New PSB Beam Control Clock issues

1. Participants: Pablo Alvarez-Sanchez, Maria Elena Angoletta, Alfred Blas, Petri Leinonen, John Molendijk, Jorge Sanchez-Quesada, Maarten Schokker, Javier Serrano, Erik Van Der Bij New PSB Beam ControlClock issues Working group meeting 01/02/2010

2. Clock source New PSB Beam ControlClock issues Working group meeting 01/02/2010

3. Clock in its context (typical simple loop) New PSB Beam ControlClock issues Working group meeting 01/02/2010

4. Clock Jitter requirements To take full advantage of the ADC/DAC dynamic range (N bits), the peak-peak jitter amplitude should be limited to: http://www.nanophon.com/audio/jitter92.pdf (author: Julian Dunn) New PSB Beam ControlClock issues • ωi is the expected bandwidth of the system • For a 16 bits ADC and a 10 MHz sampled signal , J should be less than 1 psP-P • For a 14 bits ADC and a 10 MHz sampled signal , J should be less than 4 psP-P Using a 16 bit conversion is typically to obtain a 14 bit dynamical performance (> - 84 dBFS of spurious) => Jitter < 4 psP-P Working group meeting 01/02/2010

5. Elements influencing the jitter within the clock generation chain 10 MHz Clock source: > 0.36psRMS phase jitter integrated from 12kHz to 20MHz 10 MHz -> 1 GHz PLL: ADF4106 (-53dBFS with 10kHz BW = 9 bit eq. - 84 dBFS with 10 Hz BW = 14 bit eq.) New PSB Beam ControlClock issues Following John’s advice, I went to see Donat and Julien who advised me to use ADIsimPLL , a free PLL simulator from analogue devices. From it, it seems like we can achieve -110 dBc/Hz (I didn’t try to optimize this value). => -84dBFS obtained with 400Hz BW = 14 bit eq. Working group meeting 01/02/2010

6. Elements influencing the jitter within the clock generation chain MDDS: AD9858 using a 10 bit DAC with AD9515 as a clock distribution unit (not used presently) New PSB Beam ControlClock issues Conclusion: Within the clock source circuitry, the PLL is the main limiting factor Working group meeting 01/02/2010

7. Setup used to measure the MDDS jitter New PSB Beam ControlClock issues Working group meeting 01/02/2010

8. FPGA performances -Stratix 3 New PSB Beam ControlClock issues Working group meeting 01/02/2010

9. FPGA performances -Stratix 3 New PSB Beam ControlClock issues Working group meeting 01/02/2010

10. FPGA performances -Stratix 3 New PSB Beam ControlClock issues Conclusion: In our context, the FPGA is not suited for clock management (250psP-P with BW = 10 MHz correspond to a 8 bit signal quality) Working group meeting 01/02/2010

11. Clock and tag to be distributed New PSB Beam ControlClock issues At 40 MHz, hclock = 4 => frf= 10 MHz, one clock period (25 ns) represents 90o of the rf phase. => We do want to have an uncertainty in the acquisition of the tag, to avoid 90o phase jumps in the feedback system • Maximum tolerated skew between tag and clock = +/- 3ns •  The tag signal should be routed the same way as the clock Working group meeting 01/02/2010

12. Summary: The clock and tag should be created with a specific low-jitter electronic circuit (no FPGA) and routed along the same path This means a complete re-design of the MDDS mezzanine! New PSB Beam ControlClock issues Here Javier believes the tag could be routed as a data on a data line within the FPGA. He believes that one could take advantage of an automatic phasing circuit within the FPGA that delays the data line so as to avoid indeterminations when latched. This idea needs to be studied taking into account the rate of change of the clock and the response time of the phasing loop. Working group meeting 01/02/2010

13. New PSB Beam ControlClock issues Working group meeting 01/02/2010

14. New PSB Beam ControlClock issues Working group meeting 01/02/2010

15. New PSB Beam ControlClock issues Working group meeting 01/02/2010

16. Jitter on the sampling of an analogue signal can be interpreted as the sampling without jitter of a signal where the time value is modulated. http://www.nanophon.com/audio/jitter92.pdf (author: Julian Dunn) New PSB Beam ControlClock issues If J.ωi is small => => A sinusoidal modulation at ωj of the time value of a purely sinusoidal waveform at ωi corresponds in the frequency domain to the creation of two side-bands ωj apart from the “carrier” at ωi . The amplitude of these side-bands, relative to the main signal is J[sp-p].ωi/4 or 20.log(J.ωi/4) dBc. Working group meeting 01/02/2010

17. Example: for a 100 ps peak-peak sinusoidal jitter and a 10 MHz sampled signal, the modulation sidebands will be at -56 dBc (corresponds to a 9 bit SNR). To take full advantage of the ADC dynamic range (N bits), the peak-peak jitter amplitude should be limited to: New PSB Beam ControlClock issues • For a 16 bits ADC and a 10 MHz sampled signal , J should be less than 1 psP-P • For a 14 bits ADC and a 10 MHz sampled signal , J should be less than 4 psP-P Working group meeting 18/09/2009

18. Brad Brannon, "Aperture Uncertainty and ADC System Performance," Applications Note AN-501, Analog Devices, www.analog.com. If the input signal to be sampled is VIN=AIN.sin(ωIN.t) The variation of VIN with respect to time is expressed as: dVIN/dt = AIN.ωIN.cos(ωIN.t) If the sample is acquired terr after its expected arrival time, the maximum error on the acquired amplitude is: Verr max = AIN.ωIN.terr If terr is expressed as an RMS value, so is Verr… Although Verr max is defined as the error at the input maximum slew, this simple model proves surprisingly accurate to estimate the degradation due to jitter. So, the assumption on terr is that it has a Gaussian amplitude distribution and the spectrum of white noise. In the time domain, sampling corresponds to a multiplication by a succession of unit pulses spaced by TS and in the frequency domain, to a convolution of the two spectra. New PSB Beam ControlClock issues The spectrum of the acquired signal is thus modulated in amplitude and frequency with terr as the modulating term. If Working group meeting 18/09/2009

