Thread: DACs and analog inputs

So I think this is what I plan to use for my Analog input DAC. I think 32 is a good number (2k1Toaster's suggestion from another thread).
But my question is would you use only 5 bit inputs, or would spacing them out over the entire range be better?
I have a xls I here where I spaced out 32 inputs over the 256 combination range, so that I got even voltage seperation, the problem is the bit combinations are complex. Implementing that will be the hardest part of the design I suppose. Perhaps changing the output buffer so that it has a larger full scale output range, but cap it at 5V to protect the input, will work for the case of only using 5 input bits, while having a full 0-5V output.

National Semiconductor made the circuit pretty easy with their application note AN-1525, which dispells the percieved drawback that the DAC0800 requires negative suppply voltages (figure 5 is a complete single supply DAC circuit (12V)).

Anyway, creating the matrix to provide the proper unique input bit combinations is a task that I'm uncertain as to how to accomplish (sure I could draw it fine on paper, but simplified enough to put on PCB is a different story).

Which way would you lean more towards, using all 8 bits to simplify the output or simplify the input by using only 5 bits (or live with a smaller output range than 0-5Vdc)?

Other chips that might work better?

Well after playing with that excel file a little more, I started to see that the input section wouldn't necessarily be too hard. I added a a second section in sheet 2 that parses out the bits and does some conditional formatting. In doing that I began to see the familiar pattern and something else too, that should have jumped out at me before. If I use a 8 bit encoder, but tie the 3 most LSBs together, it's just like an 6 bit encoder and if all inputs connect to the LSB line, it can be made to be similar to a 5 bit encoder. That way each ascending input increments by 2^3 starting at 7. That spaces the voltages out evenly. I made a picture of what I'm talking about with a really cool graphics editor:

embarassing picture was here (I'm afraid it will sap the intelligence out of anyone who encounters it, so I burned it in the fireplace)

I think even I might be able to make this with a rather small footprint. I'm not sure if the 0.05" stagger will cause problems for me though. I just took what I have on sheet 2 of the xls and overlapped 17-32 over 1-16.

Well, thank you to all who read this and felt inclined not to say anything as to not hurt my feelings. I see it now... as any switch would always light up all of the bit inputs. so that would mean that each node would have to be driven high by a seperate line (maybe by transistors or even diodes which could be turned on in groups, thereby using one switch to turn on varied lines while providing isolation). That's a little more real estate.

I like the 5V single supply operation of the 7528 above, and also the AD7801. But I'm not too embarassed not to admit that I'm having a hard time with the CS and WR and LDAC. I'm assuming in the 7801 case I would tie the LDAC to ground and use the WR to load the inputs, but how do I fire that, do I use something like an OR looking at all 6 lines? What kind of pulse width do I need (can I get something appropriate with a capacitor to ground?). The function of CS (chip select) escapes me a little, even while looking at the timing diags.

Whew, that would be a lot of diodes (160, because I could probably just tie directly to the LSB line since they ALL send it high).

There has got to be an easier way.

Well the way I had it there in that drawing the green line is voltage available to go through the switch, through the red lines to each of the black lines going to the input bits. So yeah, kind of like a keyboard grid. I chose the connections to the bit-lines (red circles on black lines) based on the DAC.xls values I have on sheet 2 (starting at seven with with input 32 (3 LSB lines) and incrementing up to 255 by 8 at input 1 (every line,MSB through LSB). To increment by 8 I just used the 5 most significant bit-lines and tied the 3 least significant bit-lines together, the LSB line would be turned on with every switch, basically biasing everything up by 3 bits. The problem is that I can't daisy chain the bit-line connection, each connection has to be connected individually to the switch output, and diode isolated, so that one switch doesn't avalanch the whole thing on. So, more lines for each switch output and a bunch of diodes maybe.

I originally wanted to stay away from the type of DAC you recommended because I dont understand how it write-enables and what CS is for, I did figure I could select one DAC the way you recommended. What kept me looking (AD7801) was a desire for a datasheet with a really good application diagram for 5V. And I'm still looking for a good explaination on how to set the control inputs (as I mentioned above). The cascading opamps in the application info (page 8) of the TLC7528, seems complicated. I wish I could find one with the simplicity of DAC0800 (no control inputs) that works with 5V.

I'll try to get a better drawing of what I'm thinking of (with new corrections) up soon.