Tech Question about Hardware Samplers 80/90s.

Discussions about anything related to samplers and sampling techniques.

Tech Question about Hardware Samplers 80/90s.

Postby ForeverTangent » Sat Apr 13, 2013 12:07 am

I really didn't know where else on the net I could post this question and get a reasonable answer, so I thought I would try here.

In regards to hardware samplers of the late 80's / early 90's [I am specifically thinking of Ensoniq ESP / ESP16+ Emu Emax 1 and 2, AKAI 1000, Roland S-550 / S-750] basically the 16-bit ones and less ones.

How was the sample data stored in memory, and processed in terms of numeric values?
In C/C++ terms did the companies terms did the company tend to store sample values as ints or floats?
Did the samplers process the data as integer or floating point data?
Was sample data stored as 16 bits, or a bit depth that matched the hardware?

I am just curious, because I just finished a Computer Hardware course, and I know these days we take floating point operations for granted in computers. However, I know even back then they were still a big deal, so I am just curious what the companies did with Sampler hardware to do the number crunching?

Thanks in advance if there is answer.

Stan
ForeverTangent
Newbie
Newbie
 
Posts: 1
Joined: Fri Apr 12, 2013 11:47 pm

Advertisement:

Re: Tech Question about Hardware Samplers 80/90s.

Postby Ashe37 » Sat Apr 13, 2013 2:14 am

They were stored as integers. All samplers sample and store as integers. Computers sample and store as integers.

The samplers you mentions all had CPUs with 32-bit processing, but many of them only had a 16-bit data path (Motorola 68000) and no FPU, so the sounds were processed as integers when on the CPU.

Other operations, such as filtering, were done on dedicated VLSI chips- essentially custom built DSPs. (The notable exception would be the Emax I, which had analog filters). The EPS 16+ could actually load new routines onto its DSPs from disk (Waveboy effects). These chips were 16, 20, or 24-bit depending on the model under discussion. The chips still worked with the sound as integer data, more often than not. Often, the 'filter' chip and the 'effects' chip were separate... Ensoniq recycled their seperate effects chip into effects units.

This is drastically simplified, but that's the gist of it.
Ashe37
Moderator
Moderator
 
Posts: 2681
Joined: Mon Jan 19, 2009 3:43 pm
Location: Los Angeles, CA
Real name: Unpronounceable
Gear: Ensoniq SD-1/32,SQR,VFX,ESQm
Virus Indigo, M3-61 , MS2000BR, Volca Bass
Emu XL-7, Matrix 6r
TG-33, K3m, Blofeld, Micron, Mopho, Bass Station II
Band: Eridani V

Re: Tech Question about Hardware Samplers 80/90s.

Postby piRoN » Tue Apr 16, 2013 6:15 am

I would also assume that most of the early samplers would have used a fair bit of fixed-point processing internally, for the sort of operations you'd do in floating-point nowadays.
Ensoniq EPS-m | Elby Designs ASM-2 | DSI Evolver | Yamaha TQ5 | Akai S3000XL | Akai X7000 | Novation KS-4 | SCI Prophet 2000 | Kawai K1m | Roland S-50 | Roland D-10 | Roland MKS-7 | Roland TR-606 | Yamaha RX11
User avatar
piRoN
Active Member
Active Member
 
Posts: 558
Joined: Sun Dec 19, 2004 7:04 am
Location: The crotch of The Lucky Country
Gear: A fair bit.

Re: Tech Question about Hardware Samplers 80/90s.

Postby Ashe37 » Tue Apr 16, 2013 6:31 am

piRoN wrote:I would also assume that most of the early samplers would have used a fair bit of fixed-point processing internally, for the sort of operations you'd do in floating-point nowadays.


Sure, they didnt really have FPUs in them anyway...
Ashe37
Moderator
Moderator
 
Posts: 2681
Joined: Mon Jan 19, 2009 3:43 pm
Location: Los Angeles, CA
Real name: Unpronounceable
Gear: Ensoniq SD-1/32,SQR,VFX,ESQm
Virus Indigo, M3-61 , MS2000BR, Volca Bass
Emu XL-7, Matrix 6r
TG-33, K3m, Blofeld, Micron, Mopho, Bass Station II
Band: Eridani V

Re: Tech Question about Hardware Samplers 80/90s.

Postby piRoN » Wed Apr 17, 2013 5:43 am

Yeah, just clarifying that for the OP's question about processing.

With fixed-point processing you can use standard integer operations for the most part, and use software implementations of the operations that behave differently (multiply/divide etc) using simpler algorithms than software floating-point. The trade-off is less range and precision, but seeing as digital audio operates in a predictable, limited range that's not a great issue, and the precision is more than adequate given the quality of other parts of the signal chain at the time.
Ensoniq EPS-m | Elby Designs ASM-2 | DSI Evolver | Yamaha TQ5 | Akai S3000XL | Akai X7000 | Novation KS-4 | SCI Prophet 2000 | Kawai K1m | Roland S-50 | Roland D-10 | Roland MKS-7 | Roland TR-606 | Yamaha RX11
User avatar
piRoN
Active Member
Active Member
 
Posts: 558
Joined: Sun Dec 19, 2004 7:04 am
Location: The crotch of The Lucky Country
Gear: A fair bit.


Return to Samplers

Who is online

Users browsing this forum: No registered users and 1 guest