DS-DM
Jump to navigation
Jump to search
DS-DM
DarkSide-20k Global and Crate Data Manager board (GDM and CDM).
Global Data Manager (GDM):
- clock distribution to CDM boards (including GPS/atomic clock source)
- collection of trigger data from CDM boards, processing and distribution of trigger decision to CDM boards
- run control
Crate Data Manager (CDM):
- clock distribution from GDM to CAEN digitizers
- receive trigger data from CAEN digitizers
- send trigger data to GDM
- run control and dead time control
Links
edev links:
- https://edev-group.triumf.ca/fw/exp/darkside/gcdm
- https://edev-group.triumf.ca/hw/vme/dark-side-20k-data-manager-card/rev0
- https://edev-group.triumf.ca/hw/vme/dark-side-20k-data-manager-card/rev0/-/blob/master/Altium/Project%20Outputs%20for%20DS-DM-Rev0/SCH-DS-xDM-Rev0.PDF
- https://edev-group.triumf.ca/hw/vme/dark-side-20k-data-manager-card/rev0/-/blob/master/Testing/Clk3_XO_125_to_fpgaIN_recoveredMGTclk_to_IN2_Si5394-RevA-Project.slabtimeproj
Xilinx links:
- Platform Cable USB II: https://docs.xilinx.com/v/u/en-US/ds593
Enclustra links:
Onboard hardware
- jtag chain: arm_dap_0 0x5BA00477, xczu4_1 0x04721093
- Eclustra Mercury+ XU8 module: ME-XU8-4CG-1E-D11E-R2.1
- Xilinx® Zynq Ultrascale+™ MPSoC XCZU4CG-1FBVB900E
- DDR4 ECC SDRAM (PS) 2 GB
- DDR4 SDRAM (PL) 1GB
- ethernet mac chip: AT24MAC602-SSHM-T
- USB UART for Enclustra serial console, micro-USB, 115200n8
- LEDs:
- LED_FP A/B/C/D 0/1/2/3
- led1 - 3V3_SW_ON, SOM_POWER_GOOD - Enclustra FPGA module 3.3V power is good
- led2 - LTM4624 PGOOD
- led3 - FPGA_DONE - FPGA has booted
- led4 - TP-S-1, PCLK_P
- led5 - TP-S-2, PCLK_N
- LEMO connectors (top to bottom)
- J4 - input (NIM/TTL) (EXT_IN_LV(1), EXT_IN_LV(2))
- J5 - input (NIM/TTL) (EXT_IN_LV(3), EXT_IN_LV(4))
- J6 - external clock (GPS 10MHz and PPS)
- J7 - output (NIM/TTL) (EXT_OUT(1), EXT_OUT(2))
- SMA connectors
- J9, J10 - CLK_CCA from U6 C.C.
- J11, J12 - CLK_TP0
- RJ45 ethernet connector (100mbit: green light on, yellow light flashes when there is traffic)
- SFP connector
- 4 QSFP connectors (GDM)
- 6 VX connectors (CDM)
Buttons, jumpers and switches
Buttons:
- PB1 - HRST - reboot FPGA (power-on reset)
- PB2 - SRST - (SRSTn) - reboot ARM CPU
Switches:
- SW1 - boot mode BM0, BM1 [-->]
- SW2 - LEMO output NIM<->TTL
- SW3 - LEMO input 1 and 2 NIM/TTL
- SW4 - LEMO input 2 and 4 NIM/TTL
- SW5 - ???
- SW6 - serial console select. [PS<--PL] PS is ARM CPU, PL is FPGA.
