I am using REW to set up the 12 oscillators (top octave of notes) for an electronic organ, as well as the four octave divider notes generated from each of the 12 oscillators (ie. a 12 x 5 = 60 note span). Each oscillator is valve-based LC, so may have some noisy variation of frequency, and some drift with time, but I'm hoping to expedite the setup of all 12 oscillators by using REW on the bench under consistent operating and test conditions.
RTA allows easy recognition of the harmonic structure of each note, whether the note is from the highest frequency oscillator, or a divided note from that oscillator. I am using the frequency measurement display in the Distortion Box to tune each top-octave oscillator. The aim is to set each oscillator to a frequency resolution of circa 0.1Hz, eg. 1760.0Hz for the top A oscillator (for a 440Hz middle A on the keyboard), which is then within 1 beat in 10 seconds as a difference between an accurate frequency. The oscillators are exhibiting a repeatability of within 0.1Hz so far.
Following the REW help advise, I am using a flat-top window with a reasonably high FFT length, and a lowish sampling rate of 48kHz. I am presuming that gives a nominal 48/256 = 0.19Hz bin size for a 256k FFT length. My uncertainty is that the frequency measurement result has a 0.01Hz resolution, and I can observe a bit of change in the last digit as I change FFT length, but not really any significant change (ie. <0.05Hz) in value for FFT lengths from 128k up.
I guess that if the actual note's fundamental frequency is within the span of the bins, as FFT length is changed, then the calculation of frequency (to 0.01Hz resolution) won't significantly change, although I guess the absolute accuracy of the measured frequency still has an uncertainty related to the bin size? If so, then I can happily set the frequency at a lower FFT length, and then confirm that frequency doesn't change at a larger FFT length.
Ciao, Tim
RTA allows easy recognition of the harmonic structure of each note, whether the note is from the highest frequency oscillator, or a divided note from that oscillator. I am using the frequency measurement display in the Distortion Box to tune each top-octave oscillator. The aim is to set each oscillator to a frequency resolution of circa 0.1Hz, eg. 1760.0Hz for the top A oscillator (for a 440Hz middle A on the keyboard), which is then within 1 beat in 10 seconds as a difference between an accurate frequency. The oscillators are exhibiting a repeatability of within 0.1Hz so far.
Following the REW help advise, I am using a flat-top window with a reasonably high FFT length, and a lowish sampling rate of 48kHz. I am presuming that gives a nominal 48/256 = 0.19Hz bin size for a 256k FFT length. My uncertainty is that the frequency measurement result has a 0.01Hz resolution, and I can observe a bit of change in the last digit as I change FFT length, but not really any significant change (ie. <0.05Hz) in value for FFT lengths from 128k up.
I guess that if the actual note's fundamental frequency is within the span of the bins, as FFT length is changed, then the calculation of frequency (to 0.01Hz resolution) won't significantly change, although I guess the absolute accuracy of the measured frequency still has an uncertainty related to the bin size? If so, then I can happily set the frequency at a lower FFT length, and then confirm that frequency doesn't change at a larger FFT length.
Ciao, Tim
