The use of AI in the repair and study of SLRs

[Andreas Thaler]

Better late than never, I’ve started looking into AI. Specifically, how it can help me with servicing and repairing SLRs, with technical understanding, and with researching repair manuals.

To do this, I used the AI feature on Google Search, which is publicly accessible via a web browser.

I asked three questions:

1. about soldering in 1980s SLRs,
2. about the electronic controls in late 1970s SLRs, and
3. where I could find information on a specific topic in the Nikon FA repair manual (available online).

The answers (see below) are available within one to two seconds. I have prior knowledge regarding all three questions, so I should be able to evaluate the results.

---

Questions and answers

1. When soldering 1980s SLRs, should I use lead-free or lead-based solder?

Lead-based solder (specifically 63/37 or 60/40 Sn/Pb) is the correct choice for soldering 1980s SLR cameras. [1, 2]

Why Lead-Based Solder is Required​

• Metallurgical Compatibility: 1980s electronics were manufactured entirely using leaded solder. Mixing lead-free solder with existing leaded solder creates a weak, brittle alloy prone to mechanical failure.
• Lower Melting Point: Lead-based solder melts at around 183°C (361°F), whereas lead-free alternatives require temperatures above 217°C (422°F).
• Component Protection: The delicate flexible circuit boards (flex cables) and heat-sensitive ICs in 1980s SLRs degrade or delaminate easily under the higher heat required for lead-free solder.
• Prevention of Tin Whiskers: Lead-free solders are highly susceptible to growing microscopic "tin whiskers" over time, which cause short circuits on densely packed camera circuit boards. [3, 4, 5, 6, 7]

Recommended Specifications​

• Alloy Type: Use 63/37 Sn/Pb (Eutectic) solder. It transitions instantly from liquid to solid, minimizing the risk of cold joints if the component moves while cooling. 60/40 Sn/Pb is a highly acceptable alternative.
• Diameter: Opt for thin wire between 0.5mm (0.020") and 0.8mm (0.031"). Camera contacts are tiny and precise.
• Flux Core: Choose a Rosin Core (Type R or RMA) solder. Avoid acid-core solders completely, as they corrode electronic components. [8, 9, 10, 11, 12]

Essential Safety Practices​

• Ventilation: The primary health hazard comes from breathing in the vaporized rosin flux fumes, not the lead itself. Always work in a well-ventilated space or use a dedicated fume extractor.
• Hygiene: Lead does not absorb through the skin, but it can be ingested accidentally. Never eat or drink while soldering, and wash your hands thoroughly with cold water and soap immediately after finishing. [13, 14, 15, 16, 17]

If you want to prepare your workspace for camera repair, let me know:

• Which specific SLR model you are working on
• What tools you currently have (e.g., adjustable temperature soldering iron, flux pen)
• The type of repair you are attempting (e.g., replacing a wire, fixing a battery compartment, swapping a capacitor)

I can provide tailored steps or recommend specific equipment for your project.

[1] https://www.toolworld.in
[2] https://www.coilcraft.com
[3] https://www.allpcb.com
[4] https://www.sciencedirect.com
[5] https://www.reddit.com
[6] https://www.allpcb.com
[7] https://www.allpcb.com
[8] https://www.sciencedirect.com
[9] https://www.tch.es
[10] https://www.allpcb.com
[11] https://www.allpcb.com
[12] https://www.reddit.com
[13] https://forum.arduino.cc
[14] https://blog.gotopac.com
[15] https://jhdpcb.com
[16] https://www.delphiglass.com
[17] https://www.farnell.com

---

2. I’m interested in the basic operation of an electronic SLR, using a late 1970s SLR as an example. Is the following description accurate?

An analog value—for example, a voltage drop across the resistance track of the ASA dial caused by a wiper—is transmitted to the control electronics. There, an A/D converter digitizes this value into a binary code. The binary value is routed through fixed logic gates and combined with other binary values; the result is converted by a D/A converter from a binary value into a voltage, which then moves a measuring needle, for example.

Your description is highly accurate for electronic SLR cameras of the late 1970s (such as the landmark Canon AE-1). During this transitional era, manufacturers relied on exactly the hybrid analog-digital architecture you outlined.

