Just too good to be true

Steven Koepke, who goes by koepkesd @ Stocktwits and Steven @ Yahoo, has been busy reading this blog while trying to justify the lies by the Nanox CEO that somehow the $200 (or $100, depending on the day) Nanox.Tube can replace the $150,000 modern CT tube (these are "list prices," of course).

For the longest time I was trying to figure out how Ran can claim that his field emitting device (FED) can generate x-rays on par with the high end x-ray tubes used in CT machines. Those large tubes can generate 800 to 1,000 mA at 120 kV. They also cost $120,000-$150,000. Here is the math. The MEMS (FED) chip has an active area of 0.126cm^2 (4mm diameter on chip in diameter and the power level was communicated at 2.5A/cm^2. The power is concentrated down via focusing device onto the tungsten anode. The basic math provides the power of the beam to be 314mA and 120 kV (per conference call last week). That's quite close and running multiple sources in parallel amplifies the power. Cost comparison: 5 small tubes @ $100 vs $150,000 for singe large CT device tube. Micro-X has a similar arrangement working today with a CNT device (also Field Emitting Device)

 

So, what's wrong with his reasoning?

There is no such thing as a field emitting device.  FED refers to a failed display technology -  a field emitting DISPLAY.  It does not contain "a field" of emitters, as Nanox CEO believes - it emits electrons induced by an electrostatic field.  It is not a more efficient or a cooler way to generate x-rays - all x-ray tubes, whether using a cold cathode (based on the field effect) or a hot cathode (using a hot filament) to emit electrons, have about 1% efficiency as 99% of the energy applied to the tube gets wasted as heat at the anode.  A hot filament uses a lower voltage - about 4V - than the 40V (or way more) needed by a cold cathode.  Roentgen discovered x-rays in 1895 using a cold-cathode (gas discharge) tube.  GE invented the hot-cathode x-ray tube in 1913 and obliterated the cold-cathode ones.

The proposed Nanox.Source chip is not MEMS, as there is nothing mechanical about it.  The chip is not real, or commercially available, of course, as Nanox has no ability to manufacture it commercially, at least not yet.

The current density of 2.5A/cm^2 comes from a fraudulent, that is, intentionally misleading, 2015 datasheet by Nanox predecessor, which I have previously linked here on this blog. 

The Nanox.Tube cannot do 314mA and 120 kVp.  The one used in the Nanox.Cart can do up to 2mA and up to 40 kVp, at most (or 0.08 kW, per 510K summary).  The CEI one can do up to 1mA and up to 100 kVp (or something like 0.1 kW, per CEI video).  The tube used by GE in the predicate device for Nanox.Cart can do about 40 kW - it has a rotating anode.  The CT tube can do about 120 kW (using Steven's numbers). So, to replicate the power of a $150,000 CT tube, one needs to use, oh, something like 1,000-1,500 Nanox tubes that cost $100,000 or more.  An after-market CT tube will cost less than $100,000, of course.  All this has been already discussed last year.

Micro-X has a 4.8 kW tube (a bulky stationary-anode one) - it uses carbon nanotubes, which Nanox says is impossible - it sells a few units a year.  The biggest CEI tubes are smaller sizes than Micro-X's and go up to about 2.5 kW (also stationary-anode ones).  

Update June 10, 2021:  Investors will eventually blame Nanox CEO for their delusions.  Steven continues:

The anode temperature becomes the challenge with the NNOX tube. CEI states that their tubes can handle about 60KJ. The RSNA video shows the bed is moving through the sources quickly (15-20 seconds for whole body). My guess is that NNOX is using high current short bursts to keep the anode temperature under control. In the video they may have used 300mA for up to 0.2 seconds to make 10 shots (capturing 8" per shot) while the cart moves through. 300mA x 0.2 X 10 shots = 60KJ. You can't shot this with dental tubes like that. They don't have the current and the image gets too blurry.

He is right that a typical dental tube (which has a better performance than the Nanox tube) cannot do 300mA.  He is also right that at some point, the heat capacity of the anode becomes a challenge (the anode temperature is not really a problem - it is the temperature of a part of the anode, the tungsten target, that is the challenge).

But Steven does not understand what heat capacity means.  Yes, one of CEI's most powerful medical tubes, OX125-06, can handle 60 kJ (CEI only makes stationary-anode tubes).  But that does not mean that it can do 300 mA or that 300 mA  x 0.2 s  x 100 kVp  x 10 shots = 60 kJ.  Nor does it mean that you can do 5 A x 0.2 s  x 60 kVp x 1 shot.  CEI provides nice charts in its datasheets to explain the interplay between heat capacity, tube current, tube voltage, and time.

