Noise on Clipv1 (split from ClipV2 thread)

[egbxclip]

I have been noticing this low level high frequency whistles as well. These are quite apparent in fadeouts or very low level passages. I made some silence files and did some testing. The noises are not constant state, they tend to modulate at track starts, and are not always the same. However they are always present, to varying but noticeable degree. Usually masked by the music.

I was able to trace them to using the low frequency sections of the EQUALIZER. Noise starts immediately upon boosting either the shelving LF section or a peaking filter set at similar low frequencies (50-60Hz in my case, a typical setting to compensate for sports-type headphones).

The solution is to avoid using the EQUALIZER for bass boost. You can use the "normal" BASS control instead, this doesn't make noise. Obviously not as tweakable as the full-fledged parametric EQ, but still, better than the little whistles in the background.

Perhaps this bug can be fixed. I'm not sure where to post this for the Rockbox people.

[CrowdAds]

[saratoga]

Originally Posted by egbxclip

The solution is to avoid using the EQUALIZER for bass boost. You can use the "normal" BASS control instead, this doesn't make noise. Obviously not as tweakable as the full-fledged parametric EQ, but still, better than the little whistles in the background.

Thats not really a solution. With sensitive headphones I can easily hear the noise even with no EQ applied at all.

[egbxclip]

it IS a solution in my case, perhaps I should say a workaround rather than solution. Perhaps you have a different noise issue? I can hear very clearly, without a doubt, the artifacts in my Clip (V1) turn ON and OFF when engaging any amount of Bass boost in the EQ section. I'm an audio engineer; I know what I'm hearing.

I will do further testing but on material where I could clearly hear the artifacts and be bothered by them, it is dead quiet and much more enjoyable now.

PLEASE UNDERSTAND I AM TALKING ABOUT ARTIFACTS, discrete whistle-like tones and chirps, NOT BROADBAND NOISE/HISS, an area in which the Clip is not bad at all; with my inexpensive but high efficiency sonys, a moot point.

It would be interesting to see how the noises change in relation to the frequency of boost, I only tested with very LF settings (50-60Hz), high frequency boosts in the 10k range didn't introduce any artifacts that I can hear but perhaps it is just that the artifacts move up with boost frequency and at a certain point fall out of range and/or audibility.

I'm not a coder by any means, but I know writing decent EQ algorithms is not a trivial pursuit. Perhaps there is room for improvement in the rockbox implementation. DO not interpret this as a complaint or grudge - I'm simply amazed and grateful that I can get such good performance from such a small and inexpensive device, and FREE software.

[saratoga]

Originally Posted by egbxclip

it IS a solution in my case, perhaps I should say a workaround rather than solution. Perhaps you have a different noise issue? I can hear very clearly, without a doubt, the artifacts in my Clip (V1) turn ON and OFF when engaging any amount of Bass boost in the EQ section.

Actually, yes this is a different issue. This thread is about CPU noise on the Clipv2, which is a different player then what you have. Now that I realize you're not talking about the Clipv2, I'll put your posts in a new thread.

As far as I know, no one has ever reported a noise issue on the Clipv1. Could you post your config.cfg file as it is with the noise?

[egbxclip]

here it is.

I'm gonna try to get a good recording of these noises and investigate further. It'll have to wait 'til monday so I can use use a decent audio interface.

[egbxclip]

just realized the previous .cfg file is with the settings that don't cause the artifacts. So here is a second .cfg file, the only change is in the "graphic" eq, a LF boost. And volume all the way up. The volume changes the level of the artifacts, but interestingly, it also seems to cause changes in the artifacts themselves

The artifacts do not respond in a linear fashion to boost amount or frequency. They are also sometimes present with playback paused. They seem to get worse when skipping to another track.

[egbxclip]

I've made a recording of a silence file playing, first without any eq, then with the eq applied (same bass boost as before). On headphones you can clearly hear the hi freq tones appear and disappear when enabling/disabling the eq.

viewing the spectrum of these recordings, I can clearly see discrete spikes at 2,262Hz and related harmonics: 4524, 6786, 9048, 11310, 13572, 15837 Hz

These are clearly visible even though the laptop input noise floor is substantially above that of the Clip itself. So in reality, these spikes are higher in relation to the noise floor than what shows in the graphs.

