Fairlight-CMI

Peter,

I completely agree with Tobias on this topic.

It has much been talked about and I am sure you are well aware of the technical details 
+ allthough I run the risk of repeating what Tobias has already stated,
I feel, In order to make the point (to others involved in this discussion), I need to 
somewhat start from the beginning.
So here goes again:


Sound Quality...

The reason these old sampling workstations (Fairlight, Synclavier, and in partially PPG, E-
mu II, III (not IIIX) still sound so good clearly and audibly lies in their basic design.

They all use a dedicated DA-converter for each voice played back.
The signal from each DAC is then led through a VCA to impose volume envelopes or real-
time effects and summed with other voices in the analog domain.

In a system like this, a 16Bit signal stays 16bit, even if it is played back at very low 
volumes.
Surely the VCA would add some analog noise to the signal,
but it is essentially the same technology as used in big SSL or Neve desks
and people seem to live with the results from these pretty well.

As Tobias has allready stated, the digital readout rate of the samples on these old 
samplers is used to vary playback pitch much like on a vari-pitch analog tape. 
Again no basic digital degradation of the sound. 
Every single sample gets played back exactly as it was recorded, only in sped-up or 
slowed down succession. 
If you listen to a slowed down analog tape playback It sounds is amazing, smooth, 
organic, powerful. 
Compared to this ANY algorithmic pitch change sounds rediculous.

In a standard hardware sampler like a Akai S1000 (same goes for all existing Software 
samplers) the playback of a recorded sound is manipulated in the digital domain.

If you play a 3-voice chord,  the samples are merged in realtime (recalculated) and played 
back as one new sound file.
The same goes for any change of pitch or volume. 
These calculations impose audible errors onto the sounds. 
If you take into acount that these errors add up, then by the time you run a complete 
playback from your standard sampler in realtime, what you hear is only a vague 
aproximation of what was once recorded. 
And quite different from what you would hear if all these recordings were played back by 
lets say individualy pitchable high-quality analog reel-to-reel recorders.

The difference between these two technical approaches (the hardware- and the software-
aproach) to sampling are very very very audible indeed.


On a Synclavier I oftentimes find myself to prefer a short single sample to using a big 
multi sampled instrument.
This is because the sampler's transposition sounds so good. It is not so much a matter of 
realism but of musicality.
If you look at the excellent CMI II sound library, some of the best sounds consist of just 
two short samples, mapped across the whole keyboard.
These sounds have stood the test of time far better then version 15 of the 50Gig 32 
Samples per key Gigapianowhatever.


Don't get me wrong here, I am not tying to make a point for 150kb size samples in 
general.
I just try to describe what the essence of sound quality on an instrument like the CMI-IV 
sould be measured against.

It would be amazing to be able to combine the sound quality of the old high-end systems 
as described above, with today's huge multisamples. 
In my opinion the key to creating a classsic today would be:
Lots of memory and fast/easy access to huge sound librarys, played back with lots of 
voices in the best possible quality.
(User interface and form factor are a different topic)

If it is possible to achieve this using todays technoligies GREAT! 
I do not care if it is done by hardware or software, in the analog or digital domain...
...as long as it stands up against the yardstick sound, set by these old Fairlights and 
Synclaviers some 20years ago.

Oh, yes, allmost forgot: I wouldn't mind If the CMI-IV came to sound even better than the 
aforementioned... 
...and yeah, right Tobias: Some filters would be cool!


