprising me

Second thing: different EQ settings per channel with Waves GEQ doesn't help at all. Guess it causes more phasing problems because it's not a linear phase EQ? For the current track I'm working on I'll try l/r channels in Logic with linear EQ.

[camsr]

You don't want the mono as it is now, you want the L/R pair to recombine into better mono. Linear phase might be the solution, but it's a very complex problem with the already present comb filtering.

[nielsdolieslager]

You don't want the mono as it is now, you want the L/R pair to recombine into better mono. Linear phase might be the solution, but it's a very complex problem with the already present comb filtering.

Yeah, working on a track right now where I only use two octaves. Delaying the right with 32 samples sounds best and the linear phase EQ sounds better then the G-EQ.

And a new update, I took the MAutoAlign for a 14 day testdrive and even when I let it analyse the same l/r samples it gives me different results everytime and most of them do not improve the situation. Guess a detuned piano is too difficult for software, will have to use my ears per key

[egbert]

To fix the comb filtering you could (if you could be bothered) split all the stereo pairs into mono files and align the transients (ie remove the relative delay). Then sum them to mono. Comb filtering should be gone. If there is a plugin that will automatically align the transients then it could save you a lot of fiddly work.

[4damind]

Maybe you should try different plug-ins.
Alicia Keys has some options for the stereo image. There are different positions (artist,audience) and the stereo image has a knob which goes from Mono to wide stereo. Mono sounds for me very good and I use it also sometimes in combination with a small amount of tape delay.

[nielsdolieslager]

To fix the comb filtering you could (if you could be bothered) split all the stereo pairs into mono files and align the transients (ie remove the relative delay). Then sum them to mono. Comb filtering should be gone. If there is a plugin that will automatically align the transients then it could save you a lot of fiddly work.

So you think it would work visually? That's good news. I don't know if MAutoAlign was designed to work with transients, it sure can't handle this piano. Will give Radix AutoAlign a try too. I only found TAP by Advanced AudioWaves for aligning transients, VST3/iLok only, so not for me.

[nielsdolieslager]

Maybe you should try different plug-ins.
Alicia Keys has some options for the stereo image. There are different positions (artist,audience) and the stereo image has a knob which goes from Mono to wide stereo. Mono sounds for me very good and I use it also sometimes in combination with a small amount of tape delay.

AK is great but most of the times I just want to work with my old broken piano, that's why I want to solve the problem instead of avoid it

[egbert]

Actually - there is a shortcut with the editing - you might get good results in an audio editor with the equivalent of "trim silence" in Logic. You set a threshold and samples are trimmed to those zones with audio louder than the threshold. You could batch convert all the stereos to mono - split to L and R versions. Then "strip silence" and all the new files would have the key strike start right at the beginning.

It would not be sample accurate though - I don't believe so anyway. You could try fine tuning and see how close you could get to the results you get aligning by eye. If the error is 20 samples or less (half a millisec at 44.1kHz) then the results might be pretty acceptable. For any that aren't sounding right you could make adjustments.

You would still need to eyeball the samples to make sure the sample wasn't started by some stray noise - ie something like a seat squeak or something in the mechanism preceeding the actual note start

[nielsdolieslager]

Wow!
I finally downloaded a demo of Waves InPhase, that's the solution!
It calculates average correlation over 2 seconds, not real time correlation. When I shift one of the channels a few milliseconds I can see the average result changing and it remembers the best and the worst results. Of course the phasing problems differ per key so the current version of this piano will never be without phasing problems, but I'm having less phasing problems now (I replaced the file on Soundcloud). And for the samples it's going to work.

The other plugins had their own problems. Sound Radix Auto-Align and Melda MAutoAlign gave different results every time, in Voxengo PHA-979 it wasn't clear what setting was best. Aligning the attacks visually wouldn't work either because the waveforms are too different.

I can even work with all three recordings now. I can easily align the left and right samples to the mid sample and mix the mid in, capturing more character without phase cancellation

Going to take a looooong time before the next version of this piano is done, lol. 85 keys, cutting, mapping etc