A microcontroller made from 555 timers

[tggzzz]

Using a MCU to implement a 555 is one of those topics guaranteed to have greybeards muttering "...youngsters..." - even though with exceedingly cheap/small MCUs it isn't quite as stupid as it appears.

But what about implementing an MCU using 555s? One of 0.1% of really interesting Hackaday articles points to a project by Tim Böscke.

My TL;DR is:

• a decade ago he implemented a CPU in a 32 cell CPLD, with documentation at https://github.com/cpldcpu/MCPU
• that MCU definition has 4 instructions, 8-bit datapath and a 6-bit address
• he is reimplementating that using 555s, as shown in the project logs
• so far he has simulated it in LTSpice, and has an interesting and sensible workflow

Unfortunately the first Hackaday comment beat me to it: "Can you emulate a 555 with it?"

As an aside, this post is a neat example of the limitations of hierarchical taxonomies. Should it be in "Projects, Designs, and Technical Stuff" (no because it isn't my project), or "Microcontrollers" (no because you can't buy/use it), or "Open Source Hardware" (because it is), or "PCB/EDA/CAD" (because of his interesting use of workflow and tools)?
Basically taxonomies are basically useful for indicating where there no point in looking for a topic, e.g. "Jobs" in this case. The corollory is that sweating over finely crafting/tuning/changing hierarchies is a great waste of time.

But that's a different topic (which might belong in "News/Suggestions/Help").

[jpanhalt]

High Tech Chips,Inc used an MCU to emulate many of the functions of the 555.  Part numbers were HTC555, HTC550, HTC750, and HTC751.  That was done around Y2K.

Its original site: hightechips.com seems to be in the gaming business now.  Anyway, just for reference, I have attached a pdf of its datasheet for a multivibrator (HTC550).

Of course, with the 12F series PIC's, ground (pin-8) and Vcc (pin1) are different from the 555.

[rpiloverbd]

Hackaday never ceases to surprise me. Thanks for sharing the article.

[ataradov]

I personally did not find it interesting at all. You can create a NAND gate from a lot of things and it is obvious you can create computing devices from NAND gates.

I found computers built from individual transistors to be far more interesting, since they scale up and illustrate something that actually happens (or happened) in real life.

[SiliconWizard]

Indeed, what the author did was basically get NAND and NOR gates out of a NE555, at least from the short part I read. Of course from then on, everything follows. You can implement almost anything using NAND and NOR gates. Using a NE555 for this just looks like one of those fun tricks leveraging some fetishism for popular old chips.

Likewise, you can implement "linear" amplifiers with CMOS gates. That's cute too. (But it could have some practical applications for lowering costs if you have one or more spare gates in a given design!)

But hey, whatever tickles anyone's fancy.

[magic]

I think quad NAND should be doable with TL074 by exploiting its phase reversal.
Unfortunately external resistors are still needed, to guarantee that IN+<IN- when both inputs are driven high.

[tggzzz]

Indeed, what the author did was basically get NAND and NOR gates out of a NE555, at least from the short part I read. Of course from then on, everything follows. You can implement almost anything using NAND and NOR gates. Using a NE555 for this just looks like one of those fun tricks leveraging some fetishism for popular old chips.

Likewise, you can implement "linear" amplifiers with CMOS gates. That's cute too. (But it could have some practical applications for lowering costs if you have one or more spare gates in a given design!)

But hey, whatever tickles anyone's fancy.

Of course!

Inverting the "use an MCU to replace a 555" trope does tickle my fancy

[T3sl4co1l]

Yes, would be much more interesting if it took advantage of the more analog properties of the device.  Example: if input signals are well buffered, then summing them together with resistors gets you an analog level, which compares to the internal threshold, and there you go, a COUNT(N; a_1, a_2, ..., a_n) operator.  Which, if the threshold is N=1, you get a standard (N)OR gate; or N=n, (N)AND.

Would also like to see tooling to accompany the architecture -- four measly instructions is the most boring thing I can think to assemble.  Doing basic arithmetic and flow control would be nice; maybe implementing a VM with a nicer architecture, albeit even slower in operation.

Or, another interesting question might be: at the expense of adding one or a few instructions, or lengthening instructions (say to allow more memory space, or various addressing ranges, or modes even), how much nicer/easier/faster does that next level VM get?

Ultimate question, for any combination of HW/SW architecture: how long until it runs C?   (After which, Doom is only a few keystrokes away, natch.)

Tim

[tggzzz]

Yes, would be much more interesting if it took advantage of the more analog properties of the device.  Example: if input signals are well buffered, then summing them together with resistors gets you an analog level, which compares to the internal threshold, and there you go, a COUNT(N; a_1, a_2, ..., a_n) operator.  Which, if the threshold is N=1, you get a standard (N)OR gate; or N=n, (N)AND.

Would also like to see tooling to accompany the architecture -- four measly instructions is the most boring thing I can think to assemble.  Doing basic arithmetic and flow control would be nice; maybe implementing a VM with a nicer architecture, albeit even slower in operation.

Or, another interesting question might be: at the expense of adding one or a few instructions, or lengthening instructions (say to allow more memory space, or various addressing ranges, or modes even), how much nicer/easier/faster does that next level VM get?

Ultimate question, for any combination of HW/SW architecture: how long until it runs C?   (After which, Doom is only a few keystrokes away, natch.)

Tim

It has three more instructions than necessary to be Turing complete, I believe.
It has a datapath 8 times wider than the Motorola 14500.

That should be enough for anybody

[SiliconWizard]

Indeed, what the author did was basically get NAND and NOR gates out of a NE555, at least from the short part I read. Of course from then on, everything follows. You can implement almost anything using NAND and NOR gates. Using a NE555 for this just looks like one of those fun tricks leveraging some fetishism for popular old chips.

Likewise, you can implement "linear" amplifiers with CMOS gates. That's cute too. (But it could have some practical applications for lowering costs if you have one or more spare gates in a given design!)

But hey, whatever tickles anyone's fancy.

Of course!

Inverting the "use an MCU to replace a 555" trope does tickle my fancy

Well ok, it's kind of "Revenge of the NE555".

But yes, it would have been more interesting using the NE555 in a more clever way than just making it act as a logic gate.

[Zero999]

Likewise, you can implement "linear" amplifiers with CMOS gates. That's cute too. (But it could have some practical applications for lowering costs if you have one or more spare gates in a given design!)

But hey, whatever tickles anyone's fancy.

I made an analogue comparator with a couple of CMOS gates awhile ago. It worked, but was obviously nowhere near as good as the LM393.