Front panel
| top | | LED-FP1 | LED_FP(0,1,2,3) | | SFP J??? | | LEMO J4-LEFT, J4-RIGHT | EXT_IN_LV(1), EXT_IN_LV(2) | LEMO J5-LEFT, J5-RIGHT | EXT_IN_LV(3), EXT_IN_LV(4) | LEMO J6-LEFT, J6-RIGHT | | LEMO J7-LEFT, J7-RIGHT | EXT_OUT(1), EXT_OUT(2) | | J-VX-1 | J-VX-2 or QSFP | J-VX-3 or QSFP | J-VX-4 or QSFP | J-VX-5 or QSFP | J-VX-6 | | RJ45 J3 ethernet | | bottom
VX adapter board
LVDS I/O connector
34 pin connector: 0|:::: :::: :::: :::: :|15,16 n/c 0 -> VX2_RX(3) 1 -> VX2_RX(2) 2 -> VX2_RX(0) 3 -> VX2_RX(1) 4 <- CLK 5 <- VX2_TX(0) 6 <- VX2_TX(1) 7 <- VX2_TX(2) 8 -> VX1_RX(3) 9 -> VX1_RX(2) 10 -> VX1_RX(1) 11 -> VX1_RX(0) 12 <- VX1_TX(0) 13 <- VX1_TX(1) 14 <- VX1_TX(2) (set by jumper routed here or to VX CLKIN SYNC) 15 <- n/c (CLK routed to VX CLKIN CLK) 16 - n/c
Board schematics
Clock distribution
125 MHz oscillator - U5 fan out - q0 -> CLK_XO_125 - FPGA PCLK pin - (NOT IN CDM PROJECT) q1 -> U6 C.C. in1 q2 -> CLK3_XO_125 -> FPGA IO pin -> SFP RX reference clock q3 -> CLK2_XO_125 -> disconnected on the board (was -> FPGA SCLK pin) U6 C.C (clock cleaner) - in0 <- CLK_EXT1 (presumably GPS 10 MHz ref clock) in1 <- 125 MHz oscillator via U5 in2 <- (was: CLK_CC_IN <- FPGA MGT_CLK pin <- SFP RX clock, 125 MHz) in2 <- CLK2_XO_125 <- FPGA SCLK pin <- mgt_link_data_to_processing.rx_data_clk (SFP RX data clock, 125 MHz) in3 <- CLK_FB out0 -> CLK_CCA -> U12 (125 MHz) out1 -> CLK_CCB -> VX1..6 (62.5 MHz) out2 -> CLK_CCC -> VX7..12 (62.5 MHz) out3 -> CLK_FB into in3 CLK_CCA -> U12 (125 MHz fan out) -> SMA J9/J10 -> CLK_CC_OUT0 -> FPGA IO pin (-> MGT clock input ???) -> CLK_CC_OUT1 -> FPGA SCLK pin -> SFP TX clock (MGT clock input 125 MHz) -> unused output CLK_TP0 - FPGA N7, N8 - Enclustra C69, C71 - DS-DM SMA J11, J12 (NOT IN CDM PROJECT) We have 3 questions: 1) CLK_XO_125 (125 MHz osc) is not in the CDM project, why? 2) CLK_CC_OUT0 (U6 C.C. 125 MHz) is not in the CDM project, why? -> it is an MGT clock, connected now. 3) 62.5 MHz VX clock does not go into the FPGA, please confirm? -> confirmed > And 1 comment: > > a) CLK_EXT0 going to in1 of U5 does not seem to be useful. only permitted frequency is 125 MHz and we normally do not distribute 125 MHz clocks on LEMO connectors and LEMO cables. (we use LVDS signals over SATA/eSATA cables and connectors). We remove it in the next revision of the board? Proposed modifications: - CLK_CC_IN: from FPGA output pin (J-ENC-C 142/144) to clock cleaner in2 (instead of CLK2_XO_!25 pins) - CLK_CCA -> U12 -> currently unused out3 -> CLK2_XO_125 (FPGA SCLK pins)
Clock domains
GDM
- AXI clock (100 MHz) - AXI registers
- 10 MHz GPS clock - to clock cleaner
- 125 MHz oscillator - to clock cleaner, to QSFP MGT RX reference clock pins
- clock cleaner 125 MHz clock: (main clock domain)
- QSFP MGT TX clock pins (via ...)
