Morning Star,
I've been thinking about what you said. I thought about it again.
It is possible that maybe the sound quality would be better through a a really high quality DI. I doubt it, but to be perfectly honest, the only way to be sure would be to try it. And I don't currently have a DI nor am I about to buy one.
For comparison:
I happen to have an Alesis QS synth that I also connect directly to my Audiophile 2496 soundcard through the exact same cable as I used for the StreetBoxx samples. Using the same software and soundcard settings, the noise floor of the soundcard is very very low even with the synth plugged in an outputs turned up anywhere from 50% to all the way. Very easily I can get very pristine recordings of the synth without having to go through a DI or mic pre or anything like that.
In contrast, the StreetBoxx isn't exactly noisey or have humm or buzz or any kind of ground loop issue. However when zooming in on the recorded waveform, it's evident that there is some extra low (under 20Hz) frequency content in there which is modulating the audible waveform giving it some modulating DC bias type of thing.
It's not a lot in most cases, it's just something I noticed when editing.
I'm not even sure if it's all that bad. It might actually add something positive to the recording. Maybe it's from the turntable of the original samples in some cases.
I almost decided to filter it out, using an extremely steep -144dB FFT filter in Cool Edit 2000, but it seemed like it was adding a bit of subtle bass character and I just couldn't be sure of how much of it was from the original source sample that Zoom/Beat Kangs used or whatnot.
The hiss inherent in some of the original samples was also pretty easy to detect because it was clearly in some samples but not at all in others. Since I didn't change my cable or soundcard hookups when going from one sample to another, then this clearly means that some of the original source samples had hiss in them. If it was just my soundcard or something of mine, then pretty much all of the samples would have hiss.
Again, sometimes the hiss is part of the sound character. So I decided to leave it in most of the time. The few times I did filter out hiss, I tried to make the (customizeable CoolEdit 2000 FFT) filter as subtle as possible.
The classic hiss example was in the "00 SINEWAV" bass patch. It clearly shows exhibits hiss on the spectrogram and you can hear it when you compare it to a real sinewave which has no harmonics by definition. But I left the StreetBoxx's sinewave version alone instead of filtering it because it actually has a certain (digital aliasing?) flavour to it somehow. The pure computer-generated perfect sinewave is clear and noiseless, but somehow doesn't capture the same feeling. This is hard to describe, but this kind of issue was on my mind a lot while working on these samples.
I hope this clarifies what I did and didn't do better. But when it comes down to it, I don't have a DI so I couldn't A/B test with and without a DI to see which recording is better.
On a lighter note:
I opted not to multisample the basses, but I include them. At first I multisampled a few of them, but then I noticed that most of the original bass sounds were not multisampled either. I could here the "munchkinization" of the StreetBoxx hardware sounds getting shorter and shorter going up the musical scale. Most of the bass sounds that I thought were looping were actually just long one-shot recordings of a sample looping and eventually fading out after about 11 seconds.
To be sure, I loaded up the samples of basses I made (using the middle C note held) into HighLife VST sampler and A/B compared them with the StreetBoxx hardware's onboard samples. I played up and down the trigger keyboards and trigger pads comparing the sounds. In most cases the difference was minimal.
The main adjustment that needed to be made in HighLife was to set the ADSR amplitude release to 74ms to match the ADSR amplitude release of the StreetBoxx's basses.
So how identical are the samples?:
They are similar enough that you could probably redo your StreetBoxx beats using these samples loaded up into a sampler and get a mix that sounds the same musically. Bit it wouldn't be bit for bit the same. For one thing, there is not a way for me to sample the StreetBoxx's reverbs or other effects--not even with convolution IR's--there's no way to load up an IR to trigger the effects to sample.
And of course I sampled most everything in mono at 44.1kHz. A higher sampling rate and in stereo would pick up stuff better, but most of the source samples are already basically monaural anyways. They aren't perfectly mono if you look at a Lissajous phase meter of them, but they are close enough to mono that the comparison is barely noticeable.
I could've sampled at 48kHz or 96kHz, but I didn't feel that the cost in storage space was worth it.
On the positive side however, as samples these sounds can be better processed and sculpted as individual tracks if set up as such. You could make the sounds dirtier or cleaner with additional processing, you could make them fatter or thinner.
Conclusion:
After all this talk and work, I've realised that sampling and multisampling are a lot more complicated than it might seem at first glance. I'm gonna be disappointed if it turns out that I can't release these samples. But I'm going to contact Zoom and SamsonTech to find out first. Sorry to have gotten your hopes up and dropped them if it doesn't work out in the end.
