LTC26xx series IO levels – or your favorite DAC chip.

I find that the LT2600 series of DAC devices are quite nice in many respects. 5V outputs are easily supported, separate reference inputs for each DAC. Pin compatible 12, 14 and 16 bit versions. Unfortunately, they do not have a VCCIO supply pin, so the SDO pin will be at 4.6V or more, which is not compatible with a 3.3V processor for example.

There are more than a few DAC manufacturers, and there is a huge range of options. Does anyone know of an alternative DAC that has similar specs, and handles the IO level issue better?

My first thought was to simply add a level shifter between your 3.3V SPI bus master and the 5V LTC26xx, but I'm assuming you are asking the question to avoid the need to add this additional chip to your board.

After looking over the datasheet for the LTC26xx series it says in the electrical characteristics of page 4 that the worst case digital input logic levels for SDI are 2.4V to register as high and 0.6V to register as low. If you compare this with the worst case output levels of the microcontroller you plan to use as the SPI bus master, and they are above and below these numbers respectively, it seems it should be possible to interface the LTC26xx directly to the 3.3V microcontroller without the addition of a level shifter. If you haven't attempted this yet, it may be worth trying based on this.

As far as your concern about the output level of the SDO pin of the LTC26xx being 4.6V or more, I'm not sure I understand why this would be an issue unless you are trying to connect this output back to the 3.3V SPI bus master or to another 3.3V logic level part. I'm not sure why this would be done, as far as I can tell the SDO pin is only used for the "Daisychain" option of the device. If you are daisychaining to multiple LTC26xx chips then this wouldn't be a problem as they are naturally 5V logic level compatible. If you are instead daisychaining back to the microcontroller or to a separate 3.3V device as mentioned, then you are correct and would need to check the logic levels of the downstream device to ensure it would accept the higher 5V input without damage, or else add a level shifter as mentioned previously.

The other thing you mentioned was the separate reference inputs for each DAC. As far as I can tell from the datasheet, I only see a single reference pin that's common to all 8 of the DACs inside the chip, and I couldn't find any information in the applications section about the ability to set different references for each DAC internally using the SPI bus for communication. Perhaps I've missed something, but I'm not sure there are in fact separate references. The only case I could think of where that would be true would be again in the daisychaining case and you are talking about setting different output references for the 8 common outputs of each LTC26xx chip in the chain, which should of course be possible.

If you are determined to find an alternative DAC, take a look at the following part on the Digi-Key website from Texas Instruments.

http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail&name=296-18198-1-ND

It only comes in a 12-bit resolution and is a little more complex, but it does support 5V analog outputs and does explicitly state that it is 1.8-5.5V logic level compatible. It also does have 4 separate analog output reference pins, one for every two of the 8 DACs.

If you don't want the added pin complexity, and don't want limit yourself to 12-bit resolution, then take a look at the following link to a list of parts on the Digi-Key website for the Analog Devices AD56xx series.

AD56xx Series

This series is very similar to the LTC26xx, but in this case you loose the daisychain ability if you were using it. I only mention it because the AD56xx series has more forgiving high and low logic levels for SDI listed in the electrical characteristics (2V high, 0.8V low).

I hope at least some of this information is useful to you and answers your question.

Yes, the SDO is the only problem pin. I was hooking it up out of habit, and to have the option to read back the values. I had not considered that the SDO might be just for daisy chaining, and not really intended to be linked back to the micro.. that may explain why it is the way it is.

Check out the LTC2604 - quad DAC. This one has separate REF pin for each DAC. This works well for out high voltage DC-DC supply controller. The High Voltage (0-500V) DC-DC converters generate a 5V reference. I can use this as the REF input, then the 16bit DAC output of the LTC2604 should be well aligned to the supply, better than using an external 5V that is not linked with the DC-DC converter.

I see how that could potentially add robustness into your design by feeding back the SDO to the microcontroller. If you are constantly checking in software that what you sent is what you get back you may be able to program in some corrective action or at least record the data transmission discrepancy. Excellent idea if you've found a way to integrate that into your system.

Also a great idea to use the internal 5V reference of the converter as the REF input to the DAC. Avoid adding the external reference on your board and as you said automatically get nice alignment to the supply itself. Plus, from the DC/DC controllers from Linear Tech that I've looked at they always seem to incorporate excellent precision internal references with low temperature drift.

And thanks for clarifying about the separate REF pins for the DACs. I had limited my parametric search to only parts with 8 or more DACs per chip after I saw that spec in the first datasheet for the series that I opened up. My brain must have jumped to the false conclusion that all members of the series were octal. This filtered out the separate datasheet for the 4 DAC versions of the series that use the freed up output pins as separate REF pins instead.

I guess working at Digi-Key with so many different vendors and parts has trained me to always take an aggressive approach to parametric searching to try and drill down to that "just right" part. I should write some new neural algorithms for myself to scale back the attenuation on my own mental filter when I'm searching on individual manufacturer websites.