As the charts show, the tube cannot do more than 35mA at 60kVp for 0.1s or more than 20mA at 110 kVp for 0.1s. But it can do 15 mA at 100 kVp for 10 seconds.  The RSNS 2020 demo, which we now know was fake, never demonstrated a full-body scan - it scanned three phantom "organs." The "hand" scan consisted of 45 "shots" or images ( 5 tubes x 9 tilts/translations) - it took about 50 seconds for the images thumbnails to appear on the display.  That is about 1 second per shot, not 0.2 seconds (and we don't know what that's even real).

The CEI OX-70, a dental tube,  can do about 32mA at 60 kVp for 0.1s or more than 20mA at 90 kVp for 0.1s.  It can do 10mA at 90 kVp for 10 seconds.  Here is some summary table from CEI's datasheets.  Stationary-anode tubes all look kind of the same.  The tubes that do less than 100 kVp are "dental" and typically tolerate half the current than the "medical," and are a bit smaller. 

Model	Voltage  kVp	Current mA, 0.1s	Current mA, 100kV	Focal sp. (mm)	Diam. (mm)	Length (mm)
Small/Dental tubes
OX/70-P	70	19	N/A	0.8	30	72
OX/70-5	70	11	N/A	0.5	30	72
OCX/65-G	70	12	N/A	0.8	30	76
OX/70-4	70	9	N/A	0.4	30	72
OX/70	90	21	N/A	1.2	30	82
OX/90	90	9	N/A	0.5	30	83
OCX/70-G	70	12	N/A	0.8	30	65
OCX/70-G4	70	8	N/A	0.4	30	65
Medical/Mobile tubes
OPX/105	110	18	2.5	0.5	42	125
OPX/105-4	105	17	2.5	0.4	42	95
3D/cone-beam CT tubes
OCX/100	105	20	4	0.5	46	140
OX/100	100	26	1.5	1.0	35	85

Recall, the Nanox tube in Nanox.Cart can do only 2mA at 40 kVp (for 0.1s -1 s).  The CEI Nanox tube can do only 1 mA at 100 kVp for 40 seconds (per CEI video).  The CEI OX-70 dental tube can do about 40 mA at 40 kVp for 0.1s, about 25 mA at 40 kVp for 1s, and about 3 mA at 90 kVp for about 40 seconds (per datasheet charts).  If Nanox tubes perform like poorly-made hot-cathode dental tubes, they probably are.  No mystery Nanox.Source chip required.

Update June 10, 2021:   Just to clarify, regular dental tubes (just one or 5 ) can definitely replicate the fake RSNA 2020 Nanox.Arc demo.  The "hand" scan took about 50 seconds for 45 images.  Let's see whether a dental tube can do 45 images at 45 seconds, that is, a bit faster.  A CEI dental tube operating at 90 kVp can do 3mA continuously for 45 seconds, so each exposure (image) will be 3 mAs at 90 kVp.  The Nanox.Cart demo at RSNA 2020 image needed just 1.5 mAs at 40 kVp (so, significantly less than 1/4 of what the CEI dental tube can supply).  Commercial fluoroscopy equipment does ok with 100 kVp and 1 mAs for each frame (image) at 30 fps.  So, sure, with a good enough (expensive enough) detector, you can do the Nanox.Arc tomosynthesis within 45 seconds.  But the detector (regardless of the tubes used) won't cost anywhere near $10,000.  And no one would like to look at the images (the American College of Radiology never considers a tomosynthesis procedure to be "usually appropriate,"  except for breast, which the Arc cannot do).

The "nano" in Nanox isn't

So, Nanox has been boasting in the technology section on its website: 

Over nine years of development by a Japanese and Israeli engineering team, produced a ‎stable Cold-Cathode field emission MEMS silicon‎. Using proprietary Micro-Electrical-Mechanical-Systems (MEMS) techniques, ‎millions of nanoscale gates are fabricated on each silicon chip. ‎Nanox emitters are far more uniform than carbon nanotubes (CNT) and are orders of ‎magnitude smaller than conventional Spindt-type cathodes‎. ‍

But is any of it true?  For example, has Nanox been able to miniaturize a technology that was a complete failure and make it potentially successful?  The answer is:  No!

