Page 1 of 2

24/96 Recording - seem to get some distortion

PostPosted: Tue May 04, 2010 7:35 pm
by ajleach
Hello Derek (and others),

I bought Wave Corrector a few years ago and started digitising some of my vinyl. But my setup wasnt brilliant.
Recently I upgraded to a M-Audio 24/96 sound card and put a decent amp and speakers on the output. The sound is really pretty good now.

Last week I had some records cleaned with a Keith Monks cleaning machine and recorded one using 24bit 96 Khz. The sound was really clear and clean. There were few large clicks but some faint crackle - which I could probably live with, as it really is faint.
But when I tried declicking, I seemed to get a "muddier" sound than the uncorrected version. I tried again but reduced the threshold to 1 - it was a bit better but still distorted.
The record was in reasonably good condition, but WaveCor seemed to turn every little bit of crackle into a click. It found hundreds of thousands of clicks.

Is there something I've missed? Should I change settings in the Advanced section, to compensate for the higher sampling rate??

I dont intend to burn CDs, but store the sound files digitally (probably in FLAC format) and play them using a streaming system (or chop them down to MP3s for travelling.)

I'g appreciate any ideas.

PostPosted: Wed May 05, 2010 1:47 am
by Glenn
Hi AJ, welcome to the forum. I had a similar problem with my 24/96 when I plugged in a device that used a switch mode power supply in the middle of a recording. I was using an external ADC at the time and I believe the noise was picked up by the coax cable to the s/pdif in on the MAudio.

In my case using a s/pdif cable with better sheilding solved the problem, but what should work in your instance is to raise the ASIO/WDM buffer in the Delta control panel to between 384 - 512 samples. It worked for me until I could replace the cable.

BTW, too many corrections will affect the clarity of audio just as you described. For records that are in decent condition I use the level 1 setting; minimum size=7; noise discrimination=0, and the AGC=3. With these settings the number of corrections is limited to 500~1000 in most cases. With this small number, I'll scroll through the correction list one-by-one to move through the file and make adjustments. In most cases, all that remains is clean music, with a very fine, barely audible crackle that's easily ignored.


PostPosted: Sun Nov 28, 2010 8:01 pm
by ajleach
For various reasons (mainly work :( ) I havent had the chance to use Wave Corrector for a few months.
So today I tried a simple experiment, I recorded the same piece of music twice - once using 16bit 44.1 KHz, then again using 24/96.
It was a fairly good quality classical recording but only about 1min 30 secs long.
I scanned for clicks using the settings recommended by Glenn.
For the 16 bit recording it found 12 clicks.
On the 24 96 one it found over 2000!
Do I have a setting wrong??

Any ideas would be great!
Maybe I should just decide that 16/44.1 is really good enough - especially for my (fading) hearing!


PostPosted: Mon Nov 29, 2010 12:02 am
by Glenn
Hmmm, 2000 clicks over a 90 second sample is the kind of noise I was getting with the s/pdif cable (it still happens on rare occasions, but I have the ASIO buffer set to the minimum allowable for playback reasons ).

What size is your ASIO buffer?

Perhaps there is an issue with the MAudio and Wavecorrector at 24/96. I do experience hiccups in the audition playback where there will be no sound present; I must hit the escape key to cancel playback and re-initiate. I assumed this was due to the external ADC as the MAudio's clock is synched externally, but it could just as easily be the MAudio card....and again it could be the small buffer size.

I'll raise it before my next session to see if this is the case.

PostPosted: Wed Dec 01, 2010 4:10 pm
by Derek
Andrew can you send me the recording so I can investigate this. It may just be that there is some high frequency noise on the original but I wouldn't expect such a huge discrepancy.

I'll pm you with the information on how to upload the file.

PostPosted: Wed Dec 01, 2010 10:22 pm
by Glenn
Derek, if by "original" you mean the record, I doubt this is the case. In my instance, the clicks are short spikes only 2 or 3 samples wide superimposed on the waveform. Wavcor catches them all, but of course the result is unlistenable.

Oddly, the distortion caused by the spikes is inaudible during record monitoring, and appears only on the written file. I've never attempted a duplicate recording at 16/44 so Andrew's observations are a bit of a revelation to me.

Please keep me apprised of your progress with Andrew. If I come across a sample myself in the near future I'll submit it to you if you haven't found any issue before then.


PostPosted: Wed Dec 01, 2010 10:41 pm
by citguy
I assume everyone has carefully checked input/output levels on their cards/mixers. On my M-Audio 2496, I get some peak noise if I approach 0 decibels even if WC is recording in a normal range. Just a thought.

