Off the shelf versus PC for high-end DAC + Streamer +DSP

What exactly do you mean by "huge"?

I don't find latency to be a problem AT ALL. It's under 50msec for sure. Unless you are doing something wrong? It's more about how large your buffering is (number of samples and sample rate).

Hi with latency I thought of this one: Resplendence Software - LatencyMon: suitability checker for real-time audio and other tasks

And the main point was: A new fast PC, is not necessarily better in regards to internal latency than an older one. That is if using USB 2.0:)
 
If you can't answer that question, it's clear that YOU don't understand the difference between IIR and FIR filters, what they can do and can't do, etc.

Using your logic, one could post the question that "none of the best speakers in the world use DSP, so why use it in the first place?".

I bet you know a thing or two more about FIR-filters than me. My point was not at all to say that IIR is better than FIR or the opposite. Of course you could use FIR in the same manner as IIR, adding enough tabs so that what you acchieve better accuracy in frequency response, than possible with IIR.

It is also very easy to go all in, thinking that in the digital world there is free lunch to do what you want, and that just translates into perfect sound. For instance it is easy to make a crossover with 300 dB steepness and straighten phass to linear, from 20 dB past 20 Khz.
 

Here is a quote from the web page you linked to:
The audio latency problem

Windows is not a real-time operating system. All requests to the operating system are delivered on a best effort basis. There are no guarantees whatsoever that requests are delivered within a certain time frame, which are the characteristics of a real-time operating system. That is not a problem for most devices and tasks but this is bad news for audio applications (which are considered soft real-time) because they need to deliver data to the subsystem and the hardware in buffers several times per second. If one or more buffers miss their deadlines and are not delivered in time it has audible consequences which are recognized as dropouts, clicks and pops.

This is true, but it is completely a non-issue! A computer does not process data like a hardware DSP. In a hardware DSP samples are processed via something like an assembly line. As a sample enters, another one leaves, and there are a fixed number of samples inside the "DSP factory". A precision clock deteremines when a sample enters or leaves the "factory", thus keeping an accurate sample consumption/production rate. Inside, samples file along the "assembly line" in single file from input to output. It's a real time, completely "on line", serialized process.

A computer doesn't work that way at all. A computer handles audio data using buffers. These are more like buckets of samples. In the factory analogy, at the entrance to the factory samples are loaded one by one into a bucket at the "loading dock" until the bucket is full. Then the bucket is brought to the factory floor where DSP will be performed on all the samples in the bucket. There may be other buckets ahead of it that have not yet been processed, and the process make take some time so there might be "lunch breaks" where the workers go off and eat, go the to the bank, restroom, or whatever. When the workers come back they just resume where they left off. When all the samples in a bucket have been processed, the bucket is shipped to the delivery dock. There may be other buckets ahead of it that are waiting to be offloaded. Trucks arrive to pick up the delivery that hold one bucket at a time, and take one bucket away. These trucks come on a more or less regular schedule but not exactly at the same interval. They take the bucket of samples off to their destination. At the destination, a precise clock is used to spit out one sample at a time from the bucket. The above is a good analogy (I believe) to audio processing on a computer.

The workers must take some lunch break time, because they have other things to do. This is like the CPU, which is not only processing audio but running everything on the computer. Audio samples are processed in batches, not continuously. Overall there is a steady rate of consumption and production of samples on average, but the rate may fluctuate up and down from instant to instant. Only at the samples' destination (which is the DAC) is a precise clock used to produce samples in a constant stream, one by one.

When using a computer to process audio, there must be enough "slack' in the process to accommodate the largest amount of latency of the entire process, end to end. But as long as that requirement is met, it can be thought of as a black box with samples entering and leaving at a constant rate, just like a hardware DSP.

And the main point was: A new fast PC, is not necessarily better in regards to internal latency than an older one. That is if using USB 2.0:)

The above statement is misleading. A faster PC will indeed be able to process all the samples through the "factory" faster, so the overall latency can likely be made lower. That has nothing to do with USB per se. If you are using a USB DAC so samples are being sent (via the delivery truck) over USB, there will be a maximum rate that can happen, and thus there is some latency associated with the process and THAT latency will be independent of what PC is being used to host the USB DAC.
 
CharlieLaub: Thanks for at great post! Also fun to read the anology.

To my memory adding more latency/slack, up to 500 ms, did not improve the dropouts. It is clear that the less slack one needs, the more critical it becomes. In using DAW, were short latency is often required, it is often mentioned, that settings on a device is important. For instance in this (normally critical about links and this one might not be the best, but it goes through quite som settings): What is the best way to optimize Windows 10?

Staying in the realm of the factory anology, my experience seemed to be in like of a sudden strike of thunder or an earthquake. Things that is not meant to be there but happen randomly. This is why I mentioned that a faster PC per se is not always better if for instance antivirus wrecks the line.

It might mostly be and OS issue of the past.
 
I think this post by 4real is really nailing it:

"[...] The biggest issue I think is that most people don't actually know how to create the correct filters, and also not how to do proper measurements and know how to interpret them. THAT is actually what you should focus on. THAT is what makes all the difference in the end."

Then we have taste of drivers, building cabinets, how the room interacts with the cabinet and so on. It may take several years to make something that is actually better than what you can buy second hand + simple DSP correction.

I firmly believe that for a beginner into DIY-speakers, the best starting point is a setup, where a lot of test can be done in a short amount of time. If one has never even mingled with DSP (ie. simple notches/PEGs), it is hard to believe how much a difference it can do to the perceived sound.
 
Last edited:
Disabled Account
Joined 2009
What is the best way to optimize Windows 10?

is to get rid of it.

download a trial version of windows server essentials 2019- gives you a 6 months license and you loose all the crap ware. Unfortunately some PC have been customized by their vendor and will not run it (e.g. a Dell Latitude 7300). If you like it then you buy it.

I've got an HP Probook G5 and it is running windows server standard 2016 happily. No cr@pware being shoved down my throat. The cost of pruchase is offset by not having to worry about it or fixing things. But if you are into all the free apps bloatware then this is not for you.

AM