Testing frequency curve variations among software

[ryan_570]

I’ve been feeling frustrated with my mix’s sounding different after exporting from Cubase.

I always check the export in VLC player and sometimes Windows media player(WMP), and have always noticed they sound different, usually worse particularly (to my ears) in the 400Hz and lower region.

I decided I just have to test this for my sanity, and so I wanted to share the results with you here. And I have a couple of questions,

1. Should I care?
2. Do you care?
3. Are you aware of this?
4. If you are aware of this, what do you do to counteract it?
5. If these are known curves wouldn’t it be great to have a list of options at the stereo mix that could enable listening through these various curves, that everyone is going to listen to music through, including youtube, spotify, apple etc?

Of course to make these tests fair any audio enhancements in VLC WMP are not engaged and all files types (bit depth etc) are matching.

==============================================
I ran three tests, first was with a 4 minute song
Green = Cubase
Orange = VLC
Blue = Windows Media player(WMP)
Solid colour in front is average shaded is max

https://drive.google.com/drive/folders/ ... sp=sharing

At first glance they seem quite similar but the more you look the more differences you start to see between VLC/WMP to cubase. And yet VLC-WMP are almost identical.

================================================

The 2nd test was white noise.
Green = Cubase
Orange = VLC
Blue = Windows Media player(WMP)
Purple = Groove Music
Solid colour in front is average shaded is max

https://drive.google.com/drive/folders/ ... sp=sharing

This is where the difference becomes obvious.
Cubase is beautifully flat, however VLC and WMP are seriously… just ugly, Groove music is a slight improvement but still far from flat.

================================================

The 3rd test is a sine sweep
Green = Cubase
Orange = VLC
Blue = Windows Media player(WMP)
Purple = Groove Music
Solid colour in front is average shaded is max

https://drive.google.com/drive/folders/ ... sp=sharing

Again similar results

[Hugh Robjohns]

I've no idea how you made these tests, but comparing long-term frequency averages isn't really very helpful in pinpointing what's going on... and it also looks to me that there are some fairly characteristic time-based measurement system anomalies going on.

I wouldn't be surprised if Windows Media did some strange EQ-based tweaking to the replay files, but I'd be surprised if VLC did.

A better start would be to make an audio file with a few seconds of fixed tones at a constant level but at different frequencies -- third octave increments, for example.

If the system response is flat you would see a constant replay level on the meters when you replay the whole file. If there are response amplitude variations they would show on the meter.

[ryan_570]

The tests were quite straight forward. I ran two separate computers, one to play the files/cubase and the other to monitor the output via Voxengo.

My expectation was if I see the response as flat in Cubase then I should see the same result in the other players, if there is a difference, they cannot be flat right?
I was incredibly surprised by the result not just how different Cubase -VLC/WPM looked but how similar VLC/WMP came out.

I can definitely do a test how you describe.

Cheers for the reply

[RichardT]

Your audio may actually be different after export!

For example, if you use Kontakt, it will use the ‘perfect’ interpolation algorithm during audio mixdown, and a less good setting live in Cubase unless you have specifically configured it not to. This algorithm is used to generate the notes with pitches that are not explicitly covered by samples. This can make an audible difference.

Soothe does something similar.

Also check that you don’t have any processing in the control room when you do the comparison.

[Sam Inglis]

What sample rate are you working at, and are the files that get exported from Cubase at the same sample rate? I'm wondering whether these other applications are having to resample on the fly because for example the audio interface is working at 48k and they are trying to play back a 44.1k file. Although I can't quite see why that should make such a big difference.

[ryan_570]

Your audio may actually be different after export!

For example, if you use Kontakt, it will use the ‘perfect’ interpolation algorithm during audio mixdown, and a less good setting live in Cubase unless you have specifically configured it not to. This algorithm is used to generate the notes with pitches that are not explicitly covered by samples. This can make an audible difference.

Soothe does something similar.

Also check that you don’t have any processing in the control room when you do the comparison.

The file i'm using to play through VLC is actually pre export (white noise and sinesweep were downloaded, and the music was a CDrip), the files in cubase are the same, imported, without any conversion.
For extra security I actually did a comparison with the file generated by Cubase located in the pool, however this file and the downloaded file were exactly the same, suggesting that the file that Cubase creates is an exact copy and therefore any difference is not the file itself. The only variable is the player.

No processing of any kind has been done at any stage, these are clean empty projects, created just for these tests.