The System Workflow:​

• Analog Inputs: The ASA and shutter speed dials used mechanical wipers on continuous potentiometers to output raw voltage drops.
• A/D Conversion: A custom Large Scale Integration (LSI) circuit converted these analog voltages into binary code.
• Digital Logic: Fixed digital logic gates performed binary arithmetic to solve the logarithmic exposure equation.
• D/A Conversion: An on-chip resistor ladder (DAC) converted the calculated digital answer back into a variable analog current.
• Analog Display: This current directly drove a mechanical galvanometer needle up and down inside the viewfinder.

Why this Only Applies to the Late 1970s:​

By 1980, this complex analog-to-digital-to-analog process was quickly abandoned. 1980s designs switched to digital code wheels for inputs and solid-state LED or LCD panels for displays, rendering both data converters and the fragile mechanical needle obsolete.
If you would like, I can provide more details on:

• The exact resistor ladder (DAC) layout used in late-70s ICs
• How the silicon photodiode's light readings were digitized
• A comparison with a specific camera model like the Canon AE-1 or Minolta XD

Let me know which technical aspect you want to explore further.

---

3. Where in the Nikon Repair Manual for the FA is the adjustment of the multi-segment metering (AMP) described?

In the official Nikon FA Repair Manual, the exposure metering adjustments (which include the Center-Weighted and Automatic Multi-Pattern/AMP systems) are located in the "Electrical Adjustment" (or "Adjustment") section, typically starting around Page 41 to Page 46 (labeled as E1 to E6 in the electrical subsection). [1]

Key Adjustment Overview:​

1. Central Calibration: The manual does not outline a distinct, standalone physical potentiometer exclusively for "AMP." Instead, the AMP system relies heavily on the baseline accuracy of the multi-segmented silicon photodiode (SPD) cells and the primary Center-Weighted baseline calibration.
2. Main Potentiometers: Variable resistors located on the flexible printed circuit (FPC) board under the top cover (near the pentaprism/eyepiece assembly) dictate the main voltage offsets:
   • VR101 / VR102 / VR103 series: Handle critical calibrations such as the basic standard exposure level (EV calibration), ASA/ISO compensation, and operational amplifier offsets.
3. The Microcomputer Link: Because AMP utilizes a digital A/D conversion layout managed by the camera's CPU to process contrast and brightness distributions across five distinct zones, any general drift in the metering values must be first calibrated in standard center-weighted mode. [2, 3]

Before Attempting Adjustments:​

• The AMP Mechanical Switch: Ensure that the mechanical AMP/Center-Weighted selector pin below the lens mount is operating correctly. The internal contact brushes under the lens mount box signal the mode to the CPU; a mechanical fault here often mirrors an electronic metering error.
• FRE (Film Rate Resistor): Clean the FRE pattern under the shutter speed/ISO dial assembly first, as resistance tracking issues here simulate overall metering failures across all segments. [4, 5, 6]

Would you like the exact voltage specifications and tester terminal pinouts listed in the manual for the EV adjustment, or details on how to safely remove the top cover to access these specific trimpots?

[1] https://www.wahyan.edu.hk
[2] https://foto.aaweb.cz
[3] https://mir.com.my
[4] https://www.mir.com.my
[5] https://www.wahyan.edu.hk
[6] https://www.dpreview.com

---

Conclusion

• AI in this form brings a new dimension to working on SLRs.
• Questions are answered in a tailored manner, in more detail if desired or just as an overview.
• Follow-up questions are answered in dialogue with the AI based on the previous answers.
• Source references help assess the quality of the answer.
• Time-consuming research is eliminated.
• The quality of the answers is impressive.
• Thinking critically and verifying the information is always necessary.
• (Commercial) AI that can process documents offline—and thus has access to, for example, collections of repair manuals—will increase the amount of information available.
• Essential for my work, effective immediately.

 

[ic-racer]

I’ve been soldering and repairing electrical devices since the 1960s I’ve never seen a tin whisker, have you?

 

[Andreas Thaler]

I’ve been soldering and repairing electrical devices since the 1960s I’ve never seen a tin whisker, have you?

No, but it looks nice:

Whisker (metallurgy)

 

[Andreas Thaler]

This is also interesting for examining workflows.

In this case, however, I had to clarify my question („Alkaline“ „parts already removed“, „about to be rinsed with water“) until I received a positive response. Previously, the AI had advised me against it.

Can I clean battery contacts that have been contaminated with battery electrolyte (alkaline, AA, AAA) —and which I have already removed—using Durgol Universal Descaling Solution, provided that I rinse them immediately with water, soak them in isopropyl alcohol, and let them dry?

