I had troubles with the tuning of the filters by noise. I therefore always do it in way I do it on all other analog synths VCFs.
See, the first goal of the tuning is a sychroneous tracking of the filter frequency on all voices. no matter what the absolute frequency of the cutoff actually is, I should in any instance be the same for all voices. The absolute cutoff for all filters can easily be adjusted in the patch via the filter pot on the panel.
The first step of the tuning is to find the absolute offset frequency. Be aware that filter tuning can be properly performed only when the KCV (tracking) is present. Do not forget to activcate "track". This is actually the second step. to first step is to allow the voice boards to warm up. Depending on you local conditions 20 mins will do.
Now switch off all signal to the filter (OSC1 off, OSC2 off, NOISE off). Then crank up resonance on all voices, but only to a level where self-oscillation just begins, but is stable. Not more! Since heavy resonance might lower the resonance frequency and thus the apparent cutoff frequency of the filter which would then lead to misalignment. Resonance occurs quite easily in the range of more than 250Hz.
Check whether self-oscillation can be accheived in the same manner in all voices.
(If not, one of the CEM3320 may be out of range (replace). In most cases I had here which showed trouble with the filters, the problem was in the 741 OP-amps processing the incoming CVs.Or even in the TL084 OP-amps generating the CVs on the CPU boards. I don't know, what is the reason for this, since 741 and TL084 normally are pretty robust. So if a voice board is bad, do not always blame the CEMs for it.)
Now take a chromatic tuner, to tune all voices corresponding to the key pressed on the keyboard. Take care to adjust the voices to each other rather than to the absolute note value.
Then proceed with the slope adjustment. Do not expect a perfect tracking! This is not a digital synth and the design of the OB-Xa in this respect is poor. Acceptable tracking can be acchieved over two octaves only - at most, if not less. For your filter tuning use the note range you are using most.
Note that offset substantially changes with the adjustment of the slope.
I quickly alternate between C2 and C4 on the keyboard (activate 1 voice only, of course) and simultaneously adjust offset and slope until a two octave jump is heard.
This is laborous and tricky, but you can manage it with the filters in self resonance. It would be impossible with the noise method described in the manual.
Repeat for 4-pole.... ;-)
Then, before switching on the oscillators, play a little tune with the resonating filters on the keyboard. And make a break.
I recommend to lower the resonance setting back to "slighly below/before self-oscillation". Verify the absence of heavy self-oscillation at higher frequencies/notes to rule out any surprising "resonance attack" from one or more of the voices during playing. IMHO, the OB-Xa is not the right synth to shine at self-resonance, one should not go for it. Besides, the factory patches may not sound like they were meant to with a wrong resonance setting.
(Speaking of authetic factory patches, consider to adjust both LFOs on the CPU board to about 3 Hz with LFO pot twelve o'clock. SPLIT setting (of course with the same patch on upper and lower). I experienced heavily misaligned LFOs in my OB-Xa s which has considerable impact on the sound of some of the patches and in particular in DOUBLE mode using the same patch on both parts of the machine.)
Another alternative method for filter tuning I had once used, does not require a full self-oscillation (but high resonance!) and no chromatic tuner: Swith on HALF OSC2, pulse wave 50% (square) and adjust the filter for maximum amplitude of the sine wave signal leaving the filter. At high resonance settings the low pass acts somewhat like a band pass an the cutoff frequency, i.e, the (basic) frequency of an oscillator signal is let through at maximum when the cutoff frequency corresponds to the frequency of the oscillator.
This method works nice, but you have to be aware of that the oscillator's signal somewhat shifts or influences the filter's actual cutoff frequency. That is why Oberheim originally recommends to use noise as the input signal instead of a tuned oscillator signal.
Any comments, please share your personal experience with the filter tuning process.