BDOT Study Report 12052019 Jeff Larson Howie Pfeffer
BDOT Study Report 12/05/2019 Jeff Larson, Howie Pfeffer, Chris Jensen, Kent Triplett
Today we want to consider: Reduction of injection pulse jitter with respect to IMIN • Nothing is rock solid. • The line frequency is continually changing. • GMPS frequency, though very close to 15 hz, also varies. • Has been an on going effort. This year the jitter was reduced from 40µS to 6µS by use of crystal clock time delay. Quest to find substitute for the BDOT IMIN timing signal based on the coil in Reference Magnet. We have looked at: • V-Dividers both passive and active on select GMPS girder magnets. • Pick up coil placed along the bus • Rogowski coil placed around the bus If suitable substitute for BDOT is found: • Would allow us to remove Reference Magnet • Would achieve a more regular 48 cell structure for GMPS • Would make application of mode damping resistors more straightforward.
Jitter Example of how line frequency changes by 60 m. Hz over an hour. 0 on plot corresponds to 60 Hz, units are in millihertz Typical meandering of $0 F event times vs $0 C event times Application by Mike Kuplic Booster 15 Hz tracks the line frequency by locking GMPS reference to every fourth cycle.
The existing BDOT detection coil in the Reference Magnet is simple, very robust. It so far has provided the simplest, most reliable method to locate IMIN on the transductor signal. The reference magnet makes for a unique magnetic cell in GMPS. Is it possible to find something else that will work as well, or better? BDOT coil in the Reference Magnet BDOT coil spare
This is the electronics for generating the BDOT signals from the BDOT Coil in Reference Magnet The BDOT signal is used to detect time of IMIN because It goes quickly through zero volts at that moment and Allows for precise timing discrimination.
How BDOT Ties Into The Clock System 01/22/2015 Booster Elog Chandra, Salah, Kent BDOT crossing zero with respect to the beam, jitter is ~40µs. From one BDOT detection to the next, the jitter in the 66. 7 m. S period is less than 4µS over course of 1 second. The timing slowly changes with the line frequency shifts. Original system triggers the beam at a ‘fixed’ time delay from the previous BDOT detection. The delay is not really ‘fixed’ because the clock used is a ‘line-locked’ clock that breathes with the line frequency and tracks the slow line frequency shifts. This is an elegant system, but has the problem that the line-locking feedback introduces some noise and causes about 40µS of jitter in the timing of the beam pulse. This is the ‘original’ method of using BDOT and RSTDLY to trigger booster clock events. (BMIN is really BDOT)
12/13/2016 Booster Elog, Chandra The Bdot jitter as observed today for ~15 mins We attempted to reduce the clock jitter by starting the timing at BMAX and cutting the delay in ~ half. This made surprisingly little improvement in the jitter. We later noticed that line frequency changes caused subtle changes in the symmetry of the current waveform that can confuse the timing. On 07/14/2016, used BDOT from the high field portion of GMPS waveform, changed polarity of BDOT signal going to MAC Room.
01/08/2019 Booster Elog The data on the beam jitter after Bill's new FPGA based NIM module introduced. Now the B: RSTDLY is triggered with respect to Bill's at 1026 us. (as compared to that triggered with respect to falling edge zero crossing of Bdot signal at B: RSTDLY=30990 us). To get around the clock jitter, Bill measures the time between two BDOT triggers with a crystal-based clock. Then delays the next beam trigger by this time minus 2 m. S. This has reduced the beam trigger jitter to 6µS. On 01/08/2019 Bill Pellico implemented a new FPGA calculated BDOT.
Evaluating magnet voltage dividers as a potential BDOT signal source… Jeff’s working plan for installing magnet voltage dividers in the tunnel… A BDOT substitute: The voltage across a magnet has the same general shape as the signal from a BDOT coil inside a magnet. It goes through zero (with an offset) when the current is at a minimum.
EE support suggested using a voltage divider, like used in Main Injector, to look at the voltage across a gradient magnet. Board & components compliments of EE Support MI style divider (100: 1) with Isolation Amplifier Box layout and assembly by Jeff Larson Tunnel Installation A down side to this is having active electronics in the tunnel.
A typical Booster style voltage divider assembly in the tunnel. Actually two dividers in one assembly 100: 1 ratio 10 MegΩ (big blue resistors) with 100 KΩ (small brown resistors). Yellow devices are spark gaps. Jeff noted the circuit similarity of the MI style divider with the old ones we still have in the Booster tunnel. Jeff prepared Booster dividers to monitor the voltage across two other magnets, these signals were brought upstairs passively. This is a ‘compensated’ divider, capacitor values were chosen to adjust the magnet voltage signal in time and to improve frequency response. Capacitor assemblies were selected and put together by Dirong Chen. Jeff did additional ‘compensation tuning’ locally for the actual installation.