19. New PSB Beam ControlClock issues Working group meeting 18/09/2009

20. FPGA performances => Can we use an FPGA as a clock gating unit? Many of the traditional methods manufacturers use to specify clock jitter do not apply to data converters or, at best, reveal only a fraction of the story. It is important that you know the bandwidth and spectral shape of the clock noise so that you can properly account for them during sampling. picoseconds of clock jitter quickly translate to decibels lost in the signal path.Many trade-offs relate to jitter, phase-noise, and converter performanceIt is difficult to provide a direct correlation between clock jitter and phase noise, but some guidelines exist for designing or selecting encoding sources from either a clock-jitter or a phase-noise perspective. Random jitter is characteristically gaussian; the rms time value or standard deviation of the occurrences specifies this random jitter. When considering the bandwidth of the noise that constitutes jitter for a data converter, the range is from dc to the encoding bandwidth, which exceeds far beyond the typical 12 kHz to 20 MHz that vendors often quote for standard clock-jitter measurements. Because the concern with jitter is increased wideband-converter noise, it is easy to estimate clock jitter by observing the degradation in noise performance of a converter. Equation 1 defines the SNR limitation due to jitter: EQUATION 1 SNR=-20log(2.pi.fanalog.tjitterRMS)dB where f is the analog input frequency and t is the jitter. Solving this equation for t puts it in the form of Equation 2, which defines the clock-jitter requirement, given a frequency of operation and an SNR requirement: New PSB Beam ControlClock issues Working group meeting 18/09/2009

21. MDDS => What should be taken care of when designing a DDS? New PSB Beam ControlClock issues Ref: http://www.analog.com/static/imported-files/application_notes/475354741144165304775709740692131461831AN823_0.pdf Author: David Brandon Working group meeting 18/09/2009

22. MDDS => To minimize noise coupling, it is advantageous to have a 2-wire, balanced connection between the DAC output and the limiter input. the real-world spectrum of an unfiltered DDS output is rich in spurious content. It contains DAC related harmonic distortion as well as the images of the fundamental Frequency. The images of the fundamental reside above the cutoff frequency of the reconstruction filter. However, the harmonics of the fundamental are also reproduced in the images. Images that extend into the 1st Nyquist zone (DC to ½ fc), appear as aliased versions of the harmonics. Thus, the images may appear within the filter pass band. These in band images of the DAC harmonics as well as out-of-band images not sufficiently attenuated by the filter can contribute significantly to the jitter observed on the output of the limiter The jitter present at the output of the limiter is due to spur-induced, cycle-to-cycle, modulation of the time interval at which the limiter threshold voltage is crossed. The jitter that is generated by this process is classified as deterministic jitter; it is related to the specific frequencies of the spurious content of the signal. The process of spurious components being converted to a phase, or timing error (jitter) via the limiting function is referred to as AM to PM conversion. If the bandwidth of the filter is reduced, either by using a low-pass filter with a reduced cutoff frequency, or a band-pass filter, the amount of spurious noise is reduced. Bandwidth limiting this spurious noise in turn reduces the magnitude of time jitter that is produced. The magnitude of the time jitter produced is proportional to the magnitude of the spurious components with respect to the slew rate of the fundamental signal. Noise coupling between the DAC output and the limiter input, including device noise from the limiter itself, can also contribute to increased jitter. In general, increased slew rate at the input to a limiter translates to less sensitivity to jitter induced by coupled noise. Because slew rate is proportional to frequency and amplitude, an increase in either parameter tends to improve jitter performance. The key points in minimizing jitter in DDS-based clock systems are maximizing the DAC output slew rate and implementing effective filtering of the DDS spurious components New PSB Beam ControlClock issues Working group meeting 18/09/2009

23. MDDS From the ADC specs and the SNR measured on an FFT, one can derive the jitter of the input clock: New PSB Beam ControlClock issues Working group meeting 18/09/2009

24. MDDS Jitter data was taken for three frequencies. At each of these frequencies three different filter configurations were used to demonstrate the impact of slew rate (frequency) and bandwidth limiting on the measured jitter. The power of the output level for each setting is also shown in order to monitor the slew rate at the limiter input for each test condition. The AD9958 is a DDS using a 10 bit DAC and the AD9515 is a clock distribution unit. New PSB Beam ControlClock issues Working group meeting 18/09/2009

25. MDDS Setup used to measure jitter New PSB Beam ControlClock issues To further confirm the strong dependence on slew rate, additional data is taken with the AD9858 DDS (our present circuit in the MDDS), which is capable of delivering 40 mA of output current to a 50 Ω load. This enables an increased output power relative to a 50 Ω load and the associated greater slew rate. The AD9958 and AD9959 are multichannel DDS devices, and their outputs can be summed together to increase output power. Working group meeting 18/09/2009

26. MDDS New PSB Beam ControlClock issues We need to know how a 14 bit DAC would improve the results The THS4302 (instead of the AD9515 in the tables) used as a comparator needs to be measured in terms of jitter. The AD9515 has the advantage of the dividers and of 2 channels. Working group meeting 18/09/2009

27. MDDS New PSB Beam ControlClock issues We need to know how a 14 bit DAC would improve the results The THS4302 (instead of the AD9515 in the tables) used as a comparator needs to be measured in terms of jitter. The AD9515 has the advantage of the dividers and of 2 channels. Working group meeting 18/09/2009