Clocking in VLSI Design
Clock PLL and clock network design is a very important step in VLSI
design, particularly in deep sub-micron technologies, in which processing
variations make it harder to design reliable clocking for an IC.
Our work in VLSI clocking has included papers on dynamic de-skewing
approaches for a H-tree clock network (using a single phase detector),
which can reduce skew from 300ps to 3ps. We have also developed
approaches to design extreme high speed free-running as well as
phase-locked standing wave resonant clocking rings. Unlike existing
ring-based resonant traveling wave clocking approaches which exhibit a phase change
around the ring, our approach does not suffer from this problem. Also,
there have been no approaches to phase lock a resonant oscillator,
something which we are able to achieve by a novel coarse and fine tunable
oscillator.
We have also developed techniques to design radiation hardened clocks - focus sing on
not only the analog PLL, but also the local clock regeneration buffers.
Publications, patents and artefacts:
- "Clock distribution scheme using coplanar transmission lines", Cordero,
Khatri. Design Automation, and Test in Europe (DATE) Conference 2008, 10-14
March, Munich, Germany, pp. 985-990. Traditional ring based resonant (rotary) clocks
produce clock signals with varying phase along the ring. This paper
presents a ring based standing wave resonant clock oscillator, which achieves extreme
high speeds, low power consumption and constant phase around the ring. Presentation slides.
- "A PLL Design based on a Standing Wave
Resonant Oscillator", Karkala,
Bollapalli, Garg, Khatri. IEEE International Conference on Computer Design
(ICCD) 2009, Lake Tahoe, CA. October 4-7, 2009. In this paper, we present an extreme high
frequency clock PLL with lock range between 6GHz and 9GHz. The VCO is a
low power ring based standing wave oscillator, and is controlled in two modes (a
coarse tuning mode and a fine tuning mode).
- "Interconnected Tile Standing Wave
Resonant Oscillator based Clock
Distribution Circuits", Mandal, Karkala, Khatri, Mahapatra. 24th
International Conference on VLSI Design, Chennai, India. Jan 2-7, 2011.
This paper reports on techniques to synchronize multiple standing wave
ring-based resonant oscillators, as a means to perform clock
distribution over a large chip, using ring-based resonant clocking
techniques.
- "A Dynamically De-skewable Clock Distribution Methodology", Jayakumar,
Kapoor, Khatri. IEEE Transactions on Very Large Scale Integration (TVLSI),
vol. 16, number 9, Sept 2008, pp. 1220-1229. This paper presents a dynamic
clock de-skewing approach for an H-tree clock network. It uses a single
phase detector for the entire chip, and is able to deskew a clock signal
by two orders of magnitude.
- "An Automated Approach for Minimum
Jitter Buffered H-tree Construction",
Mandal, Jayakumar, Bollapalli, Khatri, Mahapatra. 24th International
Conference on VLSI Design, Chennai, India. Jan 2-7, 2011.
In this paper, we
employ a dynamic programming approach to realize a minimum-jitter clock
distribution network.
- "A Novel Clock Distribution and Dynamic De-skewing Methodology", Kapoor,
Jayakumar, Khatri. International Conference on Computer-Aided Design
(ICCAD), Nov 2004, pp. 626-631. In this paper, we present a dynamic clock
de-skewing technique, which can deskew a clock network to 3ps from a
initial skew of 300ps, with 35% area overhead and a 5% power improvement
compared to a non-deskewed network. Presentation slides.
- "Practical Techniques to Reduce Skew and its Variations in Buffered Clock
Networks", Venkataraman, Jayakumar, Hu, Li, Khatri, Rajaram, McGuinness,
Alpert. International Conference on Computer-Aided Design (ICCAD) 2005, Nov
6-10, San Jose, CA, pp. 592-596. This paper presents a skew reduction
approach for link-based buffered clock networks. It achieves a skew
reduction of 47%, with an improvement in skew yield from 15% to 73%.
- "SEU Hardened Clock Regeneration
Circuits", Dash, Garg, Khatri,
Choi. International Symposium on Quality Electronic Design (ISQED) San
Jose, CA. Mar 16-18 2009. This paper presents a radiation hardened clock
regenerator,
which eliminates radiation induced clock races, and reduces radiation
induced clock skew to 30ps, for deposited charge values up to 150fC.
- "A Radiation Tolerant Phase Locked Loop
Design for Digital Electronics", Kumar, Karkala, Garg, Jindal,
Khatri. IEEE International Conference on
Computer Design (ICCD) 2009, Lake Tahoe, CA. October 4-7, 2009.
In this paper, we present a radiation
hardened analog PLL, in which every node is radiation hardened by using a
novel split output design approach. The worst case radiation induced jitter
is 18%. The PLL requires 16 cycles to lock after the worst case radiation
event, a significant improvement over previous radiation hardened PLLs.