Logic, Before ICs | Hacker News Reader

Logic, Before ICs

northcoastsynthesis.com

124 pointstodsacerdoti27 comments

Nice to see RTL (resistor transistor logic) circuits again! They were my introduction to digital logic back in the mid-1960s.

I was getting into ham radio at the time, and my neighbor across the street was a sales engineer with Motorola. He saw me putting up an antenna and said "hey, it looks like you're into electronics, let me know if you can ever use some free samples of our integrated circuits. Here's a databook to get you started."

I really wanted to make an iambic Morse code keyer. This is the kind that has a paddle with two independent levers, one for dits and the other for dahs, and if you squeeze them both at the same time it alternates dits and dahs for you - thus the name "iambic" [1].

So I designed a circuit using the RTL chips, gave my neighbor the parts list, and a week later got a nice box with the parts I needed. Put together the circuit and it worked!

Of course I didn't know anything about RF shielding, and RTL circuits were particularly susceptible to interference. So when I hooked the keyer up to my transmitter and tried making some contacts, it mostly worked, but sometimes it would just start generating dits and dahs all on its own because the transmitter would feed back into my iambic circuit.

Naturally the hams I contacted started laughing at me (as much as you can laugh in Morse code) and said "what a lid!" [2]

It was fun though.

[1] "a line of verse with five metrical feet, each consisting of one short (or unstressed) syllable followed by one long (or stressed) syllable": https://en.wikipedia.org/wiki/Iambic_pentameter#Example

[2] https://boards.straightdope.com/sdmb/showthread.php?t=324119

RTL just looks horrifying to me. I had one of those Radio Shack 300 in One electronics kits, but most of the projects were analog and poorly explained. It wasn't until years later, when I saw the simple design of MOSFET logic gates, that my view of computers shifted from "sorcery" to "math".

Granted, with modern transistors getting progressively smaller, computers seem to be moving back to analog sorcery again.

You had the 300-in-one kit?! I had one of the much older Radio Shack kits, and have many fond memories of those spring terminals.

It looked a lot like this 160-in-one:

https://medium.com/@rxseger/electronic-project-kits-hands-on...

http://www.zpag.net/Electroniques/Kit/160_manual.pdf

But that lists a date of 1982. I am sure mine was many years before that. Of course it could have been in production long before 1982 - the kit definitely doesn't have an "80s" look.

I wonder how hard it would be today to build a robotic assembler of large-scale devices on discrete transistors which would take a lot of time and effort to build manually, like the "monster 6502": https://monster6502.com/.

It's called a "pick-and-place machine", and the Monster 6502 was likely assembled using one.

https://en.wikipedia.org/wiki/Pick-and-place_machine

I've waited for monster 6502 to become available. Does anyone know if it is available for purchase?

Can I buy one?

Hopefully in the near future! We are working towards a public launch now; please Sign up for our mailing list if you would like to receive updates as we get further along.

This is what was taught to me in high-school, along with analog electronics, then moving on to 7400-series ICs. I like having been exposed to electronics/computers at many different levels of detail. I couldn't say, make a CPU but have broad-stroke understanding up to software.

I find it incredibly interesting that a Rust 'move' could be implemented as an exchange operation with a thermodynamic lower energy limit of zero.

Once for kicks I made a 555 oscillator out of discrete parts. I was looking at the internal circuit diagram one day and realized I had all the parts handy: https://en.wikipedia.org/wiki/555_timer_IC

You make 1/3 and 2/3 voltage references with three equal resistors in series as a voltage divider and feed those into a pair of op amps that have their other inputs ganged together, the outputs of the op amps go to a flip-flop made of transistors (a 1-bit memory), which controls the filling/emptying of a capacitor (external to the IC) that sets the other (ganged) input to the op amps.

The Exploratorium had a physical pneumatic oscillator that worked in a very similar way. A knob on a sliding rod would trip switches at either end of its slide that toggled the left-or-right sliding of the bar by altering the path of the air (from a compressor under the table.)

Nice, but I wouldn't call an op amp a discrete component ;)

They might have built the op-amps out of discrete components, though. Never says they used an IC package. ;)

Ha! No, it's a fair point. I did look at a schematic for an op amp. And then I used half of a LM324 IC. ;-)

My analogue design engineer colleague still uses this approach. One rationale I have heard him say is that he prefers to do things with transistors instead of logic ICs because of the direct availability of transistors, compared to ICs. It's an advantage when doing repair. You always have transistors laying around.

However, when you need to repair ICs on a PCB, you need a specific IC doing your function, which you may not have in stock. I wondering why the argument doesn't apply for transistors. The assumption seems to be that it doesn't matter much which transistors are used.