Best regards,
Fabian




--- In Fairlight-CMI@yahoogroups.com, Tobias Enhus <tobias@...> wrote:
>
> Peter,
> The appeal to me in a system like say a CMI-IV, are
> not the qualities of an experimental platform like a
> Kyma system (as someone suggested).
> The appeal to me is much more basic. I rather want a
> simple, the audio sampler version of a really really
> good sushi chefs knife. Not a swizz army knife. The
> best Sushi knife there is if possible. Not totally
> unlike the original CMI III software/ hardware design,
> but with a few well chosen modern upgrades.
> 
> - My main interest in the CMI system is as a sampler
> and music instrument.
> It's a digital instrument with a strong analog feel.
> It is the sound u get when you pitch something down 2
> octaves that rings well with me. The very organic
> quality of the sound. 
> One major thing that both the Synclavier and the CMI
> III lack, are filters. This would be a very nice new
> feature for the next gen CMI. 
> 
> - It would absolutely be great to get a few disk
> recording tracks. But I don't think competing with
> current recording systems like Pro Tools, is the idea
> here. Rather go for quality and simplicity. 8 tracks.
> 16 tracks or perhaps even 24. But no need for dub
> stage size sessions here. Again, simplicity and
> superior sound quality. Features like vari speed,
> EQ/filters and dynamics would be very nice.
> 
> - User interface should be all about functionality. No
> color unless color is needed and has a function etc.
> Keep commands to actual buttons and stay away from too
> much mouse movement. Speed and tactile interaction is
> the key.
> 
> - If a custom chassi is better for ease, sound.,
> consistency, sex appeal, then great! Custom chassi it
> is! Don't hold back on a Rolls Royce. As long as it
> doesn't become 'bling". No blue LEDs for no reason
> etc. 
> 
> - Dedicated RAM based system, to ensure rock solid
> stability. And why not loads of it. I could easily see
> myself with a 30 gig RAM system, if possible.
> 
> Now, what is sound quality really? Why do I use Pro
> Tools for certain things and Synclav for other? I
> would perhaps describe it as organic personality. 
> I use my synclav in very un orthodox ways (by general
> digital audio standards). Ex I normalize a sound, then
> continue to over gain it to get distortion from the
> cards. Sounds amazing! Layer bunches of samples on
> top, all with different playback speeds. Sounds huge!
> Sample something in at super low level, then normalize
> the file. Best noise ever! Basically I bend, kick and
> torture those synclav soundcards backwards and
> forwards, just because they give interesting results.
> Unlike if you try to do the same in say Pro Tools (try
> to clip a file in PT and see how good that sounds).
> 
> This is why I wonder about the basic hardware design
> of a CMI-IV. I'm sure it is possible to port the
> software, and perhaps mimic some aspects of the
> original hardware. But, it is important to remember
> that  this is more than just nostalgia.  If I want
> nostalgia, then I would buy the original no matter
> what. What I am interested in is to improve upon an
> already successful design. Make something better. Not
> just more convenient, but sonically better. 
> If you look at the market right now, there simply
> aren't any high end pro samplers out there. Sure you
> have scores of sampler software, but nothing still
> beats a well calibrated Synclavier or CMI III, even to
> this day. Let's change that!
> 
> How would you mimic the hardware architecture of the
> voice cards? All running in tandem, but at different
> clock speeds depending on what note you press on the
> keyboard? Just the virtue of having multiple cards,
> has the same effect as stacking up multiple analog
> mono synths, a very pleasant and rich sound.
> 
> The sound you get when slowing down a sound to less
> than a quarter of its original speed. A very pleasant
> sound on a tape machine or a Synclavier, but usually
> not interesting nor good sounding on a software
> sampler.
> Do you think it's possible to solve this in software?
> If you think it is, and if you think more DSP would
> solve it, then don't hold back. This is the weakest
> link of any software sampler for sure. I would
> sacrifice a lot of extra features (read DSP) just to
> get to the very best vari speed sound possible. If it
> could even match that aspect of the original CMI III
> or the Synclav, I would be absolutely thrilled! 
> 
> I don't know how it worked on the CMI III, but on the
> Synclav you have analog VCA's that also brings a
> certain analog quality. It also effectively made the
> output of the sound always 16 bit "analog floating
> point?", as the sound always retained full resolution
> regardless of level. 
> Now VCA's are much easier to simulate in software. 
> Although easier to generate, they should not be
> overlooked as they bring a very important part of the
> sound. Perhaps analog modeling?
> 
> 
> Cheers!
> Tobias
>