I’ve checked the new spectrogram and saw that it is showing only middle channel - the sum of left and right. Unfortunatelly sounds are not shown when played in side channel.
Take a mono sample, convert to stereo and invert left or right channel then play it and you will not see anything on the spectrogram altough the sound is playing.
So…
It would be good to fix it while Renoise is in beta. Fix could be applied as an switchable option, because it takes time: two spectrums have to be calculated separately for left and right channel and frequency/amplitude data added afterwards instead mixing two channels before transformation.
If you invert one of the two channels of a mono sample, that is not a bug of the spectrum graph.The sum of left channel and right channel bits simply balance eachother out and this result you see on the spectrum.
If you would do a mix-down of both the samples that you just converted to stereo with the inverted channel, i bet you also get a “silent sample” because that is the result of mixing channels that cancel eachother out.(Try it: convert to mono with mix option)
… and this is the point - Renoise makes mixdown to mono and then does the FFT. If it had been doing sepearte FFTs for left and right channel and then had mixed it down then analysis would have shown left channel, right channel, middle channel and side channel on the same plot.
I get how this is a problem, and a way around it would be to have two scopes, one each for L and R. However… However, if you have phase inverted material across the side channel then really you’ve got a problem in your mix, because the mix will sound gutless on mono playback systems. The phase meter tool helps you here, so you could look at that as your failsafe checker.
And if you’re on Win or OSX then there are suitable alternatives with highly regarded free 3rd party plugins.
One scope for left and right channel mixed after FFT, instead one scope for L and R mixed before FFT. Doesn’t matter if my mix sounds gutless or not, just want to see side channel together with middle. (To compare if middle and side are in pitch together or something else… just to see everything that i hear)
No worries. It’s good to have other people agree and bring it up using their own description and reasoning. Hopefully it will be listened to as a lot of good work has already been done with the spectrum in this release but this is still one fairly major flaw with it as far as I’m concerned.
Adding a separate fft for each channel will increase the cpu load even more, and other people will complain about this.
I can consider adding an option to select the summing mode > Left, Right, Mid, Side, L+R but definately an fft for each channel is heavy, consider using a block of 8192 (overlap of 4) so you have 4 fft per block * 2 analysed track * 2 channels = 16 ffts… your cpu will burn, and especially the netbook user ones
We could use some specific CPU profile options (Like in the old days) in the preferences, where low profile would disable all overhead and luxury functions, medium would support most visual feedback in a basic mode and high or ultra would allow users to demand stuff that no regular laptop or desktop PC can handle, but you could manage with an i7 CPU for instance.
Another thing you could then use is the CODA GPU to do all these FFT processing (And a lot faster and better than the CPU) Kraken since every graphics card seem to have one that does a decent job, why not abuse a lot of processing time from the GPU instead of the CPU? (I don’t think graphic cards has less than 128MB of ram these days and 256 or 512MB is quite common and would also suffice for buffering i presume.)
What we’re talking about is options. You’ve introduced some very nice options in the latest spectrum area but left a mode of operating that at times can be misleading.
Isn’t that what you meant by your L+R option? Or are you going to put a separate trace for left and right? So if we have both track displays (current track and master for example) in L+R/Stereo we will actually have 4 traces on display? Or disable the second track display if we want to use stereo? Seems a shame…
Is an averaging of the FFTs that computationally hungry?
it’s not the average that will create overhead, is having one fft for LEFT and one for RIGHT instead of a single one for MONO signal that will double the cpu usage of the spectrum.
well, i can lower the overlap (aka lowering the framerate by 2x) to achieve the same cpu usage as it is now for stereo fft averaging, but the peak will respond at half the framerate.
It would be 20% of my CPU at maximum FFT block size (2 cores, 2.1 GHz). Not that bad. I use mostly prerendered samples, so lots of CPU usage is anyway useless and could be put into analysis…
I’m not sure if lots of users use maximum CPU performance, anyway it is easier and faster to prerender everything into samples. If it was a switchable option it would be possible to turn it off in case of lagging. I think there is no reason for leaving it unimplemented.
Current maximum settings use around 7% of my Core2 series 2.83GHz quad core (actually just retested since the old thread (B2?) and seems to be more like 11%.) I sure for the people with i7 series processors it is much less still. I don’t think CPU usage because of a few netbook users, when there are lower use options available, is really a valid argument.
So what is the difference between your mentioned Mid and L+R modes? Mid signal is the sum of left and right after all. So that only left the conclusion the L+R would show a curve for each or would average the FFT results.
