1. Latency measurements on the ALBA Electron units
Olmos
2nd Libera Users Meeting
3rd - 4th April 2008

2. 2nd Libera Users Meeting
Latency Measurements
Latency as Libera response time: “the time from the beginning of the step response up to the moment when the 90 % of the stable state output is reached”
Three different setups were analyzed to measure and compare the latency
First two setups were proposed by I-Tech (maybe the first by I-Tech and the second by some of you)
The last one includes a variation from I-Tech proposals
Latency basically depends on the FA processing, that for ALBA Electron is …
1.1153MHz
Dec = 112
9.96 kHz
Dec = 2
SW Dec = 84
Dec = 56
~1.6kHz
~3.3kHz
LPF -3dB / 2kHz
2nd Libera Users Meeting
A. Olmos

3. 2nd Libera Users Meeting
FA data counter
Based on the analysis and comparison of FA and Turn-by-Turn samples
Target is to create a step change on any of the Libera inputs and measure how long does it take to propagate to the output
PROCEDURE
Time-stamp the TbT sample at the trigger (when step function starts)
Look for that sample in the FA data stream
Counter_Value = MOD(MT_time_stamp/112 , 2^16)
Measure the latency as the time that it takes from that sample until 90% of the final output step level
2nd Libera Users Meeting
A. Olmos

4. 2nd Libera Users Meeting
FA data counter
2nd Libera Users Meeting
A. Olmos

5. 2nd Libera Users Meeting
FA data counter
FA Voltage C
~280µs
FA Voltage C
~420µs
This latency measurement setup seems to be not very precise because of its
wide range of latency results and poor resolution level (100µs between FA
samples)
2nd Libera Users Meeting
A. Olmos

6. By Measuring Interlock Signal (I) 2nd Libera Users Meeting
Based on the analysis of the interlock signal generated by the Libera
PROCEDURE
Interlock signal is obtained from the FA data stream
A steep change in the beam position out of a determined threshold will trigger the interlock
Latency is measured as the time that Libera takes to trigger the interlock due to a position change out of threshold
There’re two scenarios: Corrected and Non-Corrected position offsets
2nd Libera Users Meeting
A. Olmos

7. By Measuring Interlock Signal (I)
Non-Corrected position offset
Non Gated (Int OFF)
Non Gated (Int OFF)
Gated (Int ON)
2nd Libera Users Meeting
A. Olmos

8. By Measuring Interlock Signal (I)
Non-Corrected position offset
C Libera input signal
208µs
97µs
Interlock output
Xoffset=690µm
Zoffset=784µm
One FA sample
2nd Libera Users Meeting
A. Olmos

9. By Measuring Interlock Signal (I)
Corrected position offset
2nd Libera Users Meeting
A. Olmos

10. By Measuring Interlock Signal (I)
Corrected position offset
193µs
314µs
2nd Libera Users Meeting
A. Olmos

11. By Measuring Interlock Signal (II) 2nd Libera Users Meeting
Simpler: there’s no need of compensation attenuators
Scope triggered with the gate control signal
Setting the scope in persistence mode allows “automatic” max/min latency measurements for a determined interlock threshold
Again two scenarios: Corrected and Non-Corrected position offsets
2nd Libera Users Meeting
A. Olmos

12. By Measuring Interlock Signal (II)
Non-Corrected position offset
Non Gated (Int OFF)
Gated (Int ON)
2nd Libera Users Meeting
A. Olmos

13. By Measuring Interlock Signal (II)
Non-Corrected position offset
Gate control signal
Interlock output
152µs
Xoffset=174µm
Zoffset=220µm
252µs
2nd Libera Users Meeting
A. Olmos

14. By Measuring Interlock Signal (II)
Corrected position offset
173µs
268µs
2nd Libera Users Meeting
A. Olmos

15. Summarizing … Which setup should we trust?
I-Tech reports
Comparable with our MIN results
for very off-centered beam
DIAMOND Brilliance unit
Non-Corrected offset ~300µm
Xoffset=690µm Zoffset=784µm
What’s different in ALBA Electron unit?
-CIC filter contribution
-Two Notch filters instead of only one
 negligible assuming Notch
contribution ~10µs
Which setup should we trust?
Have you measured “your latency”?
-Two Notch filters instead of only one
 negligible assuming Notch
contribution ~10µs
2nd Libera Users Meeting
A. Olmos

16. Beam orbit control estimation 2nd Libera Users Meeting
Summarizing …
Beam orbit control estimation
Fast orbit feedback loop (Beam dynamics / Controls calculations)
fH=235 Hz
fV=1550 Hz
- ALBA sextupoles have extra coils to provide H/V beam correction
- Eddy currents reduce the effect of the magnetic field at high frequencies
The overall latency of the control loop is µs
µs on Libera electronics (very worst case)
~50µs distribution of data around the network
~100µs for DMA transfer + correction calculation + writing into power converters
Vacuum chamber cutoff fH=235 Hz and fV=1550 Hz  LIMITING FACTOR
Beam orbit control will damp oscillations up to 100Hz
Vacuum chamber cutoff fH=235 Hz and fV=1550 Hz  LIMITING FACTOR
Beam orbit control will damp oscillations up to 100Hz
2nd Libera Users Meeting
A. Olmos