Talk:Arbiter (electronics)

From Wikipedia, the free encyclopedia

[edit] Metastability

This article claims:

Yet metastability is an inevitable consequence of quantum indeterminacy, since if the inputs to an arbiter (flip flop) arrive almost simultaneosly the cicuit most likely will traverse a point of metastability.

Can you provide references to laboratory experiments that demonstrate this, by ruling out the non-quantum-mechanical explanations? For example, ground bounce is currently taught by EE schools as one of the many causes of metastability, and quantum is not typically mentioned. Are there texts which explicitly develop this claim in a testable, verifiable fashion? linas 23:38, 3 October 2005 (UTC)

Ground bounce is not generally thought to be the cause of metastability in arbiters. There are published articles on metastability in arbiters that trace it back to quantum physics.--Carl Hewitt 23:45, 3 October 2005 (UTC)

Can you provide references? None of the references provided in the current article even use the word "quantum". I have the luxury of being able to talk to people who work in this business. They rather question the applicability of the qantum explanation. We discussed a certain proprietary clock circuit that operates at frequencies well beyond what is currently available for chip clocks; fundamental to its design is is the fact that its driven into a curious "permanently metastable" state, with the actual clock phases pulled off in a very novel and unique fashion. It seems that a purely classical model of the circuit is able to adequtely model the fabricated devices, with regards to oxide thickness, doping, etc. Given that a leading-edge, state-of-the-art circuit designer who builds metastable circuits intentionally does not beleive in any quantum explanation for the circuit ... makes me wonder about the claims made in this article. I'd like to see some actual lab proof, as opposed to theoretical arguments. linas 00:00, 4 October 2005 (UTC)

I mean, this is kind of important. If I can fabricate quantum logic at ordinary fabs using ordinary design rules, and build circuits that stay in quantum states for very long periods of time, this would have a huge and broad set of implications for people who are trying to do quantum computing. Under current technology, it is very hard to build systems which hang in an true quantum state without having the wave-function collapse. Having the ability to use ordinary fab techniques would change the state of the art. I'd really like to investigate more; this would be very exciting if there was anything to it. linas 00:07, 4 October 2005 (UTC)

The indeterminate states that we are talking about here are not qubits so they can't be used for that kind of computing. There have been some experiments done to see how long an arbiter can be maintained in a metastable state. I am not sure what the current record is but I believe that for a well designed arbiter above room temperature that it is not a very large number of times longer than the time it takes an inverter to operate. So I would be very curious about a "permanently metastable" state for such an arbiter. Of course a well designed arbiter is supposed to exit a metastable state as rapidly as possible.--Carl Hewitt 01:47, 4 October 2005 (UTC)

Appearently, publc talks were given so its not confidential. Its a ring of 5 or 7 or 11 gates hooked up so that it oscillates at a frequency slightly above the switching time for a single gate. The magic is that there are interconects running across the ring, e.g. some gates are skipped, the result is that subsequent stages are being driven to invert even as the driving stage hasn't switched itself. Sort of like a polynomial decoder, but with more feedback, hooked up to be free-running. The ring itself just sort of floats, the clocks are reconstructed off of it. linas 00:27, 5 October 2005 (UTC)

I asked a similar question in regard to quantum indeterminacy in computation. I find the claim that metastability is a quantum mechanical effect puzzling, and I gave a reference in which this claim is disputed.--CSTAR 20:02, 4 October 2005 (UTC)

Please see my reply there. Thanks,--Carl Hewitt 21:12, 4 October 2005 (UTC)

Well, you refute the claim there, but still don't cite anything in support. If they're not qubits, what are they? What exactly, is decohering? How does one make a lab test, so as to distinguish a quantum effect from some other effect? linas 00:41, 5 October 2005 (UTC)

We are dealing with entangled quantum states (that happen not to be qubits) that are decohering. Lab experiments are done in the standard way and they clearly show metastability. What exactly is your question about these experiments?--Carl Hewitt 05:23, 5 October 2005 (UTC)

If a system has entangled states, then, if one is clever, one may construct qubits from them. The only difference between qubits and entangled states is that the former were designed to specifically be easy to manipulate and control. If a system exhibits entangled states, I'm pretty sure someone somewhere has attempted to use it to build qubits. Of course, with only variable success.

There are strong arguments why you cannot build qubits just using arbiters.--Carl Hewitt 16:46, 5 October 2005 (UTC)

Tell me about one lab experiment that is able to distinguish between two different theories of metastability. Tell me about these two different theories of metastability, and tell me how the experiment is able to differentiate between them. linas 14:29, 5 October 2005 (UTC)

[edit] References

Also,

when cititing references, even if they are "external web links", please cite as "author, title (date) publication info, (if applicable)". This format should be used even for links to web sites and oher non-refereed/self-published data on the net. Just because its not in a journal doesn't mean we should be sloppy in citing it. linas 00:19, 4 October 2005 (UTC)

The following text was copied from a post on Wikipedia talk:Requests for arbitration/Carl Hewitt/Workshop by Allan McInnes (diff here), since these seem to be useful and relevent. linas 00:54, 31 January 2006 (UTC)

1. What properties of arbiters are not adequately described by classical models?
   1. "Overview of Modelling and Analysis Techniques for Arbiters and Related Circuits" gives a comprehensive overview of the dynamical systems approach to modeling arbiters, and makes no mention of any properties not adequately represented by classical models
   2. "Synchronisation and arbitration circuits in digital systems" shows how the exponential probability of stabilization can be derived classically
   3. "Measurements on a High Speed Arbiter" demonstrate empirically that real arbiters match the classically-based physical models
   4. "Metastable States in Asynchronous Digital Systems: Avoidable or Unavoidable?" shows that metastability is a general phenomenon which arises wherever a continuous set of states has to be mapped onto a discrete one, that metastability cannot be avoided in principle with classical devices (i.e. metastability is a property of classical systems), that quantum devices can in principle avoid metastability (i.e. metastability is not necessarily a property of quantum mechanical systems), and that other quantum effects like macroscopic quantum tunneling prevent a practical application of such quantum devices (i.e. even quantum mechanical systems will suffer from metastability - which is not the same as saying that metastability is caused by quantum effects).
2. The only hits that Google returns on quantum indeterminacy and arbiter metastability appear to be Wikipedia articles that you have written, or mirrors of those articles elsewhere. The same appears to be true of quantum decoherence and arbiter metastability. Obviously not definitive, but disconcerting nonetheless.

--Allan McInnes 23:43, 30 January 2006 (UTC)

[edit] Cleanup on text

Did a bit of cleanup on the text, to improve readability and to make it clearer why this is important. Please check for correctness. Thanks.

(Having once been a systems programmer for one of the rare UNIVAC 1108 multiprocessor mainframe systems designed before sound arbiter design was figured out, I was painfully aware of this very real problem. Not fun. I like the reference to the Ginosar paper; apparently some designers are still getting it wrong.) --John Nagle 16:55, 4 April 2006 (UTC)