Head Bottle Washer

Biography has not been added


's contributions
    • I have several Windows 7 machines with USB 3 ports: an HP laptop with built-in USB3 ports and a desktop with an add-in card. Those drivers are apparently not native to Win 7, but the result is the same (people are running USB3 on Win XP). I develop imaging applications and typically get around 2.4 Gb/sec real-world through USB3, using Opal Kelly FPGA boards. I'm sure there are driver differences but in my case the limitation is on the FPGA side. I think it would be very feasible to design a much better cheap logic analyzer, with a gig of DDR3, at least 32 channels, 200 MHz external sampling, USB3, trigger in/out, and support for triggering off the protocol analyzers (this seems to be lacking in all the inexpensive logic analyzers -- post-processing captured data doesn't count). FPGAs are big and cheap enough that this would be straightforward.

    • Agreed. The acceptable outcomes to get each face to appear only once for 6 rolls of a die are the permutations of (1,2,3,4,5,6), of which there are 6!. The total number of outcomes for 6 rolls is 6^6.

    • LogicPort, like Saleae, uses compression, so the 2K buffer is not as much of a limitation as it might seem. But it definitely is less than ideal -- sometimes I have to hide the frequently-changing signals (like clocks) from a view to get longer sampling times. Personally a logic analyzer with only 16 channels and 12.5 MHz sampling would be useless for me since I work mostly with much faster FPGA signals. The combination of LogicPort and internal Chipscope covers most of my bases. I'm not convinced about the reason for not using USB 3.0 -- it was released in 2008, and it would appear that the Logic16 is more recent than that (2011 maybe? If so I'm a bit confused about why it's called the "new" Logic16 in the review).

    • I've used the similar Intronix LogicPort for years. With 34 channels, 500 MHz sampling, and more flexible thresholds I think it's worth the extra ninety bucks or so. Now I'm waiting for a USB 3.0 version with significantly more hardware memory -- small sample storage is its Achilles heel. It would seem that the Logic16 relies mostly on PC-side sample storage (streaming data via USB after the trigger event?), which would be useful if you can live with slow sampling rates (12.5 MHz for 16 channels) but need to acquire more data than can fit in on-board storage. It's not clear why Saleae didn't go with USB 3.0 with its much higher bandwidth.

    • In an email to Max I expounded on the following statements that make the proposition impossible: 1. There is no reason to consider more than 2 bits in the expansion of the ternary digits. 2. If you choose any three of the possible 2-bit patterns (00, 01, 10, 11), there will always be one transition with two bits changing. 3. The ternary gray code includes all three transition possibilities: 0<->1, 1<->2, 0<->2. So there is no way to map the ternary digits onto binary patterns without changing two bits in at least one of the transitions. It seems that gray codes with unit steps will always allow a binary expansion that is itself a gray code, while those (like the ternary gray code) that require non-unit steps, don't. The reason is, with a unit step you could avoid the transition (0->2, for example) that maps onto the forbidden binary transition.

    • Just to summarize more clearly: I answered the first question in the blog: "Now, this isn't to say that it isn't possible to come up with some sequence that passes through all nine combinations and returns to 00 with only a single digit change each time. But I'm a bit busy at the moment, so I'll hand this over to you... can you come up with such a sequence?" The harder question that followed is still unanswered, although the answer is probably "no": "And, if so, can you map binary values onto the ternary digits such that we have Gray codes in both bases?"

    • You are making an assumption about what the meaning of "Gray code" is. The whole point of ternary (and higher) codes is that we don't expand each digit in a lower-base representation, we are concerned only with that base -- there is NO need to look at the digits in binary (or decimal, re your last comment) from a theoretical point of view. If you define a Gray code as being one where only one digit (in that base) changes (which is the standard definition), then there are many ternary gray codes. If you add the additional restriction that the jumps must be +/- 1, then of course there are none. You are making a perfectly valid connection to practical things like voltages and such in your comments, whereas I (and I think the original blogger) were concerned more with the abstract concept of Gray codes extended to higher bases than 2.

    • c2: '3' is not a valid digit for a ternary sequence. Here is one base-3 gray code: 20 22 21 11 10 12 02 01 00 or 10 11 21 20 22 12 02 01 00 There are many others. The fallacy of the post by probinson is that the count sequence does not need to alternate between odd and even. It is also not necessary (in most applications) that the sequence be of increasing magnitude.