Beta Test: Earful -- a hearing test App

[pkane]

I have a i5 6600K and all the four physical cores are used when it freezes and has glitches.

I have to try again with other outputs, it happens with the standard Windows DS for sure and I am using EqualizerAPO for both headphones/IEMs and desktop speakers. But if I recall I changed output too and I think I had the same issue in WASAPI (exclusive and shared) but I didn't test properly and maybe it is just DS.

I just tried ASIO, WASAPI and DS, the CPU utilization stays around 10% even if I continuously move the slider up and and down. If you can see which process is using up all the CPU in task manager, that will help me. If all four cores are used, then it's not Earful. It uses up to two threads at any given time, so can't take up 4 cores

Here's what I get while continuously adjusting volume slider playing into a DS driver

 

[zermak]

I just tried ASIO, WASAPI and DS, the CPU utilization stays around 10% even if I continuously move the slider up and and down. If you can see which process is using up all the CPU in task manager, that will help me. If all four cores are used, then it's not Earful. It uses up to two threads at any given time, so can't take up 4 cores

Here's what I get while continuously adjusting volume slider playing into a DS driver
View attachment 74842

It is Earful sadly (edit) with WASAPI shared. In DS it looks like working alright.

 

[pkane]

It is Earful sadly.

Hmm. That's strange, as I don't see the same behavior. I'll try to reproduce with different hardware. Thanks for reporting and testing!

If glitching isn't completely awful, I assume you can still use it, correct? After you finish adjusting the volume, does the sound become normal or does it continue to glitch?

 

[pkane]

It is Earful sadly (edit) with WASAPI shared. In DS it looks like working alright.

Interesting. Does EqualizerAPO work with DS driver, can you tell?

EDIT: I just installed EqualizerAPO, added convolution, and a few other types of filters, reduced my hardware to a 1-core CPU... And still I get only about 30% CPU utilization when moving the volume slider around. Don't know what else to try to reproduce the problem. Any thoughts?

 

[zermak]

Interesting. Does EqualizerAPO work with DS driver, can you tell?

It should and it is not the issue here. I have problems with WASAPI shared (which should use EqualizerAPO too).

EDIT: I just installed EqualizerAPO, added convolution, and a few other types of filters, reduced my hardware to a 1-core CPU... And still I get only about 30% CPU utilization when moving the volume slider around. Don't know what else to try to reproduce the problem. Any thoughts?

Have you tried using WASAPI shared? That's when I have the high CPU usage issues. I am gonna try ASIO and WASAPI exclusive to see what happens.

EDIT: The most demanding is WASAPI (both shared and exclusive) and it causes glitches if I just start moving slides randomly and it then get stuck at 100% CPU usage and I have to kill the process. With ASIO it seems fine but ASIO4ALL doesn't work outputting a memory error (telling you just for debugging).

 

[ernestcarl]

I wasn't going to share mine, but what the heck...

A thunderstorm just started so I had to stop... preliminary results:

Between 20-30Hz is confusing and likely wrong as I was having trouble differentiating between what is a physical sensation of pressure vs actual audible tones.

Not possible to isolate sensitivities between L&R ears as I'm using speakers + sub.

Distance is ~0.9m with speakers calibrated to 79dB(C)S SPL with -20dBFS pink noise signal.

That unusual ~10dB dip around 450Hz corresponds with a deep null in my L+R speakers' as you can see in the MMM RTA response (unsmoothed):

I'll retry when it's much quieter.

*Maybe I'll get a different result using the other signal options.

 

[pkane]

It should and it is not the issue here. I have problems with WASAPI shared (which should use EqualizerAPO too).

Have you tried using WASAPI shared? That's when I have the high CPU usage issues. I am gonna try ASIO and WASAPI exclusive to see what happens.

EDIT: The most demanding is WASAPI (both shared and exclusive) and it causes glitches if I just start moving slides randomly and it then get stuck at 100% CPU usage and I have to kill the process. With ASIO it seems fine but ASIO4ALL doesn't work outputting a memory error (telling you just for debugging).

I’m guessing it has to do with the actual driver and the device. I’ve tried WASAPI shared and exclusive with three different devices and had no issues. ASIO works with the right device driver, but I’ll try ASIO4ALL, as I’ve never used it.