Yes, you can use this method. Since you have already removed the contacts from the device, this procedure is highly effective and chemically sound.
Alkaline battery (AA/AAA) leakage is potassium hydroxide, which is a strong base. The acidic nature of Durgol Universal will chemically neutralize it. Following up with water, isopropyl alcohol (IPA), and drying is the correct sequence to prevent corrosion.
Here is the breakdown of why your method works and how to execute it safely:

1. Durgol Neutralization

• How it works: Durgol Universal contains sulfamic acid (amidosulfonic acid). This acid rapidly neutralizes the alkaline potassium hydroxide crust and dissolves the salts.
• Best practice: Submerge the contacts only until the fizzing stops. Do not leave them in the solution longer than necessary, as the acid can eventually attack the underlying metal or its protective plating (usually nickel or tin).

2. Immediate Water Rinse

• How it works: This stops the acid reaction instantly and washes away the neutralized chemical salts.
• Best practice: Use distilled or demineralized water if possible. Tap water contains dissolved minerals and chlorides that can leave deposits or initiate micro-corrosion on the freshly cleaned metal.

3. Isopropyl Alcohol (IPA) Soak

• How it works: Pure IPA (99%) displaces water from the microscopic pores of the metal and evaporates rapidly. This prevents any flash rusting.
• Best practice: Agitate the contacts in the IPA bath for a few seconds to ensure all water is displaced.

4. Post-Drying Optimization

• Polishing: The battery acid may have tarnished or darkened the metal surface. Once the contacts are dry, gently rub the contact points with a pencil eraser or a fine brass wire brush to polish away the tarnish and restore maximum electrical conductivity.

Would you like advice on how to protect the clean contacts from future corrosion, or do you need tips on reinstalling them into the housing?

So I will use the AI's responses, as well as those from my fellow forum members and my own, as follows:

„All information provided without guarantee and use at your own risk.“

A wide range of possibilities opens up

 

[koraks]

One area of interest in using AI is how to use it effectively and responsibly. There's a couple of pitfalls, such as:
* The formulation of the output is generally highly authoritative, but this is not necessarily in line with the validity and reliability of the answers. It's easy to be lured into overconfidence.
* There's a bias in the answers towards suiting the 'perceived' preference of the user. Simply put: it's flattering.
* The content of the output can change sometimes in 180 degree turns if a subsequent prompt it used to trigger a more critical assessment.
* The output will depend highly on which model and model version is used. It's also idiosyncratic; giving the same prompt twice does not guarantee receiving the same response. This has implications for how to interpret the reliability of the information obtained.
* At this stage, the LLM's I've tried are not very good at doing architectural work; i.e. the more structural aspects to a process, design, problem-solving remain relatively weak or implicit, and solutions offered as a result do not always fit the problem very well (even though they may still technically work). This is very apparent when using AI to generate software code, for instance, where in my experience it tends to produce convoluted, inflexible structures (e.g. Copilot GPT5.1 and Claude Pro).

The main recommendation I'd like to make is to remain very critical of the output and to probe important parts of the answer, prompting for sources and actually verifying those sources against the claims made by the LLM used. I've come across many instances where AI gave a neat list of sources for its answer, while on closer scrutiny, none of those sources really supported the core of the answer.

Having said that, it can be a really powerful and convenient tool, but it's also a bit like putting an F1 car up for sale to the general public. Violent run-ins with trees are common.

 

[Andreas Thaler]

Having said that, it can be a really powerful and convenient tool, but it's also a bit like putting an F1 car up for sale to the general public. Violent run-ins with trees are common.

I’ve only recently started looking into this topic, but it seems to me that AI has rapidly evolved from being a laughingstock to a serious application—if not more. Where will we be in two or three years?

 

[FredR]

And AI will hallucinate, disgorging fantasy facts. One French lawyer is tracking the embarrassing results when used in legal proceedings: https://www.damiencharlotin.com/hallucinations/
Since AI is biased and trained to defend itself, asking AI about this issue simply generates more BS.

 

[Andreas Thaler]

And AI will hallucinate, disgorging fantasy facts. One French lawyer is tracking the embarrassing results when used in legal proceedings: https://www.damiencharlotin.com/hallucinations/
Since AI is biased and trained to defend itself, asking AI about this issue simply generates more BS.

But the information the AI provided on the technical topics (see above) was flawless. Are there some things it does better, some it does worse, and some it can’t do at all?

