Yesterday I was "burning" an AT89S52, in-situ. I have access to parallel programmers and the chip was socketed, but I wanted to try it so I did not cheat :-)

For that, I built a "burner" from P89V51RD2, some supply (7805), a crystal, and a reset circuit. All in "bird's nest" style (the "burner" is not part of the circuit to be delivered so I didn't care). Still, I believe this is much better than any of the "LPT-connected" "cable".

The power was shared between the target board and the "burner". The target board has a proper reset IC, so no concern about problems during powerdown.

The "burner" firmware allows to manipulate the reset line (mental reminder: I need to add "manual manipulation" of the rest of signals (MISO MOSI SCK) and a "readback" of the state), set "speed" of SPI interface, transmit and receive individual bytes via SPI, and - the real stuff - emulates the "intelhex" protocol for the actual programming itself (I am using FlashMagic for convenience - Andrew Ayre would please kindly forgive this abuse).

I was cheating a little bit, as this setup was already tried in an almost a "standalone" configuration (the target AT89S52 in that configuration had the MISO/MOSI/SCK pins reserved for programming).

But this time it did not work at once. Bah...

Scope would be cheating so let's see what we can do with a cheap DMM.

The first problem was in the reset. On the target, a TL7705 provided reset for AT89S52. The "positive" reset output is an open collector of a PNP (!) transistor in TL7705, so it needs a pulldown. There is one built-in in AT89S52, but I don't like these things left "floating" (for case when other derivative with possibly no internal pulldown on RST is used); so there is an extra 10k pulldown. The "burner" had no pullup on pin driving RST, and the internal set of pullups of P89V51RD2 was good only to pull it up to some 2.6V... A 1k pullup and here we go with 4.5V, that's good enough.

The second problem was that MOSI conflicted with the circuit on board. There is a rotary encoder connected to it. Of course, when I designed the board, I knew it might conflict and that I would need to set the encoder so that it's switch is open during programming. And, of course, I forgot that when it came to actual programming. But this was easy to discover, too; and would be much easier would I put the manual manipulation and readback feature into the "burner's" firmware.

After the second problem solved (encoder turned to proper position), the autobaud/erase/program/verify/run cycle was repeated several times, with absolutely no problem whatsoever.


Conclusion (valid for LPT-style "programmers", too):

- state of RESET/MISO/MOSI/SCK (and oscillator) should be known (read back), and they should be settable/resettable "manually"

- consistent behaviour is much more easier to achieve in a "standalone" programmer


Jan Waclawek