But its interesting you say the audio may be different after export, why might this be, other than bit rate changes?

[ryan_570]

What sample rate are you working at, and are the files that get exported from Cubase at the same sample rate? I'm wondering whether these other applications are having to resample on the fly because for example the audio interface is working at 48k and they are trying to play back a 44.1k file. Although I can't quite see why that should make such a big difference.

I simply matched the sample rates of the files. Which was a mixture of 16 bit/44.1 and 24bit/48. so the cubase projects were running the same.

That's a thought I had too. I wondered if VLC/WMP were converting before playback but I think if that was the case we'd see some buffering or something just for a few seconds. Imagine if the file was like an hour long, it'd take a few second to convert that surly, or could it happen realtime?

These tests are using the computers on board soundcard rather than running through the interface, as I didnt want that to be a variable. Its an interesting point you make there because (I'm yet to test this) I think using VLC/WMP and playback via USB interface, the sound appears to be flat... still have to test this.

[orange]

These tests are using the computers on board soundcard

I'm a little confused exactly how you are running these tests but are you listening/recording the files from different audio interfaces (external/internal) ?

The devil is usually in the details

[Sam Inglis]

I wondered if VLC/WMP were converting before playback but I think if that was the case we'd see some buffering or something just for a few seconds. Imagine if the file was like an hour long, it'd take a few second to convert that surly, or could it happen realtime?

These tests are using the computers on board soundcard rather than running through the interface, as I didnt want that to be a variable. Its an interesting point you make there because (I'm yet to test this) I think using VLC/WMP and playback via USB interface, the sound appears to be flat... still have to test this.

If there's a sample rate mismatch between the file and the system clock then I think the audio gets resampled in real time during playback, either by the audio application or possibly by the Windows soundcard driver. At any rate, it might be worth checking whether the sound changes according to whether the soundcard and the file are at the same sample rate.

[Hugh Robjohns]

Just because I was intrigued with the OP's complaint, I generated a 20Hz to 20kHz log-sweep tone file at -20dBFS, (sampled at 44.1k) in Adobe Audition, and then played it back from Adobe Audition, SADiE, Reaper, VLC, Windows Media and iTunes.

All replay software was configured to route via the AES outputs of one of my two RME AIO cards (via ASIO), and I recorded through my Crookwood mastering console back into the AES inputs of a second AIO card into SADiE.

Every track comes back completely flat. The meters sit at -20dBFS throughout and don't budge.

And once I'd aligned the starts whatever pair of recorded tracks I picked, if I inverted the polarity of one they nulled completely.

So I've no idea why the OP's system sounds different when replaying from different sources, but my system sounds -- and is -- identical, exactly as I suspected, so there is no intrinsic problem with WMP or VLC as players. You'll need to look elsewhere for the source of your apparent problem.

[Lophophora]

Thanks for testing, I was really intrigued as well! Although the frequency charts made it pretty clear that there was something going on somewhere anyway.

[ryan_570]

Just because I was intrigued with the OP's complaint, I generated a 20Hz to 20kHz log-sweep tone file at -20dBFS, (sampled at 44.1k) in Adobe Audition, and then played it back from Adobe Audition, SADiE, Reaper, VLC, Windows Media and iTunes.

All replay software was configured to route via the AES outputs of one of my two RME AIO cards (via ASIO), and I recorded through my Crookwood mastering console back into the AES inputs of a second AIO card into SADiE.

Every track comes back completely flat. The meters sit at -20dBFS throughout and don't budge.

And once I'd aligned the starts whatever pair of recorded tracks I picked, if I inverted the polarity of one they nulled completely.

So I've no idea why the OP's system sounds different when replaying from different sources, but my system sounds -- and is -- identical, exactly as I suspected, so there is no intrinsic problem with WMP or VLC as players. You'll need to look elsewhere for the source of your apparent problem.

This is the difference you are if you are using RME AIO cards to run your DAW and VLC software through this of course will give a flat response, however I am using a 'consumer grade, stock shipped, dell standard on board soundcard'.
If I play VLC through my Interface for example, it will improves quality.
I'm trying to listen to VLC/WMP how the average person would.

[ryan_570]

These tests are using the computers on board soundcard

I'm a little confused exactly how you are running these tests but are you listening/recording the files from different audio interfaces (external/internal) ?

