Preliminary Design Review Plan
This is an outline of the PDR scheduled for 10:30 AM January 31st, 2010. All times are rough estimates based on the current level of development each section is in, and the amount of detail available. We will allow an additional 15 minutes or so per talk for questions and suggestions. This will come to about 4-5 hours total.
Introduction and Overall project plan (20 Minutes)
Marty Bloss will introduce the panel, present the project plan, including when it started, when it should end, and the schedule that has been laid out so far.
Spectrometer specification and overall design to achieve spec compliance (30 minutes)
- Anish Roshi will present the Spectrometer specifications.
- John Ford will present the design.
ADC characterization (30 minutes)
Srikanth Bussa will report on the ADC Characterization work.
Analog and clock distribution detailed design (30 minutes)
Galen Watts will present designs for the analog routing and clock distribution.
1 PPS and Switching signal synch and distribution plan (15 minutes)
will present on distribution of all the switching signals.
Firmware design and development strategies (45 minutes)
Dan Werthimer presnents details about designs that are being done at UC Berkeley.
Software design and development plan (45 minutes)
Amy Shelton will present the conceptual design for the software to integrate and control the Spectrometer within the NRAO environment.
Project Risks (20 minutes)
John Ford will review the project risks.
Review (30 minutes)
Response to the committee report
(FWIW, This by J. Ford. Nobody else has blessed it...)
- "... need significant expansion to accomodate larger multi-pixel receivers"
- This spectrometer is not intended for large-format arrays. New technology will need to be developed for that.
- "Bandwidth for Ka observations is still slightly less than existing Zpectrometer. This will need to be explained well in the user documentation."
- This instrument was not designed to replace the coarse, wideband analog Zpectrometer. If that is needed, then we should regroup.
- "Doppler Tracking seems to be adding complexity..."
- The team failed to explain this well to the reviewers. The doppler tracking facilities are in the computing pipeline, not in the FPGA firmware. It will not be done in firmware.
- "There may be still too many modes to complete all of the work on time and on budget. The project might benefit from a review of the prioritization of lower-priority modes by a broader discussion of relative importance"
- Agreed. Would the panel agree to undertake such a prioritization?
- "While the Berkeley/Casper track record for projects is good, there is some concern about the transfer of knowledge to NRAO for maintenance, future enhancements, etc."
- We agree. The best way to accomplish this is to have the teams work together, and for NRAO to take the initiative to assign sufficient personnel to accomplish this task.
- "Berkeley is new to using GPUs for spectroscopy - NRAO did that work for GUPPI. How are the throughput numbers being validated?"
- This is being validated through analysis and prototyping.
- "Will data read/write speeds be a bottleneck to observer efficiency? Will there be extended periods waiting for the data to be available after a scan? The reviewers encourage the team to begin work on the data stream-related issues in the project as soon as is practical as these may represent a significant portion of the software work possibly illustrating insufficient resources have been assigned to the project for this work"
- This is a question requiring much discussion. The whole data-stream business is a new addition to the scope of the project, and was not in the original plan. On the performance questions, there is no reason to believe this system will suck.
- "The reviewers feel the project needs to establish a firm date for a decision on Roach 2 so that any duplication of effort caused by joint Roach 1&2 development or any other dependencies can be resolved."
- We agree with this. We showed a deadline of ~August for this decision. We have structured the program so that minimal rework will be required in either case.
- "An additional risk should be added as how to disseminate data to the users."
- This is not the province of the spectrometer project. The data will be available on a disk subsystem on the GB network for users. No other facilities are planned or funded.