Noise problem in active crossover

[docaudio]

I had this post in 'chipamps' but thought it might belong here.
I've been hot rodding an older Ashly active crossover, (new caps, opamp swap, etc.) and have been trying to track down some noise issues.
Here is what I've done so far:
I've have swapped out the opamps for LM4562s. The coupling caps (c20,40,45,50,73,76,77) have been replaced with higher values (220mf)
I also had to change the voltage regulators to + 15v.
Here are some initial observations:
Things are much quieter than before, but I'm getting a bit of 'rocky-spity' noise. Someone is not happy.
I've a little oscillation on the outputs way up high (30 Mhz area) but just above the noise floor.
Also, the sound is a bit brittle.
Any other suggestions would by greatly appreciated.
Thanks,
Bill
(Schematic)

[Nrik]

The problem with swapping opamps, is that the original circuitry around is optimized for the original type.
You might wanne check the data sheet for your new opamps and the change C38, C39, C41, C44, C68, C69, C74 and C75 accordingly.

[docaudio]

Thanks Nrik,
I had been looking at those caps as well, but I'm not sure which way to go. Bigger or smaller? Any idea what spec on the data sheet should I compare between original and replacement that will help determine the ballpark value? I had studied the 4562's data sheet carefully, but may have missed something important.
=Bill

[Conrad Hoffman]

The 4562 is IMO a fantastic part, but it's fast. You have to be careful swapping it into circuits that were laid out for slower parts. Be sure to use small ceramic or film bypass caps right at the power leads of every op-amp. If any stage has large loops and isn't stable, install a local feedback cap in the small pF range, right at the leads. Obviously you need to check everything with a fast scope. There's a fair amount of "stuff" in that circuit, and the signal to noise ratio is set right up front in the diff amp with a gain of 1. You'll never do better than what happens in that first stage, so something to think about is gain redistribution. If you change the front end gain to 2X, and attenuate the last stage by the same amount, voila, instant 2X noise improvement. Of course you lose headroom, but how much do you really need?

[docaudio]

Great info Conrad, thanks!