 

[zermak]

I used my cheap laptop for the testing and I had less problems there than on my tower

Anyway here is the graph:

I used WASAPI exclusive mode on my laptop and at 192kHz (I loaded the file on my main PC for ease, that's why you read 48kHz on the screenshot) with my KZ ZS10 Pro IEMs without any EQ correction. Forget about the last two points cause I could't go higher in SPL (and I should had to adjust my main volume and all the curve to lower values if I could ever hear such tones ahah).
I used a 1dB step cause with 0,5 I was probably imagining sounds and I needed a bigger threashold

EDIT: I used the same USB DAC on my laptop (and used WASAPI exclusive without a glitch) so I guess it's something else messing it all up on my main PC (and yes I have EqualizerAPO on my laptop too but indeed it was bypassed cause of WASAPI exclusive). The laptop runs on a Celeron N4100

 

[zermak]

I was editing when you @pkane liked my post

 

[pkane]

I used my cheap laptop for the testing and I had less problems there than on my tower

Anyway here is the graph:
View attachment 74885
I used WASAPI exclusive mode on my laptop and at 192kHz (I loaded the file on my main PC for ease, that's why you read 48kHz on the screenshot) with my KZ ZS10 Pro IEMs without any EQ correction. Forget about the last two points cause I could't go higher in SPL (and I should had to adjust my main volume and all the curve to lower values if I could ever hear such tones ahah).
I used a 1dB step cause with 0,5 I was probably imagining sounds and I needed a bigger threashold

Great! It would be interesting to see this in relation to the ISO226 curve. Just click on Load button, select the ISO file, and load as the second data set. It’ll make it easier to compare curves from different people and devices.

 

[Hugo9000]

Well, my sad old computer seems to hate it. CPU near 100%, pulsing distorted sounds (quiet, possibly at the level appropriate for what the test tones are supposed to be) that are identical whether I have selected tone, warble, or white noise.

I generally use DS, but the same happened when I tried WASAPI.

Windows 10, i3-4130 3.4 GHz (LOL, it was cheap and it's been reliable, at least.)

 

[pkane]

Well, my sad old computer seems to hate it. CPU near 100%, pulsing distorted sounds (quiet, possibly at the level appropriate for what the test tones are supposed to be) that are identical whether I have selected tone, warble, or white noise.

I generally use DS, but the same happened when I tried WASAPI.

Windows 10, i3-4130 3.4 GHz (LOL, it was cheap and it's been reliable, at least.)

Interesting. I do have a really old laptop I can try that should still work. Let me see if I can find where the bottleneck is.

 

[zermak]

Here it is. Hard to come to a conclusion but tomorrow I'll try with my K712 Pro and see what it goes to make some statements.
I also have an UMIK-1 and I could easily (?) measure the IEMs SPL for sure (just stucking it on the hole of the memory foam tips) if you give me a useful way to measure it (or I'll probably just use the ~1kHz tone).
And little suggestion: if you can add a easy way to save a screenshot of the graph, maybe with WxH default dimensions and to embed both or just one of the cruves and maybe a description for headphones/speakers used and what distance, would be great (I don't mind doing it myself but since you are developing it I am speaking my mind).

And again it's weird that on my cheap and slow laptop I had no issues and on my tower with a more powerful CPU it is struggling. I am open on testing on my main PC to help you debug it if you need some feedback.

 

[pkane]

View attachment 74886
Here it is. Hard to come to a conclusion but tomorrow I'll try with my K712 Pro and see what it goes to make some statements.
I also have an UMIK-1 and I could easily (?) measure the IEMs SPL for sure (just stucking it on the hole of the memory foam tips) if you give me a useful way to measure it (or I'll probably just use the ~1kHz tone).
And little suggestion: if you can add a easy way to save a screenshot of the graph, maybe with WxH default dimensions and to embed both or just one of the cruves and maybe a description for headphones/speakers used and what distance, would be great (I don't mind doing it myself but since you are developing it I am speaking my mind).

And again it's weird that on my cheap and slow laptop I had no issues and on my tower with a more powerful CPU it is struggling. I am open on testing on my main PC to help you debug it if you need some feedback.