Initially used bread-boarded differential amplifiers to accept passively supplied magnet voltage signals Later replaced by nice Le. Croy differential amplifier
Wide views of magnet voltage, magnet current, and BDOT
Top Scope images show that the active amplifier signal for 202 D magnet voltage has considerably more noise than the passive 19 -1 and 19 -2 magnet voltage signals. Also note the passive signals aren’t compensated for in the top images, so the passive signals lag BDOT by > 300µS, while the active signal precedes BDOT by ~100 µS In lower image the passive signals ARE compensated at divider, see that the magnet voltage signal now precedes BDOT by about 80 µS
A couple of views of the 19 -1 magnet compensated voltage signal RF noise in the Gallery is considerable View during no RF View during RF pulse It was found the 7 A 22 differential amplifiers offered pretty good filtering to remove RF noise
Plot of Magnet Voltage vs BDOT frequency DRIFTS We noticed that the zero crossings of our voltage signals drifted around by about 10µS vs the BDOT signal. We thought this might be a function of line frequency shifts, but the plot shows no correlation between the offset and line frequency. This plot made by triggering scope on the previous BDOT = 0 crossing. Then finding the next BDOT = 0 crossing. Time difference from the expected 15 Hz period was found a frequency was calculated from that. The Magnet Voltage level was measured at that time and recorded.
Homemade Long Magnetic Pickup Coil signal without filtering Coil was laid between GMPS cables going into penetration to tunnel. The beauty of a Tektronix 7 A 22 filter
Rogowski Coil diagram from Wikipedia Fluke i 430 flex Rogowski Coil borrowed from our Fluke Power Analyzer wrapped around cables going to reference magnet. A Rogowski coil measures IDOT, which is like BDOT and so a good candidate for substitution. Tektronix 7 A 22 differential amplifiers with selectable filtering
CH 1 = BDOT signal CH 3 = Rogowski signal CH 4 = BDOT Detected signal Rogowski Coil signal compared to BDOT signal Wide view - No RF Pulse - without amplifier Close view - No RF Pulse - with amplifier Wide view - w RF Pulse - without amplifier Close view - w RF Pulse – with amplifier Closer view - No RF Pulse - without amplifier Closer view - w RF Pulse – without amplifier
Rogowski Coil signal compared to BDOT signal CH 1 = BDOT signal CH 3 = Rogowski signal CH 4 = BDOT Detected signal The Rogowski Coil zero crossing about 54µS before BDOT zero crossing. Differential Amp input shorted, this is noise on amplifier output and zeroed on scope Differential Amp input shorted, see offset level of amplifier output has drifted by about 2 m. V The Rogowski Coil zero crossing drifts by about +/-10µS wrt the BDOT signal.
Booster is the only Fermilab synchrotron that has operated without using mode damping resistors. Mode damping resistors used in Booster could help power supply stability through less changes in IMIN. This is different from the magnetic field driving instabilities in the beam. The GMPS cell containing the reference magnet would require a special and non-trivial inductor to allow use of damping resistors. This is another reason to remove the reference magnet and change how the timing is determined. Diagram of one fourth of the GMPS magnet system, as it is now. This is between Power Supply 1 positive terminal and Power Supply 4 negative terminal. Note the difference the Reference Magnet has on the cell structure at period 21. Same diagram with mode damping resistors added. The special cell at period 21 would need a compensating inductance added along with the mode damping resistors if the reference magnet stays.
We are still investigating the source of timing offsets among our different sources. We will try to expand the Booster current signal to see if we can determine the IMIN time that way and compare our candidates. A 10µS drift is still small compared to our previous jitter, so we would be tempted to run on the Rogowski coil signal for some time and make sure Booster performs well. Then we can try shorting the reference magnet to see how that goes, and finally installing mode damping resistors. We would mechanically install the 50Ω power resistors. We would run their connecting leads but not connect them until we had a four hour downtime followed by a beam study. If things didn’t work out, we could disconnect in the next four hours. We have ~ 37 of 48 power resistors on hand, would need to buy more. Also there are plans to make aluminum plates on which to place the power resistors and then mount to the wall. PIPII consideration: Spice simulation needs to be done to see how the GMPS transmission line is going to be affected by proposed changes implementing 2 short D magnets for injection and 2 wide bore D magnets for extraction.
- Slides: 23