I got a TOP 2007 universal programmer yesterday, for burning EPROMs, GALs, and Flash memory. Whenever I hear the term “universal programmer”, I imagine a nerd-equivalent of the 1992 movie Universal Soldier, starring Jean-Claude Van Damme. Plot: a mild-mannered C++ programmer gets cybernetic implants and goes on a killing spree.

TOP 2007 Pros: It’s pretty cheap (about $100), and it programs GALs, unlike most low-cost programmers.

TOP 2007 Cons: Horrible, horrible software. I’ve seen slipshod, poorly-translated software before, but this really elevates it to an art form. Maybe 25% of the text isn’t translated at all, and is still in Chinese, including some status and error messages. What text is in English is so poorly translated, it’s difficult to tell what it means. “Driver not ready ,if load it now?” Damned if I know. And what’s a “cussor?” Clicking the close box minimizes the window instead. Reading from a chip overwrites the contents of any data file you previously had open for writing. Lots of options and menu items do mysterious unknown functions, or nothing at all. Device recognition doesn’t seem to work– you have to manually select the right device from a list of hundreds, and pray you got it right.

Despite these issues, programming a 128K Flash memory worked fine. I had problems with some Lattice 22V10D GALs, however. That’s especially frustrating, since GAL support is the whole reason I chose the TOP 2007 over other choices. The 22V10 is on the supported device list, and it does almost work, but not quite. What appears to happen is that it successfully programs all the 5000+ fuses needed to implement the logic expressions, but won’t program the last 20 fuses that configure the output pins to be inverting or non-inverting, registered or combinatorial.

Any sane person would have attempted to return the programmer, or junk it and get a better one. Instead I spent an afternoon trying to reverse engineer the software to see if I could fix the problem. By accident, I discovered that I could program 14 of the 20 unprogrammable fuses by adding extra fuse data to the data file, beyond the number of fuses that are actually in the GAL. I also found that 3 more of those 20 seemed to copy the data from elsewhere in the fuse map. But the last 3 fuses appear permanently stuck at 0. The net result is that with some complicated effort, I can completely configure 8 of the 10 GAL outputs. The other two are stuck in registered, inverting mode. I can probably work with that for most purposes. Worst case, those 2 outputs will just be unused.

GAL Counter: The hardware design calls for a stack pointer that’s a 24-bit up/down counter with output enable. The simplest approach would be to use six 4-bit up/down counters, and three 74LS244 drivers, requiring nine chips total. A better solution would use the hard-to-find 74LS569 4-bit up/down counter with integrated output enable, requiring only six chips. Better still would be to use three 22V10 GALs to make 8-bit up/down counters with output enables, requiring just three chips.

It turns out that programming an 8-bit up/down counter into a GAL is quite a challenge. For starters, there are barely enough pins. The 22V10 has 22 data pins, of which at most 10 can be outputs. I was able to barely cram it in by encoding some of the control inputs into a 2-bit function code. But the bigger challenge is that the logic equations for computing the new value of each counter bit require more product terms than the GAL supports. My equation for the MSB of the counter includes 19 AND terms all OR-ed together, but the 22V10 supports at most 16 terms. Here’s my equation:

/q7 := f1*f0*/q7 + /f1*/cet*/q7 + f1*/f0*/d7 + /f1*f0*cet*q0*q1*q2*q3*q4*q5*q6*q7 + /f1*f0*cet*/q7*/q0 + /f1*f0*cet*/q7*/q1 + /f1*f0*cet*/q7*/q2 + /f1*f0*cet*/q7*/q3 + /f1*f0*cet*/q7*/q4 + /f1*f0*cet*/q7*/q5 + /f1*f0*cet*/q7*/q6 + /f1*/f0*cet*q7*/q6*/q5*/q4*/q3*/q2*/q1*/q0 + /f1*/f0*cet*/q7*q0 + /f1*/f0*cet*/q7*q1 + /f1*/f0*cet*/q7*q2 + /f1*/f0*cet*/q7*q3 + /f1*/f0*cet*/q7*q4 + /f1*/f0*cet*/q7*q5 + /f1*/f0*cet*/q7*q6

cet is the count enable (active high). F=00 means count down, F=01 means count up, F=10 means load, and F=11 means no change. So this crazy equation says that bit 7 should be zero if we’re not changing and the current bit 7 is zero, or we’re trying to count up/down but counting is disabled and the current bit 7 is zero, or we’re loading and the input bit 7 is zero, or we’re counting up and the counter is at 11111111, or we’re counting up and the the current bit 7 is zero and any of bits 0-6 are zero, or we’re counting down and the counter is at 10000000, or we’re counting down and the current bit 7 is zero and any of bits 0-6 are one. Phew! If you can simplify that to 16 product terms, you’ll win a prize. Maybe I should create a truth table with 4096 entries, and build a Karnaugh map. Or not.

I need to take another look to make sure there isn’t some clever way I can simplify the equation to get under 16 terms. If not, then I’ll probably either make four 6-bit up/down GAL counters (which would be a little strange, since bytes would be awkwardly divided among different GALs), or six 4-bit up/down GAL counters, essentially replicating a 74LS569 in a GAL. So far I haven’t found any place that has 74LS569s is stock.