Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

A modern 8 bit design, built using 1950s thermionic valves

valve.computer

Post reply on HN
Print view View original

61–70 of 87 posts

Re: A modern 8 bit design, built using 1950s thermionic valves

#61

Post by kazinator »

tass wrote: post #50
kazinator wrote: post #49

Earlier quoted context omitted.

What? Silicone hardware generates heat. Neither silicone hardware nor tube hardware goes literally cold in a cold boot; just the power is cut for a moment, usually not long enough to lose a lot of heat.

I imagine these terms could have changed over time, but consider that a warm boot might have meant power off and on with these old computers. One difference is that the valves have more chance of burning out on a cold boot due to the temperature change.

Filaments experience a high inrush current on startup because their resistance is lower when they are cold. That's why incandescent bulbs often burn out on powerup, rather than in the middle of a duty cycle.

Tube filaments last a good long time though, particularly in small signal tubes.

They do not shine like light bulb filaments; they glow red.

The circuit designer also has a say in it; tube filaments can be operated over a range of currents: you can run tubes hotter or colder. In a digital application, you'd probably want to go as cold as you can get away with for longer life.

Tubes have various modes of failure in addition to burned out filaments, like "gassing out", or the cathode emission decreasing, eroding the gain. Of course, abnormal conditions like parts melting from overcurrent.

Speaking of incandescent bulbs: those are victims of planned obsolescence. A bulb can easily be made that will last 50 years; it's just not profitable because you can't charge 50 times more for it than one that burns out in a year. Once everyone has 50 year bulbs, you're out of business.

Re: A modern 8 bit design, built using 1950s thermionic valves

#63

Post by kazinator »

cjs_ac wrote: post #57
kazinator wrote: post #49

Earlier quoted context omitted.

What? Silicone hardware generates heat. Neither silicone hardware nor tube hardware goes literally cold in a cold boot; just the power is cut for a moment, usually not long enough to lose a lot of heat.

The difference is that for solid-state logic, the heat is purely a waste byproduct, whereas for thermionic valves, the heat is the energy used by the electrons to move between the electrodes, hence the name. If the valves are cold, you have to wait for them to warm up before they can do anything.

All the same, "cold boot" doesn't refer to letting tubes cool down. If power is restored while the tubes are still hot, it's still a "cold boot".

The electrons are not moved by heat; they are moved by the electric potential between the electrodes. The heat goes into maintaining the cathode at a certain temperature, at which the electrons "boil off" in sufficient numbers, but that requires only a smidgeon of the energy. Though depending on it, tubes are incredibly wasteful of heat.

So you say silicon and other parts don't require heat? Check that data sheet again! If the part's operating range is from -40°C to 70°C, then that part requires heat. If the ambient temperature happens to be -100°C, you have to heat it, perhaps by installing a heating filament into the chassis.

E.g. check the TI datasheet for the NE5532 op-amps:

Free air operating temperature:

NE5532, NE5532A: 0 to 70°C.

SA5532, SA5532A: –40 to 85°C.

(That standard part is quite delicate there, only down to zero!)

Conversely, thermionic tubes for a hot environment could be built without heating filaments. If the ambient temperature is 500°C, and that happens to be quite enough for that tube's cathode, then it Just Works, like your NE5532 IC at 3°C.

Re: A modern 8 bit design, built using 1950s thermionic valves

#64

Post by bregma »

hilbert42 wrote: post #42
doe_eyes wrote: post #17

Earlier quoted context omitted.

They are insulated really well - by vacuum! I'm actually surprised by the figure, though. A small tube requires about 300 mA at 6 V, and the trick is that you can connect the heaters in series instead of doing it all in parallel and pumping out a ton of amps at a very low voltage. They could've done 10 tubes in series at a reasonably safe 60 VDC, and they'd only need 20 amps. Back in that era, because both valves and…

"They could've done 10 tubes in series at a reasonably safe 60 VDC, and they'd only need 20 amps." Thermionic vacuum tubes of this type usually have a specified maximum heater/cathode voltage rating which varies considerably according to design. Exceeding that rating and one risks a short between the heater and cathode. For these types of tubes heaters can safely operate up to 200V negative with respect to the cathod…

When I was in high school (many decades ago now) we had a vacuum tube tester with a switch that allowed you to select the heater voltage. We quickly discovered, devils that we were, that if you stuck a simple diode with a 1.5 V heater in the tester and cranked the voltage to max it would launch like a little rocket. Eye protection recommended.