The devil is usually in the details

Two laptops one to play all the files, the other to monitor. The output from latptop one is straight from the headphone out to an interface, (set at )db and never gets touched during the test). This interface runs into a second computer, on which SPAN is then used to monitor the curve.
Hope that makes sense

[RichardT]

Just because I was intrigued with the OP's complaint, I generated a 20Hz to 20kHz log-sweep tone file at -20dBFS, (sampled at 44.1k) in Adobe Audition, and then played it back from Adobe Audition, SADiE, Reaper, VLC, Windows Media and iTunes.

All replay software was configured to route via the AES outputs of one of my two RME AIO cards (via ASIO), and I recorded through my Crookwood mastering console back into the AES inputs of a second AIO card into SADiE.

Every track comes back completely flat. The meters sit at -20dBFS throughout and don't budge.

And once I'd aligned the starts whatever pair of recorded tracks I picked, if I inverted the polarity of one they nulled completely.

So I've no idea why the OP's system sounds different when replaying from different sources, but my system sounds -- and is -- identical, exactly as I suspected, so there is no intrinsic problem with WMP or VLC as players. You'll need to look elsewhere for the source of your apparent problem.

This is the difference you are if you are using RME AIO cards to run your DAW and VLC software through this of course will give a flat response, however I am using a 'consumer grade, stock shipped, dell standard on board soundcard'.
If I play VLC through my Interface for example, it will improves quality.
I'm trying to listen to VLC/WMP how the average person would.

I’d be astonished if any inbuilt audio processing generated errors on this scale.

[Hugh Robjohns]

... if you are using RME AIO cards to run your DAW and VLC software through this of course will give a flat response, however I am using a 'consumer grade, stock shipped, dell standard on board soundcard'.
If I play VLC through my Interface for example, it will improves quality.
I'm trying to listen to VLC/WMP how the average person would.

Silly me... I thought when you said, "Cubase is beautifully flat, however VLC and WMP are seriously… just ugly" you were complaining that VLC and WMP were defective in some way. I proved that wasn't the case.

It now seems you're really complaining about a DELL consumer soundcard...

Well, yeah... all bets are off. Who knows what enhancement processing is ever being applied in consumer-world? But you can't really anticipate or compensate for it because there are way too many unknowns.

The important point, in the context of your thread title, is that the there are no inherent response anomalies in the software players themselves.

H

[James Perrett]

I’d be astonished if any inbuilt audio processing generated errors on this scale.

You'd be surprised at how much some manufacturers can mess up the sound. It can sometimes take a deep dive into the registry to fully disable the software enhancements while the worst laptop sound I ever encountered was from a big name manufacturer with a well known audio manufacturer's name emblazoned on the front.

I've not had a problem with Dell although the only ones I've seriously listened to are the Precision mobile workstation range so the cheaper ones may well suffer from the same problems as other manufacturers' devices do.

[Drew Stephenson]

I remember having a bit of a panic a couple (ahem) of years back when listening to a track of mine on my work laptop. It sounded terrible! Overblown bass, massively compressed, really aggressive in the high-mids...
Turns out the laptop (a Dell) had a set of proprietary audio controls (apart from the usual Windows stuff) that defaulted to being optimised for voice calls. But it took a while to find that particular set of controls.

[merlyn]

Room EQ Wizard has a 'calibrate soundcard' option.

https://www.roomeqwizard.com/

You connect the input to the output, and click a button. You can then see the frequency plot of the soundcard. This would provide some more data on whether it's some Windows software processing, or a limitation of the hardware.

[Martin Walker]

Good idea!

The Rightmark Audio Analyser (RMAA) peforms a simlar function via a loopback cable:

https://audio.rightmark.org/

Martin

[ryan_570]

Room EQ Wizard has a 'calibrate soundcard' option.

https://www.roomeqwizard.com/

You connect the input to the output, and click a button. You can then see the frequency plot of the soundcard. This would provide some more data on whether it's some Windows software processing, or a limitation of the hardware.

That's a great idea, and thank you, unfortunately for me Line in/ out is one single input! Madness really.

[ryan_570]

Good idea!

The Rightmark Audio Analyser (RMAA) peforms a simlar function via a loopback cable:

https://audio.rightmark.org/

Martin

This looks awesome, thank you. Sadly my Line in/out is just a single input.

[Sam Spoons]

Your problem may well be solved by you buying and using an inexpensive USB audio interface, Behringer make some extremely cheap ones and a simple two channel one is well under £20. If you intend to do anything more than playback of pre-recorded tracks something like this https://www.thomann.de/gb/behringer_u_p ... c202hd.htm would be a better choice.