• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Kernel Streaming, ASIO, WASAPI... and music players (Foobar, JRiver...)

Veri

Master Contributor
Joined
Feb 6, 2018
Messages
9,596
Likes
12,036
The sound is more analog, with more body. More pleasant. More detailed sound -> that is why I passed a few tests that PMA did in diyaudio.com. With foobar2000 I used to have difficulties as the difficulty of the test increased.

Music with acoustic instruments and "natural" voices is what I usually listen to. Old recordings, from analog master. Vinyl, CD or SACD the vast majory. Some files from HDTT web.


How to make the best FLAC. Convert, recoder and CD rip.

https://www.audiocircle.com/index.php?topic=157417.0

PS: now the SACD sounds better with JRMC too than foobar2000 + SACD plugin.

How to make the best FLAC?

'hol up now :eek: I've been doing this wrong all these years? /s
 
OP
daftcombo

daftcombo

Major Contributor
Forum Donor
Joined
Feb 5, 2019
Messages
3,687
Likes
4,068
Mine is highly unoptimized and it only uses 0.9% of an i7 processor for DSP and playback... :eek:
View attachment 26883

I'm interested in knowing how you can do achieve this low CPU usage, because everytime I use Foobar or JRiver with Convolver, the fan gets going after 10 seconds and never stops. I have a HP Envy laptop with i7, 8 Mo RAM, bought last year.

I also have a fanless PC, much cheaper (70€ used), which can play with JRiver + Convolver without problem.
I wonder what's wrong with the Envy...
 
Last edited:

BillG

Major Contributor
Joined
Sep 12, 2018
Messages
1,699
Likes
2,266
Location
Auckland, New Zealand
I use Foobar or JRiver with Convolver, the fan gets going after 10 seconds and never stops. I have a HP Envy laptop with i7, 8 Mo RAM, bought last year.

I'm not using convolution in my DSP; it can be a rather CPU intensive process. I just use EQ, Room Simulation, and Analog Amplifier Modeling in the form of iZotope Ozone MP... :cool:

As for your laptop, I'd check to see what other processes that are running to see if they're causing CPU spikes. Additionally you may want to check out your power management settings to reduce the fan-on time.
 
Last edited:

edechamps

Addicted to Fun and Learning
Forum Donor
Joined
Nov 21, 2018
Messages
910
Likes
3,620
Location
London, United Kingdom
everytime I use Foobar or JRiver with Convolver, the fan gets going after 10 seconds and never stops

How long is your FIR filter? Direct convolution is O(n²), which means it can become quite expensive with more than a few hundred taps or so. Fast Convolution algorithms are O(n log n) which is much faster with long filters; my PC with a modern CPU can do 65K+ taps on multiple channels without any noticeable CPU usage when using such algorithms.

If the convolvers in Foobar or JRiver are using Direct Convolution, that might explain why CPU usage is high when you use them.
 
OP
daftcombo

daftcombo

Major Contributor
Forum Donor
Joined
Feb 5, 2019
Messages
3,687
Likes
4,068
I'm not using convolution in my DSP; it can be a rather CPU intensive process. I just use EQ, Room Simulation, and Analog Amplifier Modeling in the form of iZotope Ozone MP... :cool:

As for your laptop, I'd check to see what other processes that are running to see if they're causing CPU spikes. Additionally you may want to check out your power management settings to reduce the fan-on time.

I've tried a lot of things, including messing with Windows fan options, updating the BIOS, cleaning the fans, changing the CPU/GPU thermal paste. The PC is very clean as I think twice before installing even a safe software.

Still, the i7 (quad core) is used at 40% since start-up, which is strange. In fact I never have problems running any software BUT convolution in Foobar/JRiver.
 
OP
daftcombo

daftcombo

Major Contributor
Forum Donor
Joined
Feb 5, 2019
Messages
3,687
Likes
4,068
How long is your FIR filter?
It is a 44.1 kHz 64 bit stereo WAV impulse using 65.536 taps. But with 4.096 taps it is more or less the same.

I think it is not as bad with EqualizerAPO but I don't remember well.
 

edechamps

Addicted to Fun and Learning
Forum Donor
Joined
Nov 21, 2018
Messages
910
Likes
3,620
Location
London, United Kingdom
It is a 44.1 kHz 64 bit stereo WAV impulse using 65.536 taps. But with 4.096 taps it is more or less the same.

You might want to switch to 32-bit float. It will likely increase performance and I wouldn't expect it to make any significant difference in terms of quality.

I think it is not as bad with EqualizerAPO but I don't remember well.

IIRC, EqualizerAPO uses Direct Convolution, so it won't be fast enough for filters this long.
 
OP
daftcombo

daftcombo

Major Contributor
Forum Donor
Joined
Feb 5, 2019
Messages
3,687
Likes
4,068
You might want to switch to 32-bit float. It will likely increase performance and I wouldn't expect it to make any significant difference in terms of quality.

