DC-coupled AWG

[gkasprow]

I have already a few use cases for general purpose, low-cost AWG. One could use Phaser for that purpose, but most of these applications require DC coupling and output swing of +/-5V or so. Essentially, we need to replace top-bench AWG with modules that are more tightly integrated with ARTIQ. I see two approaches:

• take Phaser, get rid of the upconverter, RF amplifier, and add a high-speed DC-coupled amplifier at its output.
• use the FMC approach, just place the DAC (one used in Phaser) and amplifiers. It seems more feasible because that approach works both in MTCA and EEM, as well as on devkits. It would be a lower cost. Another approach is the possibility to use it with higher-grade FPGAs when higher DAC speed is needed (ZynQ US CPCIS carrier).

I'd like to hear some thoughts and possible use-cases.

[sbourdeauducq]

take Phaser, get rid of the upconverter, RF amplifier, and add a high-speed DC-coupled amplifier at its output.

This. Keep things tidy and neatly integrated.

It seems more feasible because that approach works both in MTCA and EEM

MTCA is not an argument, it is a dead-end due to its ridiculous cost, complexity, festival of IPMI/MCH bugs (with serious consequences, e.g. once I had to pull the power cord as the power supply started smoking because the MCH firmware crashed and had decided to turn off all fans), unreliability (e.g. frequent poor contacts in MCH or RTM connectors), and physical size. M-Labs has no plans to sell or promote any MTCA systems.

It would be a lower cost.

At higher volumes?

[dtcallcock]

I like the idea of keeping momentum behind Phaser development instead of fragmenting it.

[jordens]

FMC

I have a couple use cases that need a DC coupled AWG. GS/s and 5V fits nicely.

[gkasprow]

@sbourdeauducq We are selling quite a lot of MTCA stuff. The thing that worries me the most is that Artix may have too few resources to effectively use such high speed DAC. Especially when we mount the 1.5GS/s version of it. We have another FMC carrier with ZynQ US (where we can use just the logic part). We are currently finishing Artix based FMC carrier. The demand for such AWG won't be high, so the manufacturing cost is also an issue. FMC is simple. Keeping a stock of FMCs costs less than keeping a stock of much more complex boards. For dedicated, high-volume boards like Urukul, Phaser, the integrated approach makes sense because we produce them in batches of a hundred pieces. But for modules that we sell occasionally, fragmentation is a better approach from the logistics point of view. If such AWG reaches a higher volume production level, we will naturally make an integrated version, probably with a higher performance FPGA. The same applies to Shuttler. So let's make it FMC and treat it as a prototype.

[dhslichter]

Another vote for FMC. It seems like this is basically a lower-channel-count, higher-speed cousin of Shuttler. Would make sense to do this as an FMC to test and one can make things fancier later.

[kaolpr]

TS votes for FMC.

[gkasprow]

Very initial component placement, the DC/DC converters will be placed on a mezzanine like in the Shuttler

[jordens]

Maybe worth considering (pin-) compatibility with the DAC eval board and the FMC adapter

[gkasprow]

I was thinking about it. What we can do is to make sure DAC pins are the same. The eval board contains much more chips which we don't need.