Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Paul W]

MC works fine at 96k.  Switching to 48k sometimes triggers an oscillation in the LFE channel that grows from a low level to through-the-roof.  Sounds almost like acoustic feedback except the oscillation is sustained (as viewed at the DAC) even after switching off the power amp.  Switching back to 96k (with no other changes) works fine again.  

Suggestions?

[RD James]

Upsample 48K content to 96K via DSP Studio?

[Paul W]

That would normally be okay, but the end-game involves tri-amping Netflix and convolution from the WDM driver.  Multi-channel latency is a big concern & upsampling after low-latency convolution would be tough (for me) to pull off.  I'd prefer understanding the root cause to prevent the problem.   

I'm taking baby steps toward convolution and the oscillation problem is with standard JR PEQ equalization...no convolution in the path.  Could the current PEQ settings cause instability at 48k but not 96k?

[ferday]

whats the oscillation frequency, is it slow bang or fast like a "motorboat"?

normally oscillations are related to the amp or other circuits, sometimes an impedance mismatch, etc.
it's unusual that it seems to only happen at 48K but it's even more unusual that just software (without convolution filters) could cause it

mwillems or mattkhan may have some good input...

[mattkhan]

That would normally be okay, but the end-game involves tri-amping Netflix and convolution from the WDM driver.  Multi-channel latency is a big concern & upsampling after low-latency convolution would be tough (for me) to pull off.  I'd prefer understanding the root cause to prevent the problem.   

if the filter length remains the same, the higher sample rate will cut the latency of the filter in half albeit at a cost of reducing the frequency resolution by half & increasing CPU utilisation. This might be a good tradeoff to make in your case. Having said that, what's your latency budget for this? convolution and WDM (hence video that is not delayed) is a tricky thing.

I'm taking baby steps toward convolution and the oscillation problem is with standard JR PEQ equalization...no convolution in the path.  Could the current PEQ settings cause instability at 48k but not 96k?

I think you need to post your settings & how you are testing to see if someone else can reproduce. There isn't much to go on atm.

[mwillems]

I think you need to post your settings & how you are testing to see if someone else can reproduce. There isn't much to go on atm.

I concur: post your settings and signal chain; there's no way to know what's happening without that.

[Paul W]

The oscillation is below 20Hz...physically feel it before it gets loud enough to be heard.  Sounds a little like motorboating, but more like the acoustic feedback via turntable when subs are cranked too high.  Like mentioned previously, this oscillation keeps going even after the power amp and subs are shut down.

The only thing "unusual" about the sub setup is that LCRs are all Added to User 1, U1 lowpassed at 75 Hz, then Added to the LFE channel which is then lowpassed at 300.  Otherwise the sub channel is "business as usual"...Output Format is 16 channels, no JRSS, Room Correction used only for reducing levels (could easily change that to PEQ1).  Other than Addition and LPFs, PEQ1 is a slight delay, -3db notch at 40Hz, -5db notch at 110Hz to correct response bumps.  Hardware: PC w/MC>AES16e card>Aurora 16>power amp>subs.  Test method is to watch a NF movie>unpredictable holy shet>sprint to the sub amp switch!!!

I'll try Matt's suggestion before moving on to mojave's vote for a Hauppauge Colossus 2 and input Roku through a splitter.  Only trouble with the latter is I don't know if a PCIexpress slot can be freed up...does anyone make an equivalent PCI solution?  Otherwise MB replacement...which would be a real PITA.  Unfortunately I can't get back to this for several days.

[mattkhan]

sounds v odd

if you want to debug then I would remove one thing at a time from dsp studio and repeat the test to see if you can work out what sets the feedback going. It seems like you must have a loop going in your mixing somewhere that you don't realise is there. I would probably start by removing the PEQ for modal correction then removing the room correction block, i.e. work your way back to just the routing/mixing. I would also set things up to capture the output via loopback so you can avoid blasting your ears apart

[Paul W]

Will do, although it will be a few days.  I did confirm that a PCIe slot can be made available, so there is that option too.

[Paul W]

Forgot to update this ...

Oscillation was caused by routing LCR LF to User 1 with a 75Hz low-pass and then on to the Sub/LFE channel with it's own 300Hz low-pass.  Removing the intermediate step and changing the Sub low-pass to 75Hz killed the oscillation.  Problem was not limited to 48k...it was volume control sensitive and I didn't notice the system could also oscillate at 96k.