 

[FredR]

But the information the AI provided on the technical topics (see above) was flawless....

How did you confirm that it is flawless? Have you since researched what it generated without using AI? Or will you be satisfied because it matches what you want to see?

 

[Andreas Thaler]

How did you confirm that it is flawless? Have you since researched what it generated without using AI? Or will you be satisfied because it matches what you want to see?

I entered questions whose answers I already knew. KI’s answers match.

 

[FredR]

How did you confirm that it is flawless? Have you since researched what it generated without using AI? Or will you be satisfied because it matches what you want to see?

I entered questions whose answers I already knew. KI’s answers match.

There ya go!

 

[Rayt]

I entered questions whose answers I already knew. KI’s answers match.

Your trust in information provided by others including AI is based on your own level of knowledge? If someone told me 1+1 =2 then everything else will be accurate? Certain skills can only be learned from practice and a super computer can only gather published information. There is always a gap.

 

[Andreas Thaler]

There ya go!

And so I conclude that AI will also deliver good results in other areas.

But that’s not what I’m concerned with; I’m interested in how the bot can help me with my work in the workshop.

 

[Andreas Thaler]

Your trust in information provided by others including AI is based on your own level of knowledge? If someone told me 1+1 =2 then everything else will be accurate? Certain skills can only be learned from practice and a super computer can only gather published information. There is always a gap.

How AI does that is interesting, but it doesn’t concern me.

The three examples I cited encourage me to continue exploring the use of AI in the fields of repair and engineering. From what I’ve seen so far, there’s a lot to be gained today, and even more tomorrow.

As I wrote, you have to evaluate the results, but that applies to any information—including, for example, what you’re writing right now. There will never be absolute certainty as to whether a piece of information is correct, partially correct, or incorrect. In a field I’m familiar with, that’s manageable.

 

[koraks]

I thought of a recent anecdote just now; I wanted to share. It's not intended as criticism or support for AI; just to highlight the limitations that we need to take into account with the present state of technology.

A few weeks ago, our dishwasher malfunctioned. It's not very old (4-5 years) and produces and error that's apparently a notorious failure mode for this model. I asked Copilot what the suggested approach would be to fix the darn thing. It came up with some interventions that matched with (and were based on) the same that several websites offered. They struck me as mostly battling the symptoms, and wondering whether perhaps insufficient cleaning or blockage in the drain subsystem might be a contributing cause, I put this to Copilot as a hypothesis. It said something along the lines of "absolutely, you're right on the money, try cleaning the machine etc." So I did.

It seemed to help - for a brief while. Then the problem recurred. Long story short, I opened up the machine and solved the problem by addressing the actual root cause. This was entirely different from what I suspected - and what Copilot asserted to me when I asked it!

What this shows mostly is how LLM's can easily bias towards agreeing with the user's prompt. I also noticed while working on the machine that only a single YouTube video suggested a proper solution for the problem, while all other websites and videos demonstrated a band-aid workaround. It was the latter that dominated the AI output as well - probably just because those data featured more prominently in its training data set and the quick & dirty fix mostly went unchallenged. However, getting in there and doing some proper troubleshooting, including a fairly simple analysis of how that part of the machine works, gave access to a far more robust solution. So another insight is that problem-solving consists not only of recognizing a known pattern (problem-solution pairs already published online), but also analytical thinking in terms of cause-effect relationships and architectural patterns.

I think this reasonably simple example illustrates some of the issues I mentioned above. Mind you, I still use Copilot (etc.) for troubleshooting and I still find it useful, but it seems to work the best for me as long as I pay heed to the several pitfalls and work to avoid them.

 

[Andreas Thaler]

A real-world question.

Let’s say I’m taking a camera apart and come across an E-clip for the first time. You can find more information here in just a second.

What is the safest way to remove a 4 mm e-clip? (Summary)

Here is the compact summary for safely removing a 4 mm E-clip:

• The Main Risk: The tiny clip is under high spring tension and easily flies off or gets lost during removal.
• The Best Protection: Place the entire assembly and your tools inside a clear plastic bag. This stops flying parts while keeping full visibility.
• The Tool Choice: Use a specialized E-clip tool or a small flathead screwdriver that fits tightly into the recess of the center prong.
• The Removal: Place a gloved finger over the clip to hold it down, push the tool straight away from the shaft (or gently twist the screwdriver), and catch the clip with a magnet.

