diff options
Diffstat (limited to 'introduction.typ')
| -rw-r--r-- | introduction.typ | 281 |
1 files changed, 281 insertions, 0 deletions
diff --git a/introduction.typ b/introduction.typ new file mode 100644 index 0000000..8631e46 --- /dev/null +++ b/introduction.typ @@ -0,0 +1,281 @@ +#import "@preview/unify:0.8.1": add-unit, qty + +// All units for the specification should be specified here +#add-unit("ppm", "ppm", "upright(\"ppm\")") +#add-unit("ppb", "ppb", "upright(\"ppb\")") +#add-unit("inch", "\"", "upright(\"\\\"\")", space: false) +#add-unit("digital voltage", "VD", "upright(\"VD\")") +#add-unit("volt peak-to-peak", "Vpp", "upright(V_\"pp\")") +#add-unit("milli volt peak-to-peak", "mVpp", "upright(\"mV\"_\"pp\")") +#add-unit("micro volt peak-to-peak", "uVpp", "upright(mu V_\"pp\")") +#add-unit("dBc", "dBc", "upright(\"dBc\")") // decibels relative to the carrier +#add-unit("rack unit", "U", "upright(U)", space: false) + += Introduction <introduction> + +== What is analog computing? <what-is-analog-computing> + +Analog computing is the action of leveraging the properties of physical +phenomena to model mathematical principles. The actual phenomena can be as +diverse as hydraulics, mechanics, or electrics. In an electronic analog +computer, voltages represent variables and operational amplifier circuits +perform mathematical operations on them. + +Modeling the mathematical principles being computed implies that the results are +continuous in time and amplitude. This directly contrasts digital computing, +where results are symbolic and discrete. + +== Short history of analog computing <short-history-of-analog-computing> + +=== Mechanical and hydraulic analog computers +<mechanical-and-hydraulic-analog-computers> + +Analog computing traces its roots to ancient Greece, with the Antikythera +mechanism. It re-emerged time and time again across the ages, notably with the +invention of the Slide Rule for computation of multiplications and divisions in +the 1620s, and Lord Kelvin's Tide-Predicting machine invented in 1872. +Mechanical or hydraulic analog computers (sometimes electronically controlled) +continued to be the standard until the late 1950s. + +=== First electronic analog computers <first-electronic-analog-computers> + +In 1941, Karl D. Swartzel Jr. of Bell Labs patented the first operational +amplifier (op-amp). 1942 saw the birth of the first fully electronic analog +computer for real-time computation by Helmut Hölzer, to calculate V-2 rocket +trajectories. In 1950, Gilbert D. McCann, Charles H. Wilts, and Bart Locanthi +developed the "Direct Analogy Electric Analog Compute" ("the largest and most +impressive general-purpose analyzer facility for the solution of field +problems" - James S. Small, "The Analogue Alternative: The Electronic Analogue +Computer in Britain and the USA, 1930-1975" Routledge in 2001). + +In 1960, the first educational electronic analog computer, the Heathkit EC-1, +was released. It was programmed with patch cords and featured nine operational +amplifiers (op-amps). + +In the late 1960s, electronic analog computer designs were published in +electronics magazines. This is notably the case for the PEAC (Practical +Electronics Analog Computer), published in 1968 in Practical Electronics. + +=== Golden age of electronic analog computers +<golden-age-of-eletronic-analog-computers> + +Starting with the EAI PACE TR-48, produced by Electronic Associates Inc. in +1961, hybrid computers became the industry standard for high-performance tasks. +These machines combined the speed of analog computation for solving differential +equations with the logical control and storage of digital computers. They were +heavily utilized in the aerospace and automotive industries for real-time +simulation — such as modeling the Space Shuttle's flight dynamics or nuclear +power plant operations — tasks where purely digital computers of the era were +still too slow. + +=== Decline of analog computing <decline-of-analog-computing> + +The decline began in the late 1970s and accelerated swiftly throughout the +1980s, driven by the "Digital Revolution." While analog computers were fast, +they suffered from physical limitations like noise floors and component aging +(drift), which limited their precision compared to the perfect repeatability of +digital machines. Furthermore, programming an analog computer required +physically rewiring "patch cords," a laborious process compared to the rapidly +evolving software-based programming of digital systems. + +=== Current times <current-times> + +A new rise of analog computing is currently underway, often referred to as +"Neuromorphic Computing" or "In-Memory Computing". This generation uses the +physical properties of memory devices (like resistance) to perform calculations, +mimicking the neural structure of the human brain to process information. In +2021, startup Mythic introduced the M1076 Analog Matrix Processor, a chip that +performs AI inference using the analog currents inside flash memory cells, +delivering up to 25 TOPS (Trillion Operations Per Second) at incredibly low +power. Mythic filed for bankruptcy in 2024, but managed to raise \$125 million +in December 2025, showing that there is still turbulence in the neuromorphic +startup space. + +The field is still very active, notably with the Innatera Pulsar or the Applied +Brain