cullyrichard
Experienced Member
Hi all,
I think it's worth starting out with some background as this machine is quite complex. The FPS100 assembly language manual describes the FPS100 in this way: The Floating Point Systems, Inc., FPS-100 is a peripheral device that operates independently from but under the direction of a host processor. It contains its own internal memories and 38-bit floating-point arithmetic units which are interconnected with multiple data paths, allowing parallel internal data transfers. Its arithmetic units, the floating adder and floating multiplier, are designed as pipelines (operations are performed in independent stages permitting new operations to begin before old operations are complete). This parallel processing capability and pipeline arithmetic permit the FPS-100 to perform high speed array processing.
The Floating Point Systems family of machines, the AP120B and FPS100 are largely software compatible and found themselves in many of the same applications, The key difference is that the FPS100 is quite a bit more compact and slightly slower.
The FPS family of machines had some very significant applications, those sold with PDP11 interfaces often found themselves used for high performance compute tasks, often in classified environments where other machines were not practical. These machines were also heavily used for oil and gas exploration on a variety of hosts and in research environments. The Data General interfaced machines (both the AP120B and FPS100) were used frequently in medical applications. GE Healthcare's CT8800 and CT9800 machines used the AP120B and FPS100 as their primary reconstruction processor, employing a custom card called the Backprojection Image Reconstruction Processor or BIRP, do do the arctangent and other tasks associated with image reconstruction. the CT9800 was often upgraded with two FPS100s to accelerate the reconstruction process. Ohio Nuclear / Technicare CT machines used the AP120B on a PDP11/34 host, but little is known about these machines. We have spoken at length with the engineers from GE and the FPS100 project lead, and much of this information is thanks to their efforts and willingness to help.
There are presently 2 known FPS100 systems, one FPS100R, Serial number FA00721, a GE CT scanner secondary unit that lacks the BIRP, used by GE Healthcare as their development unit and an FPS5100 (the FPS5000 series was an FPS100 that supported a coprocessor) in germany. The FPS5100 is configured quite a bit differently and relies on some condensed boards to support the coprocessor. There are a few FPS AP120Bs around, two in China and one on ebay at an absolutely ludicrous price and an additional two in a museum. There may also be another attached to a Raytheon host. Efforts are underway to build a new backplane and chassis to support the PDP11 Interface as detailed on the Usagi Electric Youtube channel.
The story of how FA00721 ended up in my posession is wild. The machine was sold as surplus by GE in the 1980s as the CT9800 matured to the CT9800 Quick, the CT9800 Quick uses a custom reconstruction processor. This machine was purchased by one of the engineers at GE and found its way through several hands before finding its way to New York. We were able to identify the machine and contact it's present owner. We then worked out a deal and I drove several hours to pick it up. The machine includes the very special FPS Data General interface card, which was developed primarily for medical applications.
The next task is to attach it to my DG Eclipse S/140. If you have any software or are interested in helping us port the PDP11 software, I'd love to hear from you. We're also looking for more FPS machines to support the revival efforts. If you have any hardware or software you're interested in parting with. Please do get in touch.
Thanks!
Cully
I think it's worth starting out with some background as this machine is quite complex. The FPS100 assembly language manual describes the FPS100 in this way: The Floating Point Systems, Inc., FPS-100 is a peripheral device that operates independently from but under the direction of a host processor. It contains its own internal memories and 38-bit floating-point arithmetic units which are interconnected with multiple data paths, allowing parallel internal data transfers. Its arithmetic units, the floating adder and floating multiplier, are designed as pipelines (operations are performed in independent stages permitting new operations to begin before old operations are complete). This parallel processing capability and pipeline arithmetic permit the FPS-100 to perform high speed array processing.
The Floating Point Systems family of machines, the AP120B and FPS100 are largely software compatible and found themselves in many of the same applications, The key difference is that the FPS100 is quite a bit more compact and slightly slower.
The FPS family of machines had some very significant applications, those sold with PDP11 interfaces often found themselves used for high performance compute tasks, often in classified environments where other machines were not practical. These machines were also heavily used for oil and gas exploration on a variety of hosts and in research environments. The Data General interfaced machines (both the AP120B and FPS100) were used frequently in medical applications. GE Healthcare's CT8800 and CT9800 machines used the AP120B and FPS100 as their primary reconstruction processor, employing a custom card called the Backprojection Image Reconstruction Processor or BIRP, do do the arctangent and other tasks associated with image reconstruction. the CT9800 was often upgraded with two FPS100s to accelerate the reconstruction process. Ohio Nuclear / Technicare CT machines used the AP120B on a PDP11/34 host, but little is known about these machines. We have spoken at length with the engineers from GE and the FPS100 project lead, and much of this information is thanks to their efforts and willingness to help.
There are presently 2 known FPS100 systems, one FPS100R, Serial number FA00721, a GE CT scanner secondary unit that lacks the BIRP, used by GE Healthcare as their development unit and an FPS5100 (the FPS5000 series was an FPS100 that supported a coprocessor) in germany. The FPS5100 is configured quite a bit differently and relies on some condensed boards to support the coprocessor. There are a few FPS AP120Bs around, two in China and one on ebay at an absolutely ludicrous price and an additional two in a museum. There may also be another attached to a Raytheon host. Efforts are underway to build a new backplane and chassis to support the PDP11 Interface as detailed on the Usagi Electric Youtube channel.
The story of how FA00721 ended up in my posession is wild. The machine was sold as surplus by GE in the 1980s as the CT9800 matured to the CT9800 Quick, the CT9800 Quick uses a custom reconstruction processor. This machine was purchased by one of the engineers at GE and found its way through several hands before finding its way to New York. We were able to identify the machine and contact it's present owner. We then worked out a deal and I drove several hours to pick it up. The machine includes the very special FPS Data General interface card, which was developed primarily for medical applications.
The next task is to attach it to my DG Eclipse S/140. If you have any software or are interested in helping us port the PDP11 software, I'd love to hear from you. We're also looking for more FPS machines to support the revival efforts. If you have any hardware or software you're interested in parting with. Please do get in touch.
Thanks!
Cully
