Psymodel

[friesendrywall]

I am not sure where to ask, but LAME and shine have similar 'calc_xmin' functions, but the shine function doesn't do anything inside other than setting l3_xmin->l[gr][ch][sfb] = 0; inside the loop. Could I get some overview of what it would take for me to port this?

If this is the wrong place, please advise the correct place to ask.

[toots]

Hi,

Any contribution would be much welcome on this front! I'm not sure what it would take except that you'd have to adapt any float-based math to use fixed-point.

[friesendrywall]

I have been working on this in a different model, more based on Lame's low pass filter. Implementing a hard low pass filter by setting upper spectral values to zero's seems to help some, but is there a way to be sure that almost no bits are allocated to the chosen spectrum?

In shine_iteration_loop in l3loop.c

	  for (i = GRANULE_SIZE, config->l3loop.xrmax = 0; i--;)
 {
		  if (i > 287){//Set all spectral values above this to 0
			  config->l3loop.xr[i] = 0;
		  }
		  config->l3loop.xrsq[i] = mulsr(config->l3loop.xr[i], config->l3loop.xr[i]);
		  config->l3loop.xrabs[i] = labs(config->l3loop.xr[i]);
		  if (config->l3loop.xrabs[i]>config->l3loop.xrmax)
			  config->l3loop.xrmax = config->l3loop.xrabs[i];
	  }