- Posts: 5
- Thank you received: 0
jsg wrote: Agree with the above comments. The process I usually use goes something like this:
1. Check the drivers are all set to the same polarity (+ve signal=forward cone motion)
2. Rough-estimate GD for each channel (obviously better to measure if the kit is available). Estimating means adding up various numbers:
- path length difference, i.e. shortest path from cone to front of stack, convert to time using 1 foot->1ms
- delay due to speaker upper rolloff, this is the effect of a low-pass filtering effect. Delay is approx order/(4*freq) ms. Order is usually 1 for direct radiator and simple horn, 2 for bandpass horn and simple bandpass, 4 for complex (3-chamber) bandpass. Use the driver's own upper rolloff, not xover freq.
- delay due to resonant peak. Hard to estimate. You could use a simulation for a rough figure. I usually don't cross over near a resonant peak though, in which case this figure can be omitted.
- delay due to xover rolloff. Same formula as speaker rolloff, i.e. order/(4*freq) ms. Use the upper cutoff freq set in xover for that range (skip this part if no upper limit). order here is slope/6dB. So eg 1KHz 24dB/oct gets you order=24/6=4 and delay=4/(4*1000)=1ms
Add all these up. Note that these are all approximate. Really, this stuff only needs to be accurate enough to get you within 1 period of the correct delay.
Now just set the relative levels to taste, set the xover slopes so that both sides are at -6dB at desired xover freq (allowing for any roloff theat the speaker is doing) and set the delay of the upper side to match using the above figures (so upper delay setting = lower delay setting + lower calculated delay - upper calculated delay).
Now flip one side into invert, run a sinewave at the desred xover freq and search for a null by tweaking the upper delay and the freq of the sine wave. Now flip off the invert and you're all good.
I do the estimation part first because, as both Rog and Chris mention, total group dealys can be a lot bigger than jsut the path length component. You can (and I have) found a phase match using the nulling method that was off by at least one entire cycle at the xover freq.
For example in the TY low mid->high mid (800Hz), the path-length difference is less than 1 ms. But one I factor in the TY mid's rolloff order (which is high due to being a bandpass system) and the xover rolloff and I got nearer 3ms difference. So I searched for a match by nulling near 3ms and was happy with the results.