If you download the latest version on the website, it now has SPL calibration enabled. You'll need to set the desired output level in dBFS, and then measure the SPL level actually produced and enter it into the calibration window. You can then view the results in dBFS or dBSPL, and switch between them at will:

 

[zermak]

If you download the latest version on the website, it now has SPL calibration enabled. You'll need to set the desired output level in dBFS, and then measure the SPL level actually produced and enter it into the calibration window. You can then view the results in dBFS or dBSPL, and switch between them at will:

View attachment 74889

I've seen that (on my laptop on which I used the latest version) so I'll give it a try tomorrow and does it matter the SPL weighting or the generated sound/tones isn't affected?

 

[pkane]

@Hugo9000 and @zermak , did you try reducing the sampling frequency of the audio device to, say 48k or 44.1k? The greater the frequency, the more load on CPU since Earful is computing all the samples in real-time and on-demand. I can try to precompute it, instead, this should reduce load significantly.

 

[pkane]

I've seen that (on my laptop on which I used the latest version) so I'll give it a try tomorrow and does it matter the SPL weighting or the generated sound/tones isn't affected?

The only thing the SPL display option changes is the display Internally, all the data remains untouched and you can even load an older measurement and view it as dBSPL or dBFS without redoing the measurements.

The calibrator is using a 1kHz warble tone, so A- or C-weighting for SPL measurement should work the same.

 

[zermak]

@Hugo9000 and @zermak , did you try reducing the sampling frequency of the audio device to, say 48k or 44.1k? The greater the frequency, the more load on CPU since Earful is computing all the samples in real-time and on-demand. I can try to precompute it, instead, this should reduce load significantly.

On my main PC I was using it at 48kHz while on my slower laptop (on which I did the test cause of being fanless) at 192kHz without any issues.

The only thing the SPL display option changes is the display Internally, all the data remains untouched and you can even load an older measurement and view it as dBSPL or dBFS without redoing the measurements.

The calibrator is using a 1kHz warble tone, so A- or C-weighting for SPL measurement should work the same.

Yes, I figured it out and I switched to SPL to see what it was about

Alright. I'll calibrate it for the IEMs and will show you the output and then I'll move on the headphones to add more data (and @solderdude could show me how to measure them with the CD hole method even if I kinda figured it out ahah).

 

[Hugo9000]

@Hugo9000 and @zermak , did you try reducing the sampling frequency of the audio device to, say 48k or 44.1k? The greater the frequency, the more load on CPU since Earful is computing all the samples in real-time and on-demand. I can try to precompute it, instead, this should reduce load significantly.

I had it on 44.1k first, then switched to 96, didn't try 48 or 192 yet. When I opened the program "as administrator," the CPU load was below 80%, and I could hear tones and not just the weird glitching sounds that I heard initially. It's probably some odd incompatibility with something. Recently, out of the blue, Windows "Spatial Sound" thing became available lol. It had always been greyed out, even though it should have been available quite a while ago based on my Windows build/update status. I have tested a few programs like OOYH recently (all trial versions, that have since expired), so perhaps if I remove those. I don't know if that could affect anything, I have normal working audio for youtube, foobar, MusicBee, Spotify, etc.

I'll try again periodically after updates to my computer, or if I make any program changes, just in case it starts working for me. Please don't expend any extra effort on my behalf, since there probably aren't many people affected by whatever this is. I almost posted yesterday as soon as you had created the thread, but I didn't want to make a big thing out of what is likely simply an issue on my end. But when I saw @zermak 's post, I thought perhaps it was a similar issue.

I'll let you know if I figure out what's going on.

 

[pkane]

Updated v1.0.4 of Earful is now available. The changes in .3 and .4 are as follows:

• Speeded up audio signal generation for slower processors
• Added dBSPL display option
• Added SPL Meter calibration
• Changed data file loader to use proper regional settings for decimal separator

Please try it if you had problems with sound on older or slower computers. Another suggestion is to make sure that all Windows sound enhancements/DSP options are turned off, as these add significant load on CPU.

Please share your results, bug reports or issues and improvement/enhancement suggestions and ideas.