Would you like to know how to reinstall the clip safely, or do you need tips for a rusted/stuck component?

 

[Andreas Thaler]

Naturally, that gets me thinking. How much do I know, and how much does the AI know? If I can have problems solved from different perspectives in two seconds, have manuals summarized, or have circuit operations analyzed—what does that mean for the way I acquire knowledge?

None of my books tell me how the circuitry of the Canon AE-1 is designed and how it works. I’d have to read even Larry Lyell’s work intensively to be able to visualize that from the perspective I want. All the AI needs from me is a specific question.

It’s not perfect, it has risks, but it’s a step toward entirely new possibilities.

In my future repair projects, I'll incorporate AI and report on the results.

 

[retina_restoration]

I showed chatGPT a Kodak Retina IIIc type cocking rack and asked it to identify the part. It wasn't even close to getting it right, even when I told it what camera it was from. I didn't bother to explore further. YMMV.

 

[MattKing]

The AI related concerns and advantages discussed in this thread are parts of the reason we have set up a separate sub-forum for AI in photography related discussions.
At present, we are considering moving the thread there.
If you are curious, the sub-forum is here: https://www.photrio.com/forum/forums/ai-in-photography-and-other-arts.408/

 

[Andreas Thaler]

The AI related concerns and advantages discussed in this thread are parts of the reason we have set up a separate sub-forum for AI in photography related discussions.
At present, we are considering moving the thread there.
If you are curious, the sub-forum is here: https://www.photrio.com/forum/forums/ai-in-photography-and-other-arts.408/

Thanks, Matt!

 

[mshchem]

What do you do when you find that you are Ai and the Ai is Human???

 

[Andreas Thaler]

One challenge with 1980s SLRs

is finding verified information about their electronic circuits.

Original sources include the manufacturers’ service manuals, the SPT Journal, and the C & C Troubleshooting Guide.

These documents typically provide only electrical specifications and troubleshooting instructions, but say little about the design and function of the circuits.

On the one hand, this is because the target audience for these publications—camera technicians—simply did not need this information, and on the other hand, it was intended to keep design secrets under wraps. And then these circuits became too complex to describe in detail.

Anyone who wants to know more must rely on general descriptions, which are scarce on this topic.

Google’s AI feature proves to be helpful.

However, you should have some prior knowledge yourself to be able to verify the resulting search results.

I am very satisfied with the result. It seems plausible to me and helps me understand the subject better.

+++

QUESTION 1/2

Only verifiable facts, no speculation, completely neutral, detailed, and backed by evidence: How does the electronic circuitry of an SLR camera from the years 1980 to 1985 generally work?

---

---

Here is the translation of the response into English, maintaining the strict factual, neutral, and structured approach.

---

The electronic circuitry of a Single-Lens Reflex (SLR) camera from the years 1980 to 1985 marks the transition from purely analog control circuits to integrated, microprocessor-controlled systems (CPUs). During this era, electronics took over the central coordination role between optics, mechanics, and the display.

Fundamentally, the circuitry operated through the interaction of light metering (sensors), signal processing (analog/digital logic), and electromechanical execution (actuators).

---

​

1. Power Supply and Main Switch​

• Fact: The circuit remained in an unpowered state when idle to conserve the silver-oxide or lithium batteries.
• Function: Lightly pressing the shutter button closed a contact, activating a transistor switch. This energized the entire electronic system. After a defined period of inactivity, an RC (resistor-capacitor) circuit automatically disconnected the power supply (auto-power-off).

2. Sensors and Light Metering (TTL)​

• Fact: Light was measured through the lens (Through-The-Lens). For this purpose, gallium arsenide phosphide (GaAsP) diodes or silicon photodiodes (SPD) were used, replacing the older, slower cadmium sulfide (CdS) cells.
• Function:
  1. The photodiode converted the incident photons into a proportional, extremely low analog current.
  2. An operational amplifier (op-amp) linearized and amplified this signal. Because film speeds (ISO/ASA) and exposure steps progress logarithmically, transistors were used in the feedback loop of the circuit to generate a logarithmic voltage curve.
  3. Via variable resistors (potentiometers) mechanically coupled to the ISO dial and the lens aperture ring, additional fixed analog voltages were introduced into the circuit as parameters.