JRiver v24 is 64bit, with 64bit gain control (in "internal volume" mode) so I thought it would be better to keep all at 64bit, but you are probably right.

In RePhase, when the number of taps is reduced but the red & blue curve are very similar even below 50Hz, there is no big deal for sound quality right?
 

edechamps

Addicted to Fun and Learning
Forum Donor
Joined
Nov 21, 2018
Messages
910
Likes
3,620
Location
London, United Kingdom
JRiver v24 is 64bit, with 64bit gain control (in "internal volume" mode) so I thought it would be better to keep all at 64bit, but you are probably right.

Well, if JRiver really is 64-bit end-to-end, it might just automatically reconvert your 32-bit filter to 64-bit and you're back to where you started. Doesn't hurt to try, though.

(The fact that the whole pipeline is 64-bit might be why you're observing poor performance. IMHO 64-bit is overkill even in an audio processing chain. 32-bit float is enough.)

In RePhase, when the number of taps is reduced but the red & blue curve are very similar even below 50Hz, there is no big deal for sound quality right?

Yes. Feel free to reduce the number of taps if the frequency response of the filter doesn't change - it will give you better performance with no downsides.
 

maty

Major Contributor
Joined
Dec 12, 2017
Messages
4,596
Likes
3,160
Location
Tarragona (Spain)
Your disqualifications remind me of others I suffered years ago when I stated that I easily appreciated the differences between MP3 320kbps and FLAC 16/44 in my two systems.

And later with 16/44 vs 16/48, 16/96, 16/192, 24/44, 24/88, 24/96 and 24/192.

As for the set of instructions I doubted it myself but I decided to take the test and notice the difference.

My formation is not esoteric precisely. Hence the importance that I give to measurements when judging audio equipment.
 

maty

Major Contributor
Joined
Dec 12, 2017
Messages
4,596
Likes
3,160
Location
Tarragona (Spain)
As I wrote before, I notice the difference often when I edit ripped files, in a quick listen.

I suspect it has to do with some type of jitter generated by the CPU and how it accesses the data, but I have not found any measurement in this regard, so it only remains in hypothesis. But the improvement is undeniable.

PS: I know a lot more things than just programming.
 

edechamps

Addicted to Fun and Learning
Forum Donor
Joined
Nov 21, 2018
Messages
910
Likes
3,620
Location
London, United Kingdom

Burning Sounds

Addicted to Fun and Learning
Joined
Dec 22, 2016
Messages
523
Likes
883
Location
Co. Durham, UK
If the two are configured the same way to use the same output pipeline, there will be no difference. I have tested this occasionally and that is the case.

The first time I tested foobar I thought it sounded a lot better. I had assumed it was using exclusive mode. Then I read that by default it was not. So I did a blind test and all the differences I thought I heard, went out the window.

I've recently completed a 60 day trial of Roon (offered via Qobuz) and could not find any difference in sound quality between JRiver 25 and Roon 1.6 even when using the volume control in each as long as the levels were matched. Neither could I hear a difference when using Acourate convolver files (x-over and IIR eq only - no FIR room correction) for my LX521s.

If you enter x-over/IIR eq parameters (esp. shelving filters) directly into Roon or JRiver's DSP you will hear a difference as they implemented differently. But this also applies to digital x-over/eq implemented in hardware such as MiniDSP's various boxes as well as other software. They are not all implemented the same way. Measuring the transfer function is the only way to be sure everything is implemented correctly if you are using digital x-overs etc.
 

Veri

Master Contributor
Joined
Feb 6, 2018
Messages
9,596
Likes
12,036
As for the set of instructions I doubted it myself but I decided to take the test and notice the difference.

Yes but is this difference verified ? Allow all of us to base our thoughts as definitely 'true', and you end up on a SBAF-like echo chamber forum.

As I wrote before, I notice the difference often when I edit ripped files, in a quick listen.

You see, this is not good enough. Not for a scientific-minded forum. No disrespect to you, but no one would take you / believe you on your word.
 

andreasmaaan

Master Contributor
Forum Donor
Joined
Jun 19, 2018
Messages
6,652
Likes
9,399
How long is your FIR filter? Direct convolution is O(n²), which means it can become quite expensive with more than a few hundred taps or so. Fast Convolution algorithms are O(n log n) which is much faster with long filters; my PC with a modern CPU can do 65K+ taps on multiple channels without any noticeable CPU usage when using such algorithms.

This is very interesting! I’d always thought the delay created by a FIR filter was a simple function of the sample rate and tap length:

Delay (in samples) = (N - 1) / 2

Presumably, fast convolution can’t happen faster than this; does that mean that direct convolution is actually slower?
 
Top Bottom