Digital pots

[PRR]

> Analog Devices AD5920

Probably spelled "AD5290".

Why is the 256-step part cheaper than the 128-step part?

Why is an electronic giant unable to embed an Ohms symbol in HTML"?"

AD7376 sheet fig 22 clearly shows THD+N versus level. Unless something very odd is happening, below 0.1V the "THD" is entirely "N". For the 10K part, set mid-code, 0.5% at 0.001V is 5 microvolts. This is not unlike audio noise of a perfect resistor pot set at mid-way. You expect the 100K pot to read 3.16 times higher and it is about 3 times higher.

However the tabulated specs for the 128-step 10K part cite an impossible number "0.9nV/rtHz". The 256-step sheet has "9" which does not defy room-temp physics.

Bandwidth is not stated, but fig 21 suggests they got a 20KHz low-pass.

http://www.analog.com/UploadedFiles/Data_Sheets/AD7376.pdf

Hell of a jump from 0X01 to 0X02.

Not sure why all the high-loss codes are so far off from what they should be. Ah... the 5290 sheet shows the real problem. How many CMOS gates can you put in a signal path?

PSRR is not fabulous, but I guess we can R-C it away.

 

[jdbakker]

Interesting parts.

[quote author="PRR"]Why is the 256-step part cheaper than the 128-step part?[/quote]
It looks like the 128-step part uses a resistor ladder, with only one pass gate in the signal path; the 256-step part uses an R2R-like tree as you note, with much more switches in the signal path but way less switches and (precision) resistors on-die overall. Most of the price difference may well result from laser trimming and other post-processing/QA phases.

Considering this difference in architecture I don't get why the distortion graphs are so similar, to the point that I can't eyeball any difference between AD7376 figure 21/22 and AD5290 figure 18/19.

Speaking of the distortion graphs, I wish AD had specified under what conditions these were measured. I'm puzzled as to why the distortion starts increasing at a lower voltage for the 10k parts; the only way I could see that happening is if they'd measured with a low-impedance probe. Assuming identical CMOS pass gates I can't see why THD would be any different between 10k, 50k and 100k parts when measured with a high-Z probe (modulo capacitive effects masking higher-order harmonics, naturally).

JDB.

 

[bcarso]

Wish we had a Rochey equivalent for AD and Cirrus in here.