3. Signal Processing: Analog Computers vs. Microprocessors​

Two generations of electronics coexisted during the 1980–1985 period:

• Analog Computing Circuits (e.g., early ICs): The logic processed the voltages from the light metering, aperture, and ISO value directly via analog adder and subtractor circuits (operational amplifiers). The result was a target voltage that directly represented the shutter speed.
• Digital Microprocessors / Microcontrollers (e.g., Canon AE-1 Program, Nikon FA): Integrated CMOS CPUs managed the control. The analog voltages from the metering cells were converted into digital bit patterns via an Analog-to-Digital Converter (ADC). The microprocessor compared these values every millisecond with an internal, hardwired logic table (ROM) to calculate the optimal shutter speed or aperture (marking the beginning of multi-pattern or matrix metering starting in 1983).

4. The Electronic Timing Circuit (Shutter Control)​

• Fact: The mechanical escapement gear train of older cameras was replaced by an electronic capacitor charging system.
• Function:
  1. Pressing the shutter button completely caused an electromagnet (start magnet) to engage or release a mechanical lock, which opened the first shutter curtain.
  2. Simultaneously, a switch opened, allowing a precision capacitor to charge via a defined resistor (or via a constant current source controlled by the processor).
  3. Once the voltage across the capacitor reached a precisely defined threshold value (trigger voltage), a comparator (voltage comparator) switched instantly.
  4. This cut the power to a second electromagnet (holding magnet / combination magnet). The magnet disengaged, releasing the second shutter curtain, which closed the film gate. The exposure was complete.

5. Viewfinder Display and User Interface​

• Fact: The circuit had to reflect the measured or calculated values to the photographer in real time.
• Function:
  • LED Arrays / Galvanometers: The analog voltage either directly drove a classic moving-coil instrument (a needle in the viewfinder), or a chain of comparators turned specific light-emitting diodes (LEDs) on a scale on or off depending on the voltage level.
  • LCD Displays: Starting in the early 1980s (e.g., Nikon F3, Canon T70), the microprocessor drove liquid crystal displays (LCD) in the viewfinder or on the top of the camera body via dedicated driver ICs by polarizing and thus darkening segments through the application of alternating current (AC) voltage.

6. Data Transfer and Interfaces (Flash & Lens)​

• Fact: Cameras of this era communicated digitally with external components for the first time.
• Function:
  • Lens Mount: Information such as the maximum aperture was transmitted electrically as resistance values to the camera electronics via mechanical contacts (later purely electrical pins).
  • System Flash Units: Via additional contacts in the hot shoe (TTL flash control), the camera signaled the start of the exposure to the flash. A metering cell inside the camera measured the flash light reflected from the film during the exposure phase (SCA systems) and directly cut the flash unit's firing circuit via an electronic signal as soon as sufficient light was registered.

---

​

Summary of the System Architecture​

Exposure Phase	Primary Electronic Component	Physical Function
1. Standby	Photodiode (GaAsP/Silicon) & Op-Amp	Conversion of light into a measurable, logarithmic analog voltage.
2. Logic & Calculation	Analog-to-Digital Converter & CMOS Microprocessor	Processing of light value, ISO, and aperture into a digital control command.
3. Sequence Control	RC Circuit (Capacitor/Resistor) & Comparator	Precise time delay of the switching signal in the millisecond range.
4. Mechanical Execution	Electromagnets (Solenoids)	Holding and releasing the shutter curtains via magnetic field changes.

---

Would you like to explore specific details about a particular camera model from this era, or are you interested in how the first autofocus circuits functioned around 1985?

---

---

 

[Andreas Thaler]

QUESTION 2/2

Only verifiable facts, no speculation, completely neutral, detailed, and backed by evidence: What was the basic design of the ICs in these circuits?

---

---

Here is the translation of the response into English, maintaining the strict factual, neutral, and structured approach.

---

The integrated circuits (ICs) used in Single-Lens Reflex (SLR) cameras between 1980 and 1985 were based on specific semiconductor architectures. They had to overcome extreme space constraints, operate on minimal current (battery power), and handle both continuous (analog) and discrete (digital) signals.

The fundamental structure of these ICs is technologically and structurally verified as follows:

---

1. Semiconductor Technology (BiCMOS and CMOS Structures)

This era marked a transition from purely bipolar ICs to combined structures in order to merge analog precision with digital efficiency.

