Main Injector Synchronous Timing System

loop that is equal to the cable delay from the LLRF to the local receiver. However, for the Main Injector LLRF such a delay would complicate the LLRF system and might even lead to instabilities.
An added complication is the temperature dependency of the signal carrying me- dium. For example, for a single-mode fiber this can be up to 30 ps/Ckm. With a 10 ºC variance in duct temperature and a cable length of 5 km, the total time differs about 1.5 ns. That is enough of a variation that the Synchronous Timing System needs tempera- ture compensation. The STS design aims to provide a synchronous trigger with timing noise of around a 100 ps RMS. This will satisfy the requirements for instrumentation like dampers or synchronous beam position sampling systems and support beam insta- bility studies.
The following sections in this paper describe the reconstruction of the phase, the signal distribution infrastructure, and the design of the receiver and transmitter boards.

SYNCHRONOUS TIMING

The STS must correct the phase to account for the difference in travel times between the reference signal and the particle speed as well as the variations in the group delay due to temperature changes. This chapter first analyzes the phase as a function of frequency (RF) followed by an analysis of the phase as a function of temperature.

Phase as a Function of Frequency

The LLRF is trying to keep the particles at the same phase for each turn no matter what the frequency, as the number of buckets remains the same. This is true for any lo- cation around the ring. As the frequency increases, the phase of the reference signal shifts as more cycles fit within the same cable length. Therefore the local receiver must "unwind" the phase of the RF signal to construct a synchronous phase. This is done by subtracting a delta phase that is the product of the change in frequency and the time-of- flight of the RF signal. In addition, a phase intercept is added to align the phase of the reconstructed RF with the phase of a detector:

j_sync=j _RF- Dj+j _intercept

(1)

Dj=Dw[!]t

with Dw=w-w _base, w _base= radian frequency at beginning of cycle,t= time-of-flight of RF, and j_intercept= phase intercept.

Phase as a Function of Temperature

The temperature dependency is modeled as an extra delay in the cable as a function of temperature:

Dt=(v(T
0^)-v(T
1^))[!]l

(2)

with l= length of cable, T_i= temperature at time i, and v(T_i) = signal speed at T_i.

The Dtwill be measured indirectly by comparing the phase of the RF signal gener- ated by the transmitter board and the phase of the RF signal received after it has gone around the ring. The receiver for this signal will be on the same board as the transmitter. The phase shift due to the Dtis:

Dj _temp= Dw[!] Dt

(3)

Assuming that the distribution of the temperature dependent phase shift is linear along the signal path then the local phase shift is:

Dj ^local= Dj ^total
temptemp^[!]

(4)

with the value of a variable at an arbitrary location around the ring indicated by the su- perscript localand the location back at the LLRF after a full turn indicated by the super- script total.

The change due to temperature variation is very slow compared to the change due to frequency sweeping. The cable ducts are buried 4 to 7 feet and should change only about 10 ºC over the seasons. Fast changes between the generated and received phase will be regarded as errors that should be diagnosed. Even if the temperature varied by 1 ºC over a day this would only be a change of 0.1 ps per minute.

Reconstructed and Temperature Compensated Phase

The compensating phase shift to be locally added to the RF signal is the sum of the frequency phase slip and temperature compensation:

Dj^local= Dw[!]t^local+ Dj ^total
temp^[!]

(5)

This can be rewritten so that only one term, Dw+ ^{, must be transmitted:}

Dj ^local=^[!]
[!]
[!] ^Dw+

[!]t^local

(6)

SIGNAL DISTRIBUTION

The distribution of the signal from the transmitter to the receivers is depicted in fig- ure 1. The LLRF system provides the transmitter with a Voltage Controlled Oscillator

(VCO) synchronized strobe and the required digital information. We choose single mode fiber as the medium because of its low cost, low attenuation, and high bandwidth. The low cost, $1 per foot per 24 fibers, gives it a price advantage over copper cables espe- cially counting the number of channels per cable. The low attenuation, 0.35 dB/km, makes it possible to passively split off a part of the signal at each of the 10 service buildings while avoiding the use of amplifiers and associated noise. Figure 2 depicts the loss calculation and includes losses due to connectors, splitters, and 5 km of cable. The high bandwidth makes it possible to use the G-Link chip from Hewlett Packard [2] which supports a stream up to 1.2 Gbit/s. The STS will use the RF (53 MHz) as a clock and the 20 bits per clock cycle are used to carry the digitally encoded information. Fini- sar Optical modules [3] convert the G-Link signal to 1310 nm light waves and back. The STS will have for redundancy reasons two fibers with an active signal and a third fiber prepared for immediate use. The remaining 21 fibers are for future use.
The transmitters and receivers are VXI modules that, through the slot 0 processor, communicate with consoles using Fermilab's Accelerator Control Network (ACNET) protocol. This link is used for initialization, diagnostics, and calibration purposes.

FIGURE 1.The signal distribution.

FIGURE 2.The passive splitting of a signal at each of the 10 service buildings. The top numbers indicate the splitter ratio while the top (dotted) line is the optical power in the incoming fiber to the service building. The lower (solid) line is the optical power to the receivers at each location with the name of the service building below the solid line. This calculation assumes a 1 dBm transmitter with a minimum receiver power requirement of -16 dBm.

RECEIVER AND TRANSMITTER DESIGN.

Besides correcting the phase, the STS must be able to identify a particular bunch or otherwise a damper would not know which bunch to kick. A marker reset is generated by the LLRF on injection of beam into the Main Injector. The transmitter sends this marker to the receivers which then reset a local bucket counter. This counter counts each cycle (read bucket) of the synchronized RF and resets when all 588 buckets passed by, thus always providing the same counter value for the same bucket.

