Filip Tavernier Karolina Poltorak Sandro Bonacini Paulo Moreira

1. A Radiation-Hard PLL for Frequency Multiplication with Programmable Input Clock and Phase-Selectable Output Signals in 130 nm CMOS Filip Tavernier Karolina Poltorak Sandro Bonacini Paulo Moreira

2. Outline • Introduction • ePLL topology • ePLL design aspects • Measurement results Filip Tavernier - CERN

3. The bigger picture On-Detector Custom Electronics & Packaging Radiation Hard Off-Detector Commercial Off-The-Shelf (COTS) Custom Protocol Development of a high-speed bidirectional optical link for the LHC experiments upgrade program: • Versatile link project: opto-electronics • GBT project: ASIC design, especially for on-detector systems Filip Tavernier - CERN

4. The GBT system External clock reference FEModule Clock[7:0] E – Port e-Link GBTX Phase - Shifter CLK Reference/xPLL E – Port FEModule E – Port ePLLRx GBTIA DEC/DSCR CDR E – Port data-down data-up Phase – Aligners + Ser/Des for E – Ports CLK Manager clock 80, 160 and 320 Mb/s ports GBLD SCR/ENC SER E – Port ePLLTx FEModule E – Port E – Port Control Logic Configuration (e-Fuses + reg-Bank) one 80 Mb/s port GBT – SCA JTAG I2C Slave I2C Master E – Port data I2C (light) control clocks JTAG port I2C port Filip Tavernier - CERN

5. 2 design cases for the ePLL • ePLL for front-end modules • base design of the ePLL family • intended to provide clocks to the front-ends with a different frequency than the LHC clock or what is provided by means of the e-links • very flexible operating conditions: • input clock: 40, 80, 160 or 320 MHz • output clock: 40, 80, 160 and 320 MHz , programmable phase • taped-out last year, partly measured this year • main subject of this presentation • ePLL for GBTX • fine-tuned version to be integrated in the GBTX chip and especially intended for duty-cycle improvement of existing clocks • somehow less flexible operating conditions: • input clock: 40, 80 or 160 MHz, but optimized for the latter • output clock: 160 and 320 MHz • taped-out in August as part of the GBTX Filip Tavernier - CERN

6. Outline • Introduction • ePLL topology • ePLL design aspects • Measurement results Filip Tavernier - CERN

7. ePLL topology (1) • input divider: programmable divider ratio so that the output is always 40 MHz • feedback divider: fixed divider ratio of 8 because the VCO runs at 320 MHz • PFD: determines the frequency and phase difference between the 40 MHz input clocks and outputs UP and DOWN signals which are linearly dependent on the phase difference Filip Tavernier - CERN

8. ePLL topology (2) • CP: charge pump with a 6-bit programmable output current to convert the UP and DOWN signals of the PFD to the current domain • LPF: 1st order loop filter referred to the supply voltage because of the pMOS current sources in the VCO • VCO: voltage-controlled ring-oscillator at 320 MHz with 16 output phases Filip Tavernier - CERN

9. Outline • Introduction • ePLL topology • ePLL design aspects • Measurement results Filip Tavernier - CERN

10. Phase-frequency detector • linear relationship between the duration of the early/late pulses and the phase error • PFD is sensitive to phase and frequency errors which leads to a PLL with a locking range that is basically limited by the VCO tuning range • delay of the NOR gate results in the early/late pulses go high every clock cycle → ‘4’ possible output states of the PFD • ‘translation’ of signals taking into account the pMOS current source in the VCOearly → VCO is too fast → DOWN → source current into the loop filterlate → VCO is too slow → UP → sink current from the loop filter Filip Tavernier - CERN

11. Charge pump (1) • UP = low and DOWN = low • no current flows to the loop filter or in the unity-gain amplifier • UP = high and DOWN = high • no current flows to the loop filter if sink and source currents are equal! Filip Tavernier - CERN

12. Charge pump (2) • UP = high and DOWN = low • sink current flows out of the loop filter, source current flows into the amplifier • UP = low and DOWN = high • source current flows into the loop filter, sink current flows out of the amplifier Filip Tavernier - CERN

13. Charge pump non-idealities • unequal sink and source currentsThe loop filter is charged or discharged by the current difference when UP and DOWN are both high every clock cycle. → static phase error→ solution: current sources with long length and consequently high output impedance • parasitic capacitance at the drain of the current sourcesCharge sharing takes place every time one of the current sources is connected to the loop filter→ static phase error→ solution: unity-gain amplifier to equalize the drain voltage of the current sources to the control voltage Filip Tavernier - CERN

14. Voltage-controlled oscillator • 8-stage differential ring-oscillator • 8 D2S converters to generate the full-swing output phases (0°, 22.5°, 45°, 67.5°, 90°, 112.5°, 135°, 157.5°) from the analog levels in between the delay stages • inverters used to generate the other 8 phases to save the power of another 8 D2S converters • transmission gates to equalize the phases generated by the D2S converters directly and the ones generated by the inverters Filip Tavernier - CERN

15. VCO – delay cell • pMOS differential pair with pMOS current source → control voltage referred to the supply voltage • input transistors biased with small overdrive voltage because of the relatively large cell delay of 195.3125 ps • active load consisting of a current source and a diode to avoid the need for a common-mode feedback circuit and still have enough small-signal gain • gate voltage of the nMOS current sources derived by means of a replica biasing circuit Filip Tavernier - CERN

16. Radiation hardness (1) • total ionizing dose • all transistors have a gate width of at least 1 µm → no leakage • 130 nm CMOS is generally known for its good TID radiation hardness • single-event upsetsVCO is most critical because it can take a long time to recover after an SEU → simulated with current injected in 3 sensitive nodes Filip Tavernier - CERN

17. Radiation hardness (2) maximum phase error of 533 ps (current I3, 0.3 pC in 10 ps) ↓ loop corrects this in ± 2 clock cycles Filip Tavernier - CERN

18. Outline • Introduction • ePLL topology • ePLL design aspects • Measurement results Filip Tavernier - CERN

19. Some numbers • 130 nm CMOS • size of the test chip: 2 mm x 1 mm • size of the core circuit: 385 µm x 325 µm Filip Tavernier - CERN

20. Measurement of the output phases (1) Filip Tavernier - CERN

21. Measurement of the output phases (2) If needed, the 40, 80 and 160 MHz output clocks can be resampled with any of the320 MHz output phases in order to increase the phase resolution to 22.5°. Filip Tavernier - CERN

22. Measurement of the output jitter (1) jitter values < 10 psRMS are possible Filip Tavernier - CERN

23. Measurement of the output jitter (2) jitter values < 10 psRMS are possible Filip Tavernier - CERN

24. Measurement of the jitter transfer Filip Tavernier - CERN

25. Conclusions • a highly flexible ePLL has been presented • programmable input clock frequency of 40, 80, 160 or 320 MHz • output clocks of 40, 80, 160 or 320 MHz are always available • output phase can be programmed with a resolution of 90° for the 40, 80 and 160 MHz outputs and 22.5° for the 320 MHz output • measurements have shown a good jitter performance Filip Tavernier - CERN