Re: A modern 8 bit design, built using 1950s thermionic valves

#65

Post by _joel »

JJMcJ wrote: post #37
ahazred8ta wrote: post #35

Earlier quoted context omitted.

♫ "And when they turn the power on, it's sure to dim the lamps / At plus and minus 16 volts and fourteen hundred amps." -- Frank Hayes

Telephone central exchanges used to have enormous lead acid battery (like automobile batteries, the starter motor not the EV kind) arrays if they lost power. It was amazing to see the battery array discharge ammeter at 5000 amps when utility power went out.

some ISP's still do (or did, last time I was in one), at least one I worked for a few years back. Had a room full of lead acid batteries on shelves and another with a big spinning wheel to smooth out the flow.

Re: A modern 8 bit design, built using 1950s thermionic valves

#66

Post by ekidd »

kazinator wrote: post #61
tass wrote: post #50

Earlier quoted context omitted.

I imagine these terms could have changed over time, but consider that a warm boot might have meant power off and on with these old computers. One difference is that the valves have more chance of burning out on a cold boot due to the temperature change.

Filaments experience a high inrush current on startup because their resistance is lower when they are cold. That's why incandescent bulbs often burn out on powerup, rather than in the middle of a duty cycle. Tube filaments last a good long time though, particularly in small signal tubes. They do not shine like light bulb filaments; they glow red. The circuit designer also has a say in it; tube filaments can be operat…

> Once everyone has 50 year bulbs, you're out of business.

This seems to have been the issue with the expensive Philips Hue LED bulbs (the ZigBee ones). We installed ours close to a decade ago, and we took them with us when we moved to a new house. I don't think I've replaced a single one yet.

Philips seems to have de-emphasized that set of bulbs in favor of cheaper models, as far as I can tell?

Re: A modern 8 bit design, built using 1950s thermionic valves

#67

Post by Max-q »

kazinator wrote: post #63
cjs_ac wrote: post #57

Earlier quoted context omitted.

The difference is that for solid-state logic, the heat is purely a waste byproduct, whereas for thermionic valves, the heat is the energy used by the electrons to move between the electrodes, hence the name. If the valves are cold, you have to wait for them to warm up before they can do anything.

All the same, "cold boot" doesn't refer to letting tubes cool down. If power is restored while the tubes are still hot, it's still a "cold boot". The electrons are not moved by heat; they are moved by the electric potential between the electrodes. The heat goes into maintaining the cathode at a certain temperature , at which the electrons "boil off" in sufficient numbers, but that requires only a smidgeon of the ener…

The valves need to be warmed up can still be the source of the names "cold boot" and "warm boot", even if we today have given them different meaning.

Re: A modern 8 bit design, built using 1950s thermionic valves

#68

Post by Max-q »

This is so interesting! Quite often, I'm thinking about how a modern design would perform with tubes. Back in the 40s and 50s, digital computers were new. Much of the design was inherited from tabulating machines. When I study the early IBM machines, they seem very complex. When we learned more during the 60s, and ended up with microprocessors in the 70s, the design had been optimized a lot. So I have been wondering how a "modern" design would perform on tubes.

This guy actually did it. What a fun project it must have been. Super interesting read.

One detail in the introduction that made me a bit puzzled: "The Valve.Computer is an 8 bit computer, with the usual 12 bit address and data buses". In an 8 bit CPU, we had "the usual 16 bit address bus and the 8 bit address bus".

Re: A modern 8 bit design, built using 1950s thermionic valves

#70

Post by 082349872349872 »

kazinator wrote: post #58
082349872349872 wrote: post #53

Earlier quoted context omitted.

First boot in the morning was cold, and —the hardware having literally been cold— took minutes longer than subsequent, warm, boots.

That wouldn't be on account of the tubes; tube filaments do not take minutes to heat up the cathodes. Warm boots can save time by not repeating all the boostrapping steps. In a warm boot, you typically don't execute any power-on self-checks, for starters.

Maybe I'm misremembering (I was very young) or maybe I was with people who were too superstitious, but these tubes themselves were actually from the 1950s, and they did indeed take a significant amount of time before they decided the hardware was "warm enough" that one could let the software start doing its thing.
Post reply on HN