PostPosted: Wed Dec 01, 2010 11:57 pm
by Glenn
Good to hear from you again, Stan. This isn't clipping distortion, but comes across as a continuous hash on playback - at any/all dynamic levels. Rather like listening to a record with a broken needle. In my case it's quite rare, but no less irritating because it can't be heard until playback commences or the click count goes through the roof. The time spent is wasted and the recording has to be started over.

I take it you haven't noticed this with your M-Audio whilst recording at 24/96? What is your ASIO buffer set at?


PostPosted: Thu Dec 02, 2010 12:07 am
by citguy
Hi Glenn.
ASIO buffer at 256.

PostPosted: Thu Dec 02, 2010 12:23 am
by Glenn
Thanks, Stan. I have mine at 64 (the lowest possible for my card) for playback in cPlay because lowest is recommended to reduce software jitter in pc based audio playback. I've never gotten around to trying a higher setting for recording because I keep forgetting to open the control panel first. :roll:


PostPosted: Mon Jan 03, 2011 9:39 am
by Derek
Unfortunately, I'm not getting very far with this problem. Glenn has sent me a file that illustrates it quite nicely and I suspect unfortunately that it is a problem with the m-audio driver. Glenn has said that the problem never occurs when recording with Audacitiy. However this does not completely exonerate the driver because it provides several different interfaces and Wave Corrector and Audacity use different ones.

I do note on the m-audio site that there are a large number of drivers, and different ones for each operating system service pack. So it's possible if you've updated your operating system, that you also need too update your m-audio driver. Can you just check that you have the correct driver for your operating system.

In the meantime, I'll investigate the problem further.

PostPosted: Mon Jan 03, 2011 1:46 pm
by Glenn
Hi Derek, thanks for looking into this.

I updated to the latest driver based on your recommendation some time ago but couldn't recall what the issue was, so I've since rolled back to the old one that MAudio recommends for XP SP2 (my current OS).

After rolling back the driver I can see the issue was the 24 bit checkbox in the WavCor recorder is not enabled. A quick test yields a recording with 16 bit depth only, even though the Mytek ADC (via spdif in) is set to 24.

I've noticed a few differences between the way Audacity and Wavcor records. Wavcor seems to write the data directly whereas Audacity writes a temp file at 32 bits, with the final bit depth chosen by the user when the file is exported. I much prefer the method employed by WavCor as it saves time and drive wear.

As an aside, I noticed that the monitor stream through Audacity sounds identical to the final recording, whereas in WavCor it sounds different....not as clear, as though an inferior soundcard was being used. This does not seem to affect recording quality however. In fact, the results with WavCor seem to be easier on the ears, though I don't have enough samples through Audacity yet to support this.

Thanks, Derek.

PS, here is the original thread: ... highlight=

PostPosted: Tue Jan 04, 2011 9:13 am
by Derek
Audacity uses a higher level interface with the soundcard, ie one where the operating system itself does some processing before the sound data arrives at the program. Wave Corrector uses a low level interface that reads data directly from the soundcard. This partly explains the different approach to recording.

Also, in Wave Corrector the monitor stream is 16-bit which probably explains the inferior sound quality when monitoring.

PostPosted: Wed Jan 05, 2011 1:52 am
by Glenn
That's very interesting. Glad to know these old ears still work. :)

Since the distortion is not audible during recording I had assumed the monitor bypasses any processing done in WavCor, similar to how a 2 head tape deck simply routes the input back into the source monitor loop. Why was 16 bits chosen, or was this simply a matter of convenience?

In the meantime I've gone back to recording in WC, but I'm trying the Vista driver to see if it makes any difference. I'll keep you posted.


PostPosted: Wed Jan 05, 2011 9:29 am
by Derek
It is rather puzzling that the distortion is not audible during recording. The 24-bit samples are truncated to 16-bits and passed to Wave Corrector for internal processing (producing the monitor stream and generating the graphical display). So if the problem exists on the 24-bit stream it ought to be present when it's truncated to 16-bits. Perhaps the is a clue to where the problem lies.

I assume you are certain that the monitor stream is definitely clean?

Regarding the choice of 16 bits. When Wave Corrector was originally developed, 24-bit soundcards were quite rare and most computers probably wouldn't have been fast enough to do all the conversions necessary to generate the Wave Corrector displays.

24-bit processing is actually quite complicated because computer arithmetic uses powers of 2. Numbers are represented as 8 bits, 16 bits, 32 bits or 64 bits. So when you're presented with a 24-bit stream, you have to pad it out to 32 bits, one sample at a time; do your processing, and then truncate it back to 24 bits for saving. This is all rather time consuming.