DC Problems in OMI CD's?

[Python Blue]

I've been trying to work on this problem for an automated answer for a while, and was hoping EMXP updates would fix the problem.

Has anyone else experienced severe DC offset with some of the samples in the OMI CD's when converting the raw samples?

Let me give an example: I found a surprisingly long sample of string cluster effects in volume 1, a sound that was persistent in the first half of the Halloween 5 soundtrack. I'm hoping that the problem is a fault with either EMXP or CDS3Tool, but there is severe DC offset in the sample.

Normally, this could be fixed with the normalize functions in Audacity, but the problem is this: the very first point of each affected sample is centered, but after that, it changes to a DC offset, usually under center, that can be as high as -6dB! Wha this means is that correcting DC conventionally, or even applying a high pass filter with a very low cutoff, will result in a pop in the beginning of the sample, which is particularly troubling for the samples in existance that loop throughout the entire bulk of the sample.

Anyone else have this problem? If so, were you able to correct it without time-consuming manual editing of the samples? I've doing just that in the meantime, but call me a purist, especially when it comes to the smoothness of looping points.

[esynthesist]

Some factory and OMI banks indeed seem to contain samples with a DC offset.
The String Cluster banks are examples of these.

I noticed this problem years ago when I was working on the EII support in EMXP.
First I thought it was a problem with the encoding/decoding engine I'm using, but when I loaded these banks into the Emulator II and transferred their samples to the "good old" Sound Designer for EII software on Mac, the same problem occurred. I just double checked it with the samples of the String Cluster banks, and Sound Designer for EII clearly shows a DC offset in its Waveform View for these samples as well (as opposed to many other OMI samples which are normally centered around the 0db offset).

In my opinion something went wrong when creating the samples back in the eighties.
Either they were not sampled directly on the EII but rather sampled with some other gear and finally "encoded" into the companding 8 bit format of the EII: something may have gone wrong during this encoding process.
Or the offset was simply created in the sampling process itself by the recording gear (an offset level of 6dB doesn't seem like a coincidence...)
You can notice at the beginning of each sample that the sample's amplitude is quickly evolving (although not in a linear way) towards the DC offset level. After a short while this process stops and the DC offset remains stable (but "wrong").
So I guess that either for some reason the "companding encoder" used by OMI went in "overdrive" at the beginning of these samples, or that this DC offset was created by the recording gear used by OMI.
 
I was considering to add a "compensating algorithm" in the decoder of EMXP which would be applied upon detecting a DC offset throughout the majority of the sample. But the main problem is that it's hard for the decoder to notice/detect this initial signal evolution towards an offset (the signal may have this shape intentionally) and especially to detect at what specific point in the audio data the "offset shaping" stops and no further compensation is required.
I'm not saying though that it's impossible to implement this, maybe I will re-consider it.
So far I didn't further investigate it because I noticed that Sound Designer suffered from the same problem, so I consider it to be a problem of the sample, not of the software ;-)

Anyway, this explanation doesn't help you in finding a quick solution to solve the problem of course.
But I'm afraid the only solution is currently to correct the samples with manual editing.

///E-Synthesist

[Wolfram]

I noted this problem too. When I made my first experience the my EMAX with Sound Designer and Alchemy.
...25 years ago ;-)
I remember, I tried to remove the DC offset withe no luck.