Research TSP1, both released in 2025. + +== The case for modern general purpose analog computers +<the-case-for-modern-general-purpose-analog-computers> + +We believe the time is ripe for a revival of the General Purpose Analog +Computer. Reintroducing general-purpose analog computing is not an exercise in +retro-computing; it is a solution to specific modern problems. By combining the +transparency of "Golden Age" architectures with modern low-noise components, we +can achieve a new standard of computing that is sustainable, secure, and +strategically independent. + +=== Theoretical capabilities <theoretical-capabilities> + +Historically, the General Purpose Analog Computer (GPAC) was perceived as being +limited to solving differential equations, distinct from the universal logic of +Turing Machines. However, this distinction assumes a finite system. In a +sufficiently large analog system (the same way a digital computer is a Turing +Machine assuming sufficiently large memory), the combinatorial logic required +for universal computation is inherent in the topology. + +We do not rely on discrete "extensions" like Sample & Hold to achieve Turing +Completeness; we use them merely for convenience. The $I$$(I x = x)$, $K$$(K x y + = x)$ and $S$$(S x y z = x z(y z))$ Combinators (which are sufficient for +Turing Completeness) are not a special digital mode; they are a description of +signal distribution and modulation that exists in every analog patch. With +enough modules, the "patch" becomes a computational engine of unbounded +complexity, capable of recursive structures defined purely by feedback loops +rather than iterative loops. This argument will be fully developed in further +sections. + +=== Technological sovereignty <technological-sovereignty> + +In an era of chip shortages and trade restrictions, reliance on $qty(5, "nm")$ +or $qty(3, "nm")$ silicon creates critical vulnerabilities. A computer built +from standard through-hole operational amplifiers and passive components does +not require a multi-billion dollar foundry. It can be assembled in a standard +electronics lab anywhere in the world. + +With a 100% open-source design and macro-scale components, the hardware is fully +auditable. There are no hidden "black box" intellectual property cores or +potential hardware backdoors, ensuring high confidence in the computing +substrate for sensitive modeling. + +=== Cost and accessibility <cost-and-accessibility> + +High-performance simulation usually requires expensive workstations or licensing +fees for proprietary software. By utilizing off-the-shelf components (common +op-amps, resistors, standard capacitors), the Bill of Materials (BOM) is a +fraction of the cost of a modern FPGA or GPU-based accelerator. The +"Through-Hole First" design philosophy means these machines can be produced, +modified, and repaired by students, hobbyists, or labs in developing nations, +without the need for robotic surface-mount assembly lines. + +=== Energy efficiency <energy-efficiency> + +While digital computers must toggle millions of transistors at gigahertz speeds +to approximate a differential equation, an analog computer simply is the +equation. Solving a complex set of coupled differential equations on this +standard requires only the static power to bias the amplifiers. It avoids the +massive dynamic power consumption associated with the high-frequency clock +cycles of digital processors. Using an analog architecture delivers results with +significantly lower total energy consumption per solution, aligning with global +goals for reduced carbon footprints in data processing. This will be fully +developed in further sections. + +=== Sustainability through repairability <sustainability-through-repairability> + +A 100% through-hole and open-source design ensures that every component is +human-accessible. There are no microscopic BGA chips; the machine can be +maintained indefinitely, breaking the environmentally destructive cycle of +planned obsolescence. It essentially enforces the right to repair while allowing +a reduction of e-waste. + +== The SAME Analog Modular Ecosystem <the-same-analog-modular-ecosystem> + +With this specification, we introduce a new analog computing ecosystem. We +believe that with the right infrastructure, modern components and the advances +in electronics research, we can now build analog computers that address the +three main points that caused the decline of analog computing in the 1980s: +cost, precision and drift. + +=== Making analog computers reliable and sustainable +<making-analog-computers-reliable-and-sustainable> + +Electronics components are cheaper than ever right now. Op-amps that once cost +hundreds of euros now cost at most a few dozen euros. PCB manufacturing is +democratized. The material conditions are here. New advances in electronics +research have solved the two remaining issues: drift and precision: with a +combination of laser-trimmed components and topological advancements (servo +loops, automated gain control, pulse width amplitude modulation, etc.), it is +now possible to design analog circuits that are able to reach 0.001% precision +(10ppm), over a $qty(90, "dB")$ signal to noise ratio (SNR), in a bandwidth +ranging from DC to more than $qty(20, "kHz")$. + +=== What SAME is not <what-same-is-not> + +SAME is not a modular synthesizer