• I think the device should be able to be constructed with parts one can purchase at radio-shack.

• It should be constructed without integrated circuits in favor of FETs or BJTs one can feel free to exchange for various brands; however the design could be simplified with an IC should the user chose to trust a company (TI, nat semi...) and wishes to create his/her own.

• I think the so inclined user should be able to take it apart and put it back together with a soldering iron; e.g. having as few parts as possible, and provide schematics and documentation.

• I think the user should be able to exchange the "sample cell."  For what I am thinking, this would mean replacing the crystalline sample with a home-made one, if one has the chemistry skills to do so.

• Gaseous samples would not be permitted for an NQR device, but could be used with optical spectroscopy. But I do not think optical circuits are simple enough to be "provably untainted."  An optical physicist may disagree.

(Moved to top)

Bump.  Community input is needed.

Assuming I can get 1 random bit every characteristic relaxation time of Sodium Chlorate (T₂)...I could grab 256 random bits in 2.56 seconds with ease, since T₂ << 10 ms

I could conceivably measure two random bits at least with each T₂ since I can grab a metric also of the apparent field angle.

This is because the quadrupole lines are split by a static applied field, so earth's field will be split the levels; moreover, this depends on the instantaneous spin angle with the earths field.  Since earths field varies with location, and the crystal orientation can be made to have some randomness, I'm confident in 2 bits/T₂ for a single crystal preparation.

In a sample with two inequivalent chlorine sites, this is doubled.  Also, if we manage manage to measure 37Cl as well as 35Cl (the abundance of the heavy isotope is 26%) that is an 8x speed up.

There are some samples I know with even shorter T₂ whose quadrupole resonance is at nearly 100 MHz, though I've never seen what the S/N is.

RNG's today operate on thermal fluctuations and alternatively current "shot noise." These methods are in principle sources of quantum entropy.  However, they work on integrated circuits so nobody knows whats really going on inside of it.  One can perhaps test the legitimacy of the outputs, but I do not like the idea of ICs used in open source technology since they can always be backdoored - one must trust the IC manufacturer to use the IC.

This is different in that anyone with an electronics hobby can scrutinize it; these electronics presented are completely analog. One can also feel free to exchange the (sodium chlorate?) sample with their own if they believe I have doctored mine.  It's simple physics, rather than black-box electronics.



Anyone who is worried about keys generated by deterministic algorithms.  They are not guaranteed to be random, and if an attack is found, the bitcoins are gone (https://www.google.com/search?q=android+vulnerability+stolen+bitcoins&rlz=1C1CHFX_enUS565US565&oq=android+vulnerability+stolen+bitcoins&aqs=chrome..69i57j0l5.5858j0j7&sourceid=chrome&espv=210&es_sm=93&ie=UTF-8).  Perhaps it would be intended for large holders of coins or by servers who tend to generate a lot of keys...or advanced users.  

Or even hardware developers who want to offer a bitcoin product, but are afraid they will create something that doesn't offer enough entropy, making their customers lose all their bitcoins, and bringing a lot of heat upon themselves.

I would make my design open source so they can modify it as they please, asking only that their version's build characteristics can also be documented properly.



I have done much of the research already.  The exact details like this would come after I have community input on what features are actually desired.  How small do I need to make it to be desirable?  How do you guys think it should output the random bits?  Desired speeds?

Once I have answers to these questions, I can make concrete statements about dev cost and a timeline.

In short, I can make this work on my lab equipment already.  That isn't satisfactory of course; I'm using a $100,000 signal generator, and a few $30,000 oscilloscopes.  

But if I specialize the device to work for only a certain chemical, and function only in such a way as to be used in conjunction with bitcoin...it doesn't need to be made from expensive lab equipment.  So once we decide the properties, I can choose an operating frequency, and really make a nice parts list.

For a rough estimate, for the type of device I have in mind...to buy the raw electronics for a one-time build...the parts would cost under $20.  In the course of devlopment one should expect many failed attempts so the dev cost is much higher than the unit cost.

I already have most of the test equipment I would like or need to finish the job.

Thanks! Yes a lot of the mathematics will be very similar if I post a whitepaper.  This will be a good resource.  Some comments:

• This is optical - which as I noted above, is a bit above the scope of hobbyists
• They didn't really show me how to do it; moreover, their detectors are expensive
• Not at all to discredit their work - just to point out the need for a cheaper, more accessible solution

I like Microsoft research a lot.  It's spot on, it looks.  I can take hints from this for I/O.

This is a very nice a rigorous proof of some of the quantum phenomena I implicitly implied will work in my case, here with magnetization in crystals (but its physics...its always the same thing in a different can).

I gave a rough estimate of 256 bits in a few seconds above. I cannot give a precise speed comparison, except to surmise that optical phenomena are at least an order of magnitude faster than nuclear resonance as a general rule

I should say though that I did propose a method using resonance multiple species to speed up acquisition.  So its not that goals cannot be met using nuclear resonance, but I have displayed clear data for the 1x speedup - the up to 8x speedup would need to be investigated further to determine feasibility.

This in fact the speed of optical phenomena is why they seem to use computer algorithms for selection of relevant bits, because their data is coming at them a lot faster than mine would be. I instead would be using a small RLC circuit with a transistor amplifier and level detector. There would be really no software or digital circuitry involved at all. So the implementations are fundamentally different, but the mathematics (of QM) is really the same.

I envision something inherently simple to build and use. My device would be completely accessible to make at home, just by looking up the schematics and having skill in analog design, which would be a major difference that sums up the above points.

This may also have a side effect of furthering research in zero field NMR, since MRI machines in hospitals are using only high field 1H NMR.  Zero field is sort of a black art in magnetic resonance, which I think could be used cleverly in the medical industry. Much of the reason an MRI costs so much is the utilization of a superconducting magnet, wheres as the spectroscopic method I am employing here is used without external fields - no magnet necessary.  This technique has been largely ignored in the NMR and MRI communities, which aggravates me to no end, but we shant go there...but it is receiving attention again in the industry of explosives/landmine detection.

"Zero field NMR" is synonymous with "NQR" though this wording is misleading.

I'm guessing you're at a university or other institution if you have access to a few US$100,000s of lab equipment. I would strongly advise you to look for local business plan competitions, or speak with someone at the business school about building a team to take this forwards.

I suspect if you're right, you have the potential to affect a large number of fields beyond bitcoin.

This sounds like a lot of chemical jibberjabber for something that can be done with webcam noise or lava lamps.

One solution I would have is to take a software defined radio such as http://spectrum.ieee.org/geek-life/hands-on/a-40-softwaredefined-radio

Take the 3 million raw I/Q samples per second of background radio and send the data through about 10000 XORed SHA hashes to get 512 bits a second. Random.

This is digital random.

It is analog random that is proposed in this topic; its work is based on such phenomena as nuclear resonance serving as  a source of quantum entropy.

Please correct me if I was mistaken.