Return of the Vacuum Tube

"Electrons move more slowly in a solid than in a vacuum, which means transistors are generally slower than vacuum tubes; as a result, computing isn't as quick as it could be."

Oh damn, I just sold my tube tester on ebay.

That's why those old Univacs were so fast.

--

John Larkin                  Highland Technology Inc
www.highlandtechnology.com   jlarkin at highlandtechnology dot com   

Precision electronic instrumentation
Picosecond-resolution Digital Delay and Pulse generators
Custom timing and laser controllers
Photonics and fiberoptic TTL data links
VME  analog, thermocouple, LVDT, synchro, tachometer
Multichannel arbitrary waveform generators

I think I had once calculated the ballistic delay of something like 6AQ5 to be 80ns. This isn't the ultimate limitation, velocity modulation is the effect that smears your wave out; but it will be on the same order.

Considering 2N3904 goes full off to full on in less time, you'll have a hard time competing with that. The electron drift velocity might be ~m/s, but when you have a sea of electrons already available and micrometers to cross, you kind of don't care. High mobility semiconductors just get better from there.

Obviously, none have the same raw capacity for velocity as vacuum does: 300V ~= 1e7 m/s final velocity, whereas most semis have thermal drift in the 1e5 range.

Tim

--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms

"bitrex"  wrote in message 
news:ofOdnV0og5fVPyDSnZ2dnUVZ_jKdnZ2d@earthlink.com...
> On 5/23/2012 10:44 PM, Chris wrote:
>>
http://news.sciencemag.org/sciencenow/2012/05/return-of-the-vacuum-tube.html?ref=hp
>>
>>
>> Chris.
>>
>
> "Electrons move more slowly in a solid than in a vacuum, which means 
> transistors are generally slower than vacuum tubes; as a result, computing 
> isn't as quick as it could be."

I recall there being some interaction between AC grid voltage and electron velocity that results in a resistive grid impedance component at high frequencies, which hurts gain. Something like that.

There was a lot of talk, about a decade or two back, about planar vacuum tubes that had MEMS cathode-grid structures: an array of microtips, the cathode, under a plate full of holes, the grid. The spacings would be microns, and a positive grid voltage would cause field emission from the microtips, out into the vacuum gap to the plate. It would make a great RF power amp. A variant used carbon nanotubes (of course.) I think they sort of worked, but the microtips degraded rapidly. Nanotubes are great if noisy electron emitters, but they also degrade quickly.

I recall a semiconductor diode that flung electrons into space, simulating a cathode-grid structure.

I invented a tiny cold cathode for use as an electron source for a really cheap SEM, which actually might have worked, but I couldn't get anyone interested in it.

Toobs are cool. I hope they come back somehow.

I think the correct technical term is "fire-bottle".

Yeah. I'd like to see a 2 billion tube CPU.

The length of the electron path IS ALSO a factor, eh?

Simple tubes are basically high impedance devices. in which even a small stray capacitance will kill the switching speeds. Reducing the impedance levels (to 50-200 ohms) would also help.

Devices switching electron beams might be used as logical elements.

Taking it a bit further, a traveling wave tube might be modified to act as a logical element.

Well, there were small versions made for space considerations (Nuvistors) as well for satellites / radiation resistant service (Timms if i remember correctly). The baby bird (goo gull) seems to draw a blank on "timms"..

No reason an even more compact version cannot be made..AND that the electron emitter could be a solid state junction device (no Phil A. Mentt).

They sure beat the heck out of the older solid-state computing devices..

The delay was one of the prime reasons in developing the external-anode coaxial power tetrodes (4x150A and descendants) to get rasonable power at VHF and low UHF.

To get any useful gain on higher frequencies, more exotic tubes are needed, like traveling-wave tubes and klystrons.

After using tubes for decades, I do not miss them at all.

--

Tauno Voipio

Never fear - vacuum tubes may stage a come-back after all... as EMP-proof ultra-high power switching devices that can handle MV and kA to MA - power levels that are orders of magnitude higher than solid state technologies. This new breed of high power device uses high-current field emission cathodes instead of thermionic emission. Interesting, potentially game-changing technology!