If Nanox' emitters are "orders of magnitude smaller" than conventional Spindt-type cathodes, this means they are smaller than 1/100 the size of the "conventional" emitters (that's two orders of magnitude).  Here is how Nanox emitters are supposed to look like under an electron microscope (Slide 13, January 2021 JP Morgan presentation).

They look tiny, but how tiny are they?  Nanox is withholding that information.  Luckily, Nanox predecessor company has already published the original image - in early 2016, in a one-page "Nanox Technology Brochure" - with an embedded scale in it, like any regular image, or micrograph, generated from a commercial electron microscope.  All we have to do is use that scale to measure the diameter of the gate holes and the distance between them, and compare to say state-of-the-art Spindt emitters twenty years ago.

Aha.  So the holes are about 300nm in diameter and the distance between them is about 500nm.

And, what was the state-of-the-art twenty years ago, in year 2001?  Here is a picture from the Motorola paper titled "Field Emission Displays: a critical review" 

Uh-oh.  Turns out the Spindt holes from Candescent Technologies were actually smaller - 100nm in diameter, and intentionally placed at random distances.

Is it possible that somehow Nanox' team did not know about those developments twenty years ago?  Nope.  Here is why.  According to the unofficial history of the Spindt scam,  three years prior, Sony, desperate to maintain its relevance in the TV market, but completely clueless, joined forces with Candescent.

In November 1998, CTC announced an agreement with Sony Corporation for joint development of a 14-inch diagonal FED by the year 2000. Both companies pledged to spend $50 million on this effort. Most of the work would be performed at CTC's plants. A team of six Sony engineers were sent to San Jose to begin the work, with some additional staff dedicated to the project in Japan

Motorola had already canceled its project in 1999, thus the paper in 2001, due to inability to "solve some basic technology problems."  Candescent went bankrupt in 2004.  It took Sony a few more years to realize its mistake, but the $1 billion R&D spending is a complete myth.

So, to reiterate, ‎Nanox emitters are not orders of ‎magnitude smaller than conventional Spindt-type cathodes‎ - they are, in fact, LARGER.  Not that it matters, because Nanox, contrary to the false claims in the Prospectus and elsewhere, has no access to facilities to fabricate them commercially.

And the "nano" should have been a red flag anyway - MEMS in the supposed Nanox "Cold-Cathode field emission MEMS silicon" stands for micro-electro-mechanical systems.  Electro, not electrical.  Micro, not nano.  And there is nothing mechanical (moving) - the "electro-ns" do not count.

Ok, but what about those 100 million emitters in Slide 13 above?  Well, that number is possible.  Assuming those are positioned in a 10,000 x 10,000 square, and assuming 800nm distance between the tips, that gets us to a 8mm x 8mm "chip," ballpark.  But such a chip, even if it were real, is not changing anything.

In case anyone was wondering how big the Spindt emitters were 50 years ago, here is a diagram and a micrograph from the 1976 Spindt paper.  The gate hole diameter is 1,500 nm.  So Nanox emitter is just 1/5 of it, not 1/100.

Update March 29, 2021:  reworded and added the original Spindt emitter size.

Update March 29, 2021:  Wikipedia's page on Field Emitter Arrays have an interesting entry about nano-Spindts.

Nano-Spindt arrays represent an evolution of the traditional Spindt-type emitter. Each individual tip is several orders of magnitude smaller; as a result, gate voltages can be lower, since the distance from tip to gate is reduced. In addition, the current extracted from each individual tip is lower, which should result in improved reliability.

How did this incorrect and misleading entry wind up on wiki?  After checking the edit history, it turns out Nanox added it on December 22, 2015, a few days after the failed attempt by Nanox predecessor to market its fake cathode at RSNA 2015 (Nanox current CEO was then the Chief Strategy Officer).  Here is who added it:

My name is Joshua Lilienstein. I am an American medical doctor, now specializing in medical device development. I currently serve as Chief Medical Officer for Nanox Imaging, Plc., a Japan-based startup. Nanox's core technology is the field effect cathode. I will be editing entries that pertain to this technology, and specifically, in its application to medical imaging. I am aware of Wikipedia's Conflict of Interest policies, and will endeavor to abide by them.

At least he was honest about something.

He referenced an interesting paper that does not support his wiki entry in any way, but describes the long-lost x-ray detector, SAPHIRE, that was supposed to use nano-Spindts.  This is a flashback to the times when Nanox team "believed" that the x-ray detector was more important than the x-ray source.