I don't know if anyone can validate what my colleague is saying.

shrug it's fine if you like doing it that way, you're doing a small set of functions, you don't mind the parasitic capacitance, and board space and power consumption is not an issue. And desk space: https://www.edn.com/jim-williams-desk-circa-2007/ (if it was good enough for Jim Williams, it's good enough for you)

> The assumption seems to be that it doesn't matter much which transistors are used.

Weeell .. sort of, in that the variance is so high that for most designs the parameters are designed to be irrelevant. Provided you get current capacity right. Sometimes you genuinely need matched transistors though.

(One of the classic synths relied on a specific batch of "faulty" transistors, which made its properties almost unreproducible until the full-digital era)

I get the idea, and have worked with people who hold that kind of opinion. But it's getting more and more untenable, if you want to have competitive products. It's the same why you generally just can't use leaded parts anymore - almost everything is surface mount.

ICs are doing so much more these days, that it would just take up too much space, or need whole extra boards to do what a tiny QFN part does. In terms of repair, even the individual transistors (of the few we use these days) are so small that field repair is extremely difficult - it's hard to replace them without a proper lab setup, so we mostly have spares of whole boards or modules for field repairs, and then send the faulty board back to the lab where somebody can repair it with a microscope, hot air rework station etc.

If you liked this article, I believe you might find this game interesting (if you find a browser with Flash still working): http://www.zachtronics.com/kohctpyktop-engineer-of-the-peopl... (no relation, I just liked to play it a few years ago, and learned quite a bit by playing it)

This might be worth someone taking some time to figure out how to make it work with Ruffle: https://github.com/ruffle-rs/ruffle

Discrete logic is still around and new series have been introduced even relatively recently, for example single-gate ICs the size of a single transistor, for when you "really just need that one NAND" on the board.

And in some cases, smaller than a transistor (as commonly packaged).

74AUP1T97 comes in a variety of packages, some less than 1mm square. It’s a clever mix of gates internally which can be turned into many common AND/OR type functions depending on how you hook up the pins.

There's actually five of them, not just one: the ’57, ’58, ’97/’157, ’98/’158, and ’99. (Plus four more if you count the open-drain-output types released only by Fairchild, but I don't as they're too annoying to source reliably.)

These guys are super useful. Unfortunately it's somewhat unobvious how best to use them, since the manufacturer literature is written in a very obtuse way. Except for the ’99, they're all just two-input multiplexers. Some have an input inverted and some have an output inverted. The ’99 adds an output enable line and a fourth input, which it XORs with the output. It's much easier to design with these things when you think of them as multiplexers!

I've made a nice chart that I use to help design with these parts, but it's on my work machine and I'm too lazy to log in on a weekend to retrieve it. I've made a lot of nice charts and references... I should figure out somewhere to publish them....

Page 5 of this old NXP PDF was the starting point I used:

https://www.avnet.com/wps/wcm/connect/onesite/8100cb9c-39d1-...

but I removed the redundant entries (most of us know how logic families work by the time this chart is useful...) and added the descriptions of what the parts really are inside.

Yes, thanks for expanding on that. I used the ‘97 most recently so it was the part number I remembered.

The open drain ones make good level translators as well, since the output can be pulled up to a different voltage. If you find that chart, I’d love to see it!

Let's try this: https://ibb.co/HDsYHFP

(I'm not sure about that image host, but a screenshot seemed the easiest way to get it free of the corporate G Suite, with no-strings-attached absolutely guaranteed.)

There's more than just the five gates previously discussed because this was supposed to be a general list of all multifunction gates: that is, if you need to implement anything other than the obvious functions, start here first.

There were also three notes attached to the columns:

1. XNOR = 2-XOR with 1 inverted input = A XOR !B

2. A | !B = 2-OR with 1 inverted input = 2-NAND with 1 inverted input: A | !B = !(!A & B)

3. A & !B = 2-AND with 1 inverted input = 2-NOR with 1 inverted input: A & !B = !(!A | B)

I think the exclusive, if both inputs are zero the output should be zero, but it look like the output will be a one in that situation.

As far as I can see only XNOR is depicted and it is done so correctly. XNOR is just XOR followed by a not, which does give a 1 when the input is two zeros.

If the inputs are equal, low/low or high/high, the transistors are closed and the resistor will pull the output high.

If one input is low and the other high one of the transistors will open and pull the output low.

Great article with good explanations. Almost made me miss the good old transistor days.

The assembly language of the hardware world

Reply on news.ycombinator.com