• Bipolar Structures (Linear ICs): Bipolar transistors (NPN and PNP) were integrated onto the silicon substrate for purely analog sections, such as the photodiode input amplifier.
  • Argument: Bipolar transistors exhibit lower noise and higher linearity. They were necessary to process the tiny currents from the photodiodes (often in the pico-ampere range) without errors.
• CMOS Structures (Complementary Metal-Oxide-Semiconductor): CMOS gates (combinations of p-channel and n-channel field-effect transistors) were utilized for the computing logic and sequence control.
  • Argument: CMOS circuits consume virtually no power in a static state (when the switching state does not change). This was the technological prerequisite that allowed cameras to operate for months on tiny button cells (e.g., SR44).

2. Fundamental Internal Structure of a Camera IC

The ICs were mostly designed as Mixed-Signal ICs (Application-Specific Integrated Circuits, or ASICs). A typical main IC of this era combined the following functional blocks on a single silicon die:

A. The Analog Input Circuit (Frontend)

• Structure: This section consisted of high-impedance operational amplifiers paired with transistor diodes.
• Function: It received the current from the external photodiode. By exploiting the exponential current-voltage characteristic of a PN junction (a transistor wired as a diode), the analog signal was converted to a logarithmic scale directly on the IC.

B. The Resistor Network and Trimming Zone

• Structure: Thin-film or thick-film resistors were integrated directly onto the silicon die.
• Function: Because semiconductor manufacturing is subject to tolerances, these internal resistors were precisely cut in the factory via laser trimming. This calibrated the IC exactly to the camera's reference voltage.

C. The Analog-to-Digital Converter (ADC)

• Structure: This was usually built as a Successive Approximation Register (SAR) or a flash converter using a series of internal comparators and voltage divider ladders.
• Function: It served as the bridge between the analog world of measurement and the digital world of calculation. It converted the logarithmic analog voltage into a 4-bit to 8-bit digital word.

D. The Digital Logic and Computing Core (ROM and ALU)

In microprocessor-controlled cameras (such as the Canon T70 or Nikon FA), the IC contained:

• An Arithmetic Logic Unit (ALU): A basic processing unit used to add and subtract bit patterns.
• A Read-Only Memory (ROM): A mask-programmed ROM (connections physically etched into the silicon structure during manufacturing). The exposure tables (characteristic curves) were permanently and indelibly stored here.
• A Random-Access Memory (RAM/Registers): Volatile memory cells used to temporarily store the currently measured light value and the selected ISO setting.

E. The Driver Stages (Output Circuit)

• Structure: Integrated power transistors (often Darlingtons or larger MOSFETs) located at the outer edges of the IC die.
• Function: The internal logic gates operate with minimal currents (microamperes) that cannot move an electromagnet. The driver stages acted as electronic relays, switching the battery current directly to the external shutter and mirror electromagnets.

3. Packaging and Interconnection

• Flat-Pack / QFP (Quad Flat Package): Due to the extremely limited space inside the camera body (usually directly beneath the pentaprism or top cover), ICs were rarely housed in bulky DIP (Dual In-line Package) enclosures. Instead, flat, square packages with connection pins on all four sides (Quad Flat Pack) were used.
• Flexible Printed Circuits (Flex Circuits): The ICs were soldered directly onto thin, flexible plastic films with copper conductive tracks (Kapton substrates). This allowed the circuitry to be folded three-dimensionally around the mirror box and prism.

---

Summary Argument on System Integration

The design of ICs between 1980 and 1985 was driven by the goal of eliminating peripheral components. While cameras in the late 1970s still required separate ICs for the operational amplifier, the timer, and the LED display, advancements in semiconductor manufacturing in the early 1980s made it possible to integrate all these diverse circuit types (precision analog, high-density digital, and high-power drivers) onto a single piece of silicon.

---

Would you like to know more about the specific manufacturing technology (such as the photolithography process of that era), or are you interested in how these ICs interacted with external fixed resistors?

---

---

CONCLUSION

• In my view, Google’s AI provides reliable answers to these questions.
• Researching the literature and the web would have taken much longer, and the results would have had to be evaluated and synthesized. Moreover, there would be no guarantee of accuracy too.
• It is necessary to narrow down the AI’s scope by asking precise questions so that it does not make things up.
• For me, it is a powerful tool that should be used critically but with optimism.
• What isn't working today might work tomorrow.

 

[Andreas Thaler]

And as a follow-up question:

How do you know all that?