format. It is not Eurorack, Serge or Buchla. +While those synthesizer formats are based on general purpose analog computers, +they are not designed for precision computations or repeatability. While SAME +can be used for sound design or experimental composition, it does not provide +the abstractions and simplifications that make synthesizers powerful. + +SAME is not a digitally controlled or calibrated analog computer. It focuses on +fully analog computer topologies. It advocates a "what you see is what you get" +approach. It has no presets, no patch memory. But it is precise enough that a +patch sheet and a pen and paper allow full patch repeatability. We believe this +approach is needed as it is more educational, and eliminates black boxes in +programming (patches) and design (topologies). It is also more sustainable, as +a design doesn't depend on a microcontroller and code that might fail and be out +of production in five years. + +SAME is not a system on a chip (SOC). The revival of analog computing in the +2020s has been mainly driven by SOCs. We do recognize the importance of this +approach, and its many benefits, notably in energy consumption. However, SOCs +require digital control. With SAME, we aim at creating an infrastructure for +sustainable electronics. A unit built in 2026 should still work in 2076. Some +parts will have been changed, but by focusing on through-hole parts in +production at the time of designing, we ensure that a human with a soldering +iron will be able to repair them. + +=== The scope of this specification <the-scope-of-this-specification> + +In this specification, we will provide full details and rationale behind the +SAME format. We will provide schematics, gerber files, CAD files, as well as +ngspice simulations of a reference implementation. We will provide a validation +methodology for our precision claims, as well as the data to support them. + +Finally, we will propose a tiered approach to precision: we believe that a +topology that manages 10ppm precision and $qty(90, "dB SNR")$ SNR over DC to +$qty(20, "kHz")plus$ bandwidth is worth using and studying even with a lower +precision target of $qty(100, "ppm")$ or $qty(1000, "ppm")$. While to achieve +$qty(10, "ppm")$ some parts might get expensive due to their very high +precision, someone could still achieve great reliability with lower precision +parts. For this reason, while we will propose only one specification for the +SAME chassis and power distribution, for each module in the reference +implementation, we will propose three bills of material (BOM): Educational, +aiming at $qty(1000, "ppm")$; Industrial, aiming at $qty(100, "ppm")$; and +Metrologic aiming at $qty(10, "ppm")$. + +=== Reader's guide <readers-guide> + +We recommend a reader not familiar with the field to first read this +specification sequentially. Sections 2 and 3 will develop the core +specifications of the SAME ecosystem. Section 4 will specify the module design +principles that allows it to achieve its precision goals. Section 5 will explain +how to verify conformity of a SAME product with this specification. Section 6 +will present a reference implementation of the system. Finally, sections 7 and 8 +will show how and when to use a SAME Analog Computer. + +== Document conventions <document-conventions> + +To ensure interoperability between modules, chassis backplanes, and signal +routing standards within the SAME Analog Modular Ecosystem, this specification +utilizes specific keywords to denote requirement levels. + +The key words MUST, MUST NOT, REQUIRED, SHALL, SHALL NOT, SHOULD, SHOULD NOT, +RECOMMENDED, MAY, and OPTIONAL in this document are to be interpreted as +described in IETF RFC 2119. + +/ MUST / SHALL / REQUIRED: These terms indicate an absolute requirement of the + specification. Hardware or documentation that fails to meet these criteria is + non-compliant. Imperative statements (commands) that do not employ specific + keywords are also considered to be in this default state. + - Example: "The reference voltage rail MUST maintain stability within ±0.01% + of the nominal 10V output." +/ MUST NOT / SHALL NOT: These terms indicate an absolute prohibition of the + specification. + - Example: "Signal ground MUST NOT be tied directly to the chassis earth at + the module level." +/ SHOULD / RECOMMENDED: These terms indicate that there may exist valid reasons + in particular circumstances to ignore a particular item, but the full + implications must be understood and carefully weighed and documented before + choosing a different course. + - Example: "Input impedance for operational amplifiers SHOULD exceed 1MΩ to + minimize loading effects." +/ SHOULD NOT / NOT RECOMMENDED: These terms indicate that there may exist valid + reasons in particular circumstances when the particular behavior is acceptable + or even useful, but the full implications should be understood and the case + carefully weighed and documented before implementing any behavior described + with this label. +/ MAY / OPTIONAL: These terms indicate that an item is truly optional. One + vendor may choose to include the item because a particular marketplace + requires it or because the vendor feels that it enhances the product while + another vendor may omit the same item. + - Example: "The integration module MAY include a digital overflow indicator + LED." |