Resolution Considerations

The maximum frequency change between updates, 10 µs, will be 20 Hz in the Main Injector, given the 2 MHz/s slew rate. This corresponds to about a 10 ps peak to peak jitter at the maximum cable delay of 25 µs.
The G-Link chip can transmit 20 bits per frame at a frame rate of 53 MHz. As the STS must transmit different messages, 4 bits are used to identify the type of message. This leaves 16 bits to describe the frequency sweep. Given a range of 500 kHz, (300 kHz is the current sweep range, 200 kHz extra range), the 16 bit word has a resolution of 500 kHz/65536 or 7.6 Hz. With the delay of 25 µs this results in a jitter of 4 ps peak to peak. These (uniform white noise) quantization errors lead to a RMS noise of [!]((10²+ 4²)/12) [!]3 ps. This is small compared to RMS noise from the G-Link chip, about 42 ps. We do expect contributions from other parts of the electronics but not so high that we will come above a 100 ps RMS noise total.
The RF signal from the LLRF, however, is not perfect. It is estimated that, depend- ing on what cycle is running in the Main Injector, there is about a 100 ps of RMS noise in this signal that is relevant to a damper system. Improvements on the quality of the RF signal will be for a future project.

The Receiver

The design of the receiver is shown in figure 3. The STS signal comes in over the fiber and is converted into a parallel stream of 20 bits data frame with a 53 MHz strobe by the Optical Receiver consisting of a Finisar FRM 1310 and a G-Link module. Diag- nostic data from the G-Link and Finisar module are accessible over the VXI interface. The strobe is used as a clock to the Phase Shifter and Phase Calculation. The frame is decoded on an Altera chip into the correction term of equation (6) or a marker reset event. The Phase Calculation performs the multiplication in equation (6) and adds two more phase offsets. One is the intercept phase, j_intercept, to align the synchronous phase with the phase of a beam pickup. The other is a phase from a local analog signal that could be part of a PLL on a beam signal to help improve the synchronous phase. ACNET can access the local intercept phase, j_intercept, and time-of-flight ,t_local, values over the VXI interface for calibration setup. The Phase Shifter is an I&Q modulator cir- cuit. The Altera chip implements the sine and cosine functions of the corrected phase. These signals are converted to analog to be multiplied by the quadrature components of the strobe signal, summed together and bandpass filtered. The resulting signal is used to pulse the counter of the Pattern Generator and used as the synchronous RF output sig- nal. The Pattern Generator contains a single byte wide memory bank with one byte for each bucket and one bit for each of the 8 output Triggers. The Pattern Generator's mem- ory is set using ACNET. To allow easy calibration and diagnostics, the correction term, the calculated Dphase shift, and the local signal are available through the Multiplexer (MUX) unit and the D/A converter as a analog signal or through the VXI interface as a digital value.

FIGURE 3.Diagram of the receiver.

The Transmitter

The transmitter module, its diagram shown in figure 4, provides the optical signal used by the receiver modules. This signal is synchronized with the RF by the VCO of the LLRF and contains information on the frequency value, the change in fiber delay due to temperature, and the beam marker reset. The module is located close to the Main In- jector LLRF system, and it contains the same DSP, an Analog Devices' SHARC, as the LLRF system [4]. Using the same DSP simplifies communication between the two sys- tems and also maximizes data transfer speed. The LLRF DSP transfers frequency in- formation to the transmitter DSP through a fast Link Port.
Each transmitter module contains all the components of a receiver module except the pattern generator and the A/D converter. This receiver decodes the transmitter signal which has traveled the full circumference of the ring. The Phase Detector of the trans- mitter compares the synchronized output of the receiver module with the RF input of the transmitter module. Errors are accumulated and used to modify temp_adj. The Phase Cal- culation, implemented on the DSP, uses the digital frequency, freq_LLRF, the temp_adj, and the time-of-flight, to calculate the correction term. The marker reset is sent directly to the encoder to avoid processor delays which might disrupt the single bucket precision of the signal. Access to the receiver and the t^totalvalue is handled by ACNET.

FIGURE 4.Block Diagram of transmitter.

CONCLUSIONS

We have designed a timing system that delivers a trigger synchronous to the beam during acceleration. The infrastructure has been installed and work to layout the trans- mitter and receiver boards is in progress. We expect that the noise added by the STS to the synchronous trigger will be less than 100 ps RMS. As the RF signal has about a 100 ps RMS noise to it, the total noise is estimated around 140 ps. This can be improved

locally by locking to a signal from a beam pickup for those applications that don't re- quire first turn sampling. In the future we hope to improve RF signal from the LLRF to further reduce the noise in the synchronous trigger.

ACKNOWLEDGEMENTS

Many thanks go to Brian Chase, Jim Crisp, and Keith Meisner for ideas and rec- ommendations on the design issues.

REFERENCES

[1]

[2]

[3]

[4]

Steimel, J., "Trigger Delay Compensation for Beam Synchronous Sampling," 7th Beam Instrumentation Workshop (BIW 96), Argonne, IL, May 1996, pp 476-482. Yen, C.S. et al., "G-Link: A Chipset for Gigabit-Rate Data Communication," Hewlett-Packard Journal, 43(3), 103-116, (October 1992).
"FTM/FRM-8510 Low Cost Gigabit Optical Transmitter/Receiver," Finisar Corpo- ration, Mountain View, California.
Chase, B. et al., "Current DSP Applications in Accelerator Instrumentation and RF," presented at the International Conference on Accelerator and Large Experi- mental Physics Control Systems (ICALEPCS'97), Beijing, China, Nov. 3-7, 1997.