I'm attaching gifs of the "no eq" vs "eq enagaged" plots. Unfortunately not too clear; resizing the gifs to the forum max made them blurry.

at least now I know I'm not crazy and imagining things!

[saratoga]

Originally Posted by egbxclip

just realized the previous .cfg file is with the settings that don't cause the artifacts. So here is a second .cfg file, the only change is in the "graphic" eq, a LF boost. And volume all the way up. The volume changes the level of the artifacts, but interestingly, it also seems to cause changes in the artifacts themselves

Thats a gain of 12.5 dB at 50Hz with no precut right? I would expect massive distortion from those EQ settings. When using an EQ you need to precut so that you don't have positive gain at any frequencies, otherwise you'll try to push frequencies above line level and clip them off.

[egbxclip]

Well, only once or twice I've run into distortion, with a recording that already had massive amounts of low end. A precut of 2-3dB solved the problem, even with the boost of 11-12dB. At some point this went back to 0 but I haven't had any obvious clipping. Still, a valid point and probably worth investigating in detail.

But this is besides the point: what I'm referring to is NOT clipping distortion, but artifacts. There is no way for a silence track to clip anything, no matter how much boost you apply. You would have to be boosting insane/impossible amounts to have the noise floor come up so much that it clips anywhere.

And, the artifacts happen regardless of boost setting. 0 dB boost = no artifacts. +0.5 dB boost=immediate artifacts. +12db boost=no change in the artifacts level. I haven't tried cutting, I'm curious now if it causes artifacts as well.

[egbxclip]

regards boosting low freqs to the point of clipping distortion, it should be mentioned that a relatively narrow boost such as I have on my setting is not as likely to bring up the overall level to exceed the digital ceiling as it may appear to some. I made a little experiment to illustrate this:

playing back Pink Noise, I recorded a section without any EQ, then engaged EQ with a boost of 15dB at 50Hz, a Q of 2.0. This resulted in an increase of aprox 3.4dB, both in peak and RMS levels, as observed on the meters as well as later reported by the statistics tool in Sound Forge.

As far as the artifacts go, I now have observed that higher amounts of boost in the plain "bass" control of the sound settings will make the artifacts appear as well. I can hear them once you get to +10dB or more boost. This boost, by the way, is of a much broader Q (it is a shelving eq) but still only brings up the overall level about 6dB

[saratoga]

I just ran your cfg file through the sim and my clip+ playing digital silence, and the output is white noise with a peak amplitude of 2 PCM levels. Basically just rounding error from dithering of the 32 bit filter coefficients to 16 bit PCM for playback. If you're seeing something else, then its probably not due to the EQ, or else it would appear in the PCM signal.

[egbxclip]

perhaps a Clip V1 issue?

If I can toggle the artifacts on and off with the eq, both audibly and measurably, it points to the eq as a cause, IMO.

I should mention also that the stock Sansa firmware is completely absent of these artifacts, with their eq cranked up to the max on the lowest and highest freq bands.

So I still say it is something to do with the rockbox eq. It would be nice to have a solution, the parametric eq works excellent to add what my Sony sports headphones miss at the top and bottom parts of the spectrum!

[saratoga]

Originally Posted by egbxclip

perhaps a Clip V1 issue?

If I can toggle the artifacts on and off with the eq, both audibly and measurably, it points to the eq as a cause, IMO.

I don't see why the same code running on two different clip players would produce different results. You're welcome to compile the test_codec plugin and decode to wav yourself if you want to double check though.

[egbxclip]

don't know how to do that... pointers appreciated.

I'm busy this week but next week I'll investigate and try to do just that ("compile the test_codec plugin and decode to wav").

[saratoga]

Originally Posted by egbxclip

don't know how to do that... pointers appreciated.

I'll just make you a build with test_codec.c added to the plugins list:

http://duke.edu/~mgg6/rockbox/rockbox-clipv1.7z

Long select a file, choose open with, then test_codec. From there you can decode with DSP effects, which should let you test the EQ.