AVR Freaks

AVRfreaks reached critical mass some years ago -- every possible AVR issue has been raised, discussed, and consensus reached. ;)

 

OK, I suppose I have to add something on-topic.  In a couple apps we have gone the "serial backpack" route and added a small AVR at the display.  Then UART or UART+transceiver added for the distance.  Not as inexpensive as the shift-register approach but handles the "remote" problem.  In addition, that remote AVR can scan user buttons/keys associated with the display and handle associated I/O  (LEDs; buzzer; ...)

 

 

BASCOM-AVR?.

Sigh.......

Jim

Does it have to be SPI? Could you not use I2C?

Obviously (I hope?)  the SPI is completely irrelevant to the LCD itself: the LCD itself neither knows nor cares anything about SPI - it is totally oblivious to SPI.

 

All the LCD sees is the signals at its pins - so your task is to replicate from the shift register exactly the same signals that the LCD would see if it were directly connected to your microcontroller.

 

So you could also use I2C to talk to your LCD.

A couple of left field suggestions:

 

1) use a 40 pin rather than a 28 pin AVR - you will get more than enough extra IO pins that way! I bet any cost up is no more than the cost of the 595 anyway.

 

2) What is actually on the whole of PORTC and PORTD of that mega8? You may find that some form of multiplexing (multiple buttons on a single ADC input for example) frees the additional IO lines you need for a direct connect. Your picture already shows there's just 6 active wires to the LCD and 3 to the 595 so it would seem you are only actually looking for 3 additional pins.

Oh I see the (mega8) picture was just "generic" culled from Google Images was it?

 

If the problem is really finding 2 pins on a mega32 then surely to goodness there's got to be something there that can be multiplexed to free just 2 more pins?

Not even if it saves the cost of the 595 from the design? I suppose it depends whether you are making just one of this circuit or 10,000 units? If the latter then saving the cost of $0.50 chip from the design could be significant.

Then why won't you explore the possibility of reducing the other mega32 pin usage so as to free enough pins for the LCD? As I read it you seem to be saying you are only 2 pins short. You may even find that your LCD pins could be doing "double duty". The LCD controller only really responds to data presented on D4..D7 when the E line is activated so, say you have some buttons in your design - perhaps a matrix? You might be able to use the same output lines that go to the LCD as the driven outputs to your key matrix too if the design is such that it can be either addressing the LCD or the buttons at any time but not both at the same time. Just don't drive the E line when you are outputting to the buttons.

 

And that's just one example - it'd be unusual in a 40 pin chip design not to be able to find at least a couple of pins that can do "double duty".

 

Perhaps you don't have a key matrix but just 4 buttons on individual inputs? If you could shuffle things round so that all 4 would connect to one ADC pin you can often use a switched resistor network to present different voltages depending on which button is pressed. However it's true that a possible complication is if you might expect to see more than one button pressed at once. However even that's do able. Conceptually have one button switch 0.5Vcc, one switch 0.25Vcc, one 0.125Vcc and so on. Now if you read 0.75Vcc you can be pretty sure it's buttons 1 and 2 that are pressed.

 

There are many ways to skin this cat!

 

If it could save $5000 from the production cost (10,000 74595's at $0.50 each) then it's got to be worth spending at least an hour studying how you might cut it from the design?

 

Or are you entrenched in the belief that the only way to gain the 2 pins you need is with a port expander?