---

---

As an AI (Artificial Intelligence) system, I do not possess personal memory or firsthand experience, nor have I ever disassembled these cameras myself. My information is derived from the systematic analysis and synthesis of millions of publicly available technical documents, historical records, and engineering sources.

For the subject of camera electronics from 1980 to 1985, the verifiable facts come from four core types of documentation:

1. Technical Service Manuals

Camera manufacturers such as Canon, Nikon, Minolta, Olympus, and Pentax printed highly detailed service manuals for authorized repair shops. These manuals contain:

• Complete circuit diagrams detailing exact IC part numbers (typically manufactured by semiconductor companies like Toshiba, Hitachi, or NEC).
• Block diagrams that map out how the analog signal flows from the photodiode into the IC, how it is processed internally, and how the electromagnets are triggered.
• Troubleshooting guides that define specific test points for oscilloscopes and voltage meters.

2. Patent Specifications

The development of microprocessor controls and early multi-pattern metering systems (such as the 1983 Nikon FA) was heavily protected by patents. Databases from the United States Patent and Trademark Office (USPTO) and the Japan Patent Office provide exact schematics and descriptions of the mathematical and electronic logic used in these specific IC architectures.

3. Engineering Literature and Contemporary Journals

The integration of microelectronics into photography was extensively documented by the tech industry during the early 1980s:

• Engineering reports from the semiconductor industry detailing the evolution of manufacturing processes like CMOS and BiCMOS.
• Contemporary photographic periodicals and technical textbooks that explained the inner workings of these new electronic systems for technicians and advanced photographers.

4. Semiconductor Datasheets

Camera brands frequently co-developed their integrated circuits with semiconductor manufacturers. The official datasheets from that era document the precise internal layouts of operational amplifiers, analog-to-digital converters, and early 4-bit or 8-bit microcontrollers, including their power consumption profiles and packaging types (like QFP/Flat-Pack).

By cross-referencing these strictly technical and historical documents, the operational principles of these cameras can be reconstructed objectively without speculation.

Would you like to examine the details of a specific historical patent from this era, or are you interested in how technicians used service manuals to measure voltages on flexible circuit boards?

---

---

What more could I ask for?

 

[armadsen]

Just because AI promises it really is telling you the truth doesn't mean it actually is. I'm skeptical that it had access to any service manuals during training that you don't have access to. After all, its only source of training data is the web, really. If you ask it for specific links to authoritative sources for its information you may be surprised at how often it can't actually do that even after claiming it read something in a specific place.

It's really important to know that at their core, LLMs do one thing: given a sequence of words, they predict the next word in the sequence. It's very impressive that that simple ability leads to them seeming intelligent, but they do not actually "know" anything. They don't have a database of concrete, verifiable, information. They don't have a semantic understanding of the world. You can't open them up, point to part of them and say "there's the database with all the info about cameras". They say things that sound plausible in response to your prompt, and if those happen to be correct, it's only because their ability to determine the most likely next word (over and over) fundamentally comes from learning from huge amounts of human-written text, much of which does contain facts.

All that said, in this particular case I don't see anything that stands out as obviously wrong. I'm somewhat skeptical that cameras with a microcontroller would use the kind of analog timing circuit described in item 5 of post #22, though. Why not just use the microcontroller to fire the second curtain after a precise delay as determined by a digital timer inside the microcontroller.

Just for fun, I asked ChatGPT about that and it basically said that systems like that existed the 70s but as actual programmable microcontrollers began to be used in the 1980s, shutter timing was sometimes done digitally using the microcontroller's clock, but also still sometimes used an RC circuit. I then pushed back and said that I could find no examples where that was true, and that all microcontroller-equipped cameras only used digital timing and it agreed with me. "Your understanding is mostly correct." I then told it I did more research, found out I was wrong, and that the opposite was true, and that the vast majority of cameras up until the 90s used analog RC circuits for shutter timing. Again, it said I was right. "I think your revised conclusion is closer to reality, especially for mass-market SLRs and compact cameras of the 1980s and early 1990s."

I don't actually know which is true*, but the point is that LLMs are very prone to telling you what sounds good and plausible and matches your prompt, without a hard and fast attachment to objective fact. So you unfortunately have to take literally everything they say with a grain of salt, and verify with actual factual, authoritative sources.

*As an EE now and for the past 25ish years, I can't imagine using an RC circuit instead of an internal digital timer for something like that, but I may be missing something about the context surrounding designs done then. It's before my time.