Thermometer Code Chip

Can you offer a few part numbers?

Not sure why you limit this to pipelined SAR. Subranging also has to do multiple steps which can be pipelined or not. But more importantly, there can be some "issues" at the edges of a range which require correction. In face, subranging flash converters are also referred to as pipelined... a rose by any other name...

We used to use buckets of those ~30 years ago.

It was a cool application, essentially an all-analog "arb generator".

We cascaded several LM3914s. The outputs each drove a slider potentiometer. These were summed such that you setup the waveform on the slider array and it output the shape you set.

A clever bit was to inject a triangle wave dither; this interpolated between the set points so there were no steps. This also dithered the LED mounted above each slider so there was a smoothly moving blob of light rather than a stepping one, indicating the "current" position in the signal. (It was a slow signal for hydraulic positioning).

Fun. The Exact company used to make an analogue arb with about 30 pots on the front panel and a counter + mux to switch between them.

Cheers

Phil Hobbs

I once saw a really fast arb that used the reflections on a tapered transmission line to define the waveform. The line was a machined brass plate clamped over a delectric, over a ground plane. Some cool software read in the desired waveform and programmed an n/c mill to cut the brass.

I think that was used on the Nova laser, the prototype for NIF.

Job security for somebody, for sure. I've made fast pulse networks using an SD-24 TDR head and open or shorted coax stubs. That transmission line stuff really works.

Cheers

Phil Hobbs

Pretty much everything 10(+) bit-ish and high-speed from Analog devices,

like the AD9257 or whatever else that starts with AD92... or AD96...

Dimitrij

The LTC2242 family goes from 125 MHz/10 bits up to the one we use,

Here's one on a board, 12 bits of LVDS into an Altera FPGA, at 250 MHz.

We are seeing less than 1 LSB of RMS noise. That was shocking, especially considering all the switching supplies an inch or so away.

Six of one, a half-dozen of the other... :)

The LT part John mentioned must be such a mixture, with a pipeline delay less than NOB. Of course, it could always be 100% flash with serdes pipelines, but... why?

Subranging, yes, that's a good word for it.

Tim

This is not a pipelined SAR converter. From page 18 of the data sheet...

"Each stage of the pipeline, excluding the last, consists of a low resolution flash ADC connected to a switched-capacitor DAC and an interstage residue amplifier"

Pipelined - yes. SAR - no. This is just an extension of the subranging flash converter I was talking about.

I'm not sure what you are saying. They are clearly pipelined, but flash and not SAR.

Yeah, that's what I mean.

At a glance, here's an example like I was thinking of earlier (and as Dimitrij mentioned):

Tim

I'm not following what you are saying. This is another pipelined flash part, not SAR. It has the exact same text on page 25 saying each pipeline stage has a low resolution flash converter. Maybe they are not describing it well and these are SAR converters. I don't know why it would take 16 clocks for a subranging converter. But I do know they need to do corrections so that may take some of the extra clock cycles.

Seems like the only difference is how maby bits get added per pipeline stage. Back when it was one bit (SAR), N/2 bits (half-flash), or N bits (flash), the distinction was cleared than it is now.

Cheers

Phil Hobbs