Recommended PC based DSP hardware and software?

One more recommendation for the xonar u7: if you experience stuttering/desync, please try adding tsched=0 to the module-udev-detect line in your $HOME/.config/pulse/default.pa AND if you want to make it permanent to /etc/pulse/default.pa .


I don't have a patreon page. Didn't think that one would be necessary. There's a paypal donation link on the website though. Do you think there would be any advantage if I had a pateron profile? After all this software serves such a small niche market 😉
 
I do love a good white russian.
I miss so many of my favourite bars and pubs during lock down, I cant wait for them to open again.

Yeah, Futurama is hands down one of my favourite shows. I usually rewatch an episode or 2 every week, and have done for quite a few years so I'd estimate I've probably seen every episode at least 10 times.
I think I can relate to it, as a kid all i wanted was to travel into the future and have a robot as a best friend. Also in my 20's I worked in a pizza shop 😀
 
An externally hosted image should be here but it was not working when we last tested it.
I thought I share a picture of my set up with your software in action.
 

Attachments

  • 20200616_131915.jpg
    20200616_131915.jpg
    646 KB · Views: 172
Very nice, thanks! Did you actually use the support for FRD files in the software? Up to now it seems that nobody ever used that, at least I have no feedback whatsovever.



FYI: The link to open the image does not work. I copied it to a new browser tab and then it works...
 
Very nice, thanks! Did you actually use the support for FRD files in the software? Up to now it seems that nobody ever used that, at least I have no feedback whatsovever.



FYI: The link to open the image does not work. I copied it to a new browser tab and then it works...

Yeah some weirdness with google drive links, I've found forums dont like them, I'm not sure of a workaround.

I haven't used the FRD files yet, I'm still waiting for my measurement mic to arrive, but I'll keep you posted.
For now I just tested the setup with pink noise and a spectum analyser on my phone. I found the ideal cutoff for the sub where I could see the speaker bass response falling, then I level matched the speakers and sub. just something as simple as that has made the world of difference.
I'll share further adjustments when the measurement mic arrives and I'll get to grips with REW.
 
I have a technical question.
How are levels and clipping handled by your software?
If i apply an EQ boost for instance, should i lower the overall output gain of that EQ to prevent clipping, or is there internal headroom or something within the processing?
Perhaps ive got some terminology incorrect there, but I hope you understand the question.
 
It depends on what you set the sampling format of the pulseaudio server to. If you use my recommendation - float32le - then no clipping can occur until the final step of resampling for the output format of the soundcard. any values >1.0/<-1.0 will lead to clipping there.

In my setups I adjust the overall gain so the gain curve of the output (combined from all filters in the chain) never goes above 0dB or to be more precise I tend to leave 2dB of headroom for "intersample-overs".

One of the next version will have a way to monitor levels at input and output including peak detection and display.
 
It depends on what you set the sampling format of the pulseaudio server to. If you use my recommendation - float32le - then no clipping can occur until the final step of resampling for the output format of the soundcard. any values >1.0/<-1.0 will lead to clipping there.

In my setups I adjust the overall gain so the gain curve of the output (combined from all filters in the chain) never goes above 0dB or to be more precise I tend to leave 2dB of headroom for "intersample-overs".

One of the next version will have a way to monitor levels at input and output including peak detection and display.


perfect, that pretty much answered what I had assumed.
I like the idea of having meters which can be applied at different stages in the chain, that will take any guess work out of it.


I had been using s24le as my bit depth, because my interface has a maximum of 24bit, so i assumed its best to keep it at 24bit so it doesnt need to be converted, or does it not matter?
also, i find i get some off behavior when i use 32bit with 192,000khz, one channel fading in an out and stuttering from the sub.
But 32bit works ok at 96,000
 
Last edited:
It will have to be converted to float anyways because LADSPA plugins internally only work with floats. So using float32le as recommended in the online help is probably your best choice. Using float also the gain structure in the middle of the filter chain does not matter, as values >1.0 are handled gracefully, there is no clipping with float, only at the last resampling step to an integer value.
 
It will have to be converted to float anyways because LADSPA plugins internally only work with floats. So using float32le as recommended in the online help is probably your best choice. Using float also the gain structure in the middle of the filter chain does not matter, as values >1.0 are handled gracefully, there is no clipping with float, only at the last resampling step to an integer value.


OK good to know, thank you.
I've currently got it running stable with out issue as follows
DSP: 96khz/float32le
HW: 96khz/s24le
 
I got to the bottom of my clicking issues.
The laptop I'm using as my streamer and DSP was running hot. I opened it up and cleared out the air vents. there was a good 8mm of dust blocking the entire fan exhaust. I wondered why the fan was working over time lately.
The CPU's were hitting over 70 degrees C and not running at more than 20 percent usage per core, it was obviously thermal throttling causing the clicks at the higher bit depth and sample rate.
It runs cool and silent now even at 32/192 🙂
 
Yes, I'll probably have to do this to my main workhorse (also a laptop) here sometime soon. It's 2.5 years old now and the fan is running a lot more than at the beginning.


Do you have timer based scheduling on or off in the pulseaudio config? Just asking out of curiosity...
 
I am afraid synced clocks do not guarantee the playback/DMA transfers for both cards will start at the very same moment => fixed delay

I've tried this on linux with two Asus D2X cards running from the same oscillator and there was no audible delay between their outputs.
I've tested it with earbuds, one earbud was connected to one card, the other earbud to the other card. The sound was exactly in the center.

This is of course not a guarantee, since linux is not a real time operating system, so a context switch at an unlucky moment might still cause a time difference between the cards.