Bert

--
Bert Hickman
Stoneridge Engineering
http://www.capturedlightning.com***********************************************************************
World's source for "Captured Lightning" Lichtenberg Figure sculptures,
magnetically "shrunken" coins, and scarce/out of print technical books
***********************************************************************

Transit time effect? The electron stream shields the grid-plate capacitance some, which mades it depend on the plate current. Transit time phase-shifts the effect, giving rise to a resistive term in parallel with the feedback capacitance.

(I agree--tubes are neat. I started building stuff with them at about

1. They sometimes worked, for sufficiently generous definitions of "worked.")

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510
845-480-2058

hobbs at electrooptical dot net
http://electrooptical.net

In my very first job interview, with a priggish guy in the Tulane physics electronics shop, I said that I preferred tubes to transistors because tubes were harder to blow up. He said "that won't do" and dismissed me. Well, tubes were free, and transistors were expensive, and germanium transistors were fragile.

About a week later, I had another interview, and I said the same thing. The guy laughed and hired me. I designed about $200e6 worth of stuff for him.

--

John Larkin         Highland Technology, Inc

jlarkin at highlandtechnology dot com
http://www.highlandtechnology.com

Precision electronic instrumentation
Picosecond-resolution Digital Delay and Pulse generators
Custom laser drivers and controllers
Photonics and fiberoptic TTL data links
VME thermocouple, LVDT, synchro   acquisition and simulation

Nice of interviewer #1 to self-identify as a jerk--saved you the trouble of telling him, 3 months into the job, which is always awkward.

My #1 daughter, who until recently had an apartment on Dauphine Street in the Vieux Carre', says that a lot of the jerkier male tourists perform a similar service to women by wearing misogynistic tee shirts. If they only knew how much that was appreciated. ;)

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510
845-480-2058

hobbs at electrooptical dot net
http://electrooptical.net

I have a pair of 4CX250Rs laying around. Beefy.

Planar triodes (made with wire-mesh grids) went up to about 10GHz; their curves really sucked because the spacings were short and the grid openings pinch off from four sides -- not that linearity mattered in tuned amplifiers anyway.

Microwave tubes were almost all made for insertion into resonators, or had internal resonators themselves (especially klystrons).

Tim

--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms

Arguably, FETs and BJTs are high impedance devices, too. (Indeed, the vacuum triode is the only device that behaves fairly resistively over its entire operating range.)

As a matter of convenience, semiconductor impedances are just closer to those of conductors on dielectrics (= good for ICs and PCBs), rather than of free space (= good for antennas and chassis). You can easily go DC-10MHz with a 6DJ8 or 6DK6, operating with circuit impedances in the ballpark of

100-5000 ohms. Transmission lines will be wires, in free space, over a chassis (ground plane), with similar characteristic impedances, rather than dielectric-loaded planar strips, with impedances typically under 100 ohms.

I suppose if you included length as a factor in the comparison, the fact that a 6DK6, with about two milimeters from G1 to P, can achieve that much useful bandwidth, when a 2N3904, with maybe two micrometers from C to E, only achieves maybe 100MHz useful bandwidth, is pretty remarkable. I suppose a similarly sized PHEMT might achieve a few GHz (I would assume the ~30GHz devices actually in use have much smaller feature sizes). In principle, vacuum has much greater mobility, because you simply can't push electrons past 5eV or so in any semiconductor (maybe up to 10eV in diamond?), while electrons go to whatever voltage you apply to the tube.

Experimental data:

2us/div: ~150ns edges, from a small compactron sweep tube. Cathode follower driver provides a meager 10 or 20mA grid drive; it could be driven much faster. This is roughly equivalent to a 2N4401, open collector, driving an IRF510 with maybe 20mA peak and 10V. The "on resistance" is actually not even that bad: Triangle modulated PWM, so the top is tube-diode-clamped and the bottom is "on", pulling through the beam tetrode; circuit looks something like an upside-down buck converter. Peak current was 400mA, so the average output impedance is around 70 ohms -- not even all that bad, really, when you consider this beam tetrode is equivalent to a 500mA, 5kV MOSFET. The closest available equivalent, a single 2500V 1A MOSFET, has 250 ohms Rds(on)!

I've got to imagine tubes can be made with MEMS. Let's say...

1. Polycrystalline alumina substrate (or others)
2. Sputter tungsten to desired thickness
3. Mask and etch tungsten to create filaments and traces; leave openings for later substrate back-etch (which frees the filaments from the surface)
4. Deposit K-G insulator; any number of semiconductor processes could be used (SiO2, Si3N4, AlN, even polymers like kapton may be acceptable)
5. Mask off, then deposit grid (could be Al or Cu, or if higher temperature is required, Ti, Mo, Ta, W, etc.)
6. Deposit G-P insulator
7. Etch through G-P insulator, K-G insulator, and beneath the filament (into the substrate), to form the vacuum cavity

7a. A second, thinner etch step could be used to provide narrow channels between vacuum cavities, to facilitate pumping; alternately, the substrate could be through-etched or laser drilled, for pumping as well as for connecting vias to the filaments

1. Mask off, then deposit a layer of emissive material ("rare earth oxides") onto the filaments; alternately, deposit thoriated tungsten in step 2
2. Metallurgically bond the anode to the G-P insulator

9a. Before bonding, a layer of bond agent could be applied to the G-P insulator, e.g., tin solder, or copper braze alloys, etc. 9b. The anode could be a second substrate with many anodes etched upon it, allowing integration.

It may sometimes be desirable to start with, e.g., a preformed substrate in the middle, between grid and anode, so that extra thickness can be added to increase voltage standoff. This could be laser drilled, or molded and sintered, etc. In this case, the anode would be foil, metallurgically bonded to the substrate so it covers the holes, and the opposite face would be built up to form the grid and filament structures.

The whole thing could be shoved in a high vacuum furnace, where besides bonding things together neatly by diffusion (if hot enough; otherwise, by solder/braze reflow), all the adsorbed gasses are dissipated.

In step 8, the tradeoff is, oxide cathode coatings have lower current density, but much lower temperature, so they save power. W:Th has higher current density, but runs at a much higher temperature, costing more dissipation. The advantage is, W:Th has ion impact resistance, allowing >2kV operation.

This is all well within current manufacturing abilities and practices, has anyone done it??? Why can't I buy them at Digikey? I understand people hate filaments, but gee, they make multi-pin TO-247 leadframes, and I need to give IGBTs or MOSFETs their own damn isolated supply anyway, a filament supply isn't going to hurt me any.

Trick: omit the grid, saving on fab cost. Make a filament diode. Drive filament with HF AC from a transformer: switching times are thermal, but at this temperature and scale, it'll be on the order of 10-100us, good enough for HVDC switches.

Tim

--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms

to

New Orleans attracts a lot of Bible Belt yahoos, dry-county Mississippians, Texans, and other undesirables looking to drink 24/7 and generally be noisy. Sometimes a prom-night party will fly in from Dallas in a chartered jet, all dressed up. Strange place.

There are bars that have not closed for 100 years. If you try to leave a bar with a drink in your hand, they grab it from you, dump it into a plastic cup, and hand it back. The only rule is no glass on the street.

--

John Larkin                  Highland Technology Inc
www.highlandtechnology.com   jlarkin at highlandtechnology dot com   

Precision electronic instrumentation
Picosecond-resolution Digital Delay and Pulse generators
Custom timing and laser controllers
Photonics and fiberoptic TTL data links
VME  analog, thermocouple, LVDT, synchro, tachometer
Multichannel arbitrary waveform generators

to

This will be very strange, but I want to recall the memory of a good friend and bandmate, Mike Adam, who died a couple of years back.

He learned his thing in N'awlins. I recorded this at a gig we did together at the Sandbar in Cocoa Beach, Fla. Press "Lord Gimmme Shoes" at the link. It was original with Mike, far as I know.

it's Memorial Day weekend,and I Memorial Mike.

And, for Memorial Day, y'all be sure and drive slow.

-- Les Cargill