- QSFP TX data clock
- QSFP TX data
- trigger logic
- trigger logic output to QSFP TX data
- QSFP RX recovered clock (per link, 12 clocks)
- QSFP RX data to phase matching fifo to main clock domain to trigger logic
CDM
- AXI clock (100 MHz) - AXI registers
- 125 MHz oscillator to clock cleaner, to SFP MGT RX reference clock pin via CLK3_XO_125
- SFP RX data clock ("recovered clock") (250 MHz/8 bit, 125 MHz/16 bit, 62.5 MHz/32 bit data) (main clock domain)
- to clock cleaner to VX 62.5 MHz clock
- SFP RX data
- SFP RX data decoder
- (SFP RX MGT clock 125 MHz not used because of uncontrollable phase)
- SFP TX MGT clock 125 MHz via clock cleaner to CLK_CC_OUT1
- SFP TX data clock
- SFP TX data encoder
- trigger logic
- trigger logic to SFP TX
- trigger logic to VX TX
- VX TX data serializer
- VX TX LVDS clock and data
- inside the VX firmware: VX RX clock and data, data deserializer, phase matching fifo to VX trigger logic
- CLK_XO_125 125 MHz from oscillator
- not used
- CLK_CC_OUT0 125 MHz from clock cleaner
- not used
- VX RX receive clock (per link, 12 clocks)
- VX RX clock and data
- VX RX data deserializer
- phase matching fifo to main clock domain trigger logic
VX
- VX 62.5 MHz clock from CDM
- VX 125 MHz clock
- drives Yair's trigger logic
- trigger logic to LVDS TX
- LVDS TX serializer
- LVDS TX clock and data
- LVDS RX clock (125 MHz from CDM)
- LVDS RX clock and data
- LVDS RX data deserializer
- phase matching fifo to VX 125 MHz clock domain
Board test plan
To test:
- Enclustra FPGA board
- SFP port
- CDM VX ports 2x(CLK, 3 tx, 4 rx)
devmem 0x80010014 32 # read VX1_RX devmem 0x80010018 32 # read VX2_RX devmem 0x8001001C 32 0x21 # write VX1_TX=1, VX2_TX=2
- GDM QSFP ports (lanes 0,1,2. lane 3 n/c)
- J6A, J6B external clock CLK_EXT1, CLK_EXT0 (sw5 is missing how is this supposed to work?). 50 ohm termination. CLK_EXT0 expects 125 MHz ?!?
- BOOT_MODE 0 and 1
Done:
- LED_FP1A..D: tested ok. K.O. 15 sep 2022
- USB UART: tested ok. K.O. 15 sep 2022
- J4A, J4B, J5A, J5B LEMO inputs (NIM/TTL) EXT_IN_LV(1..4). TTL threshold 1.7V, NIM threshold -0.3V. 50 Ohm termination.
- J7A, J7B LEMO outputs EXT_OUT(1), EXT_OUT(2) (NIM/TTL) (tested 23nov2022, K.O.)
- TTL out no 50 ohm termination: 0=0V, 1=5V, rise and fall time ~5 ns
- TTL out with 50 ohm termination: 0=0V, 1=2.5V, rise and fall time <2ns
- NIM out no 50 ohm termination: 0=+50mV, 1=-1.8V, rise and fall time ~3ns
- NIM out with 50 ohm termination: 0=0V, 1=-0.9V, rise and fall time <2ns
- ethernet MAC i2c chip (K.O. 20sep2022: can read: i2cdump 0 0x5b, nov2022 have u-boot driver)
Failure:
- ethernet: does not connect to alliedtelesys switch. connects to my USB-eth adapter at 100 Mbit speed. uboot mii status reports connection speed oscillating between 1000, 100 and 10. K.O. 16-sep-2022
Checklist for newly build boards
- put new board on workbench
- check - vme connector present, vme extraction handles present
- check - standoff are removed from all thru-holes
- plug Enclustra module
- check - SW6 both switches are in the "PS" position
- connect micro-usb cable from linux PC
- connect ethernet from 1gige capable network switch
- connect power from lab power supply - GND, +5V and -12V nominal, +5.7V and -12.5V actual
- power up, +5V current 2.10-2.8A, -12V current 0.05A
- on linux PC, open a new terminal, run: minicom -D /dev/ttyACM0
- in minicom window, observe messages about Xilinx first stage boot loader, etc
- on the ethernet switch, observe network link is 1gige speed (not 10mbit, not 100mbit).
- if everything boots okey, there will be a login prompt, login as root, password root.
- busybox devmem 0x80010000 # read firmware version number, i.e. 0xEDAD0A77
- TBW - test LEMO inputs
- TBW - test LEMO outputs
- TBW - test VX connectors
Serial console
- check that linux computer has correct udev rules to allow access to /dev/ttyACM devices, see https://daq00.triumf.ca/DaqWiki/index.php/SLinstall#Configure_USB_device_permissions and https://daq00.triumf.ca/DaqWiki/index.php/Ubuntu#Configure_USB_device_permissions
- connect micro-USB cable to connector J-UCB, other end connect to linux computer
- observe /dev/ttyACM0 was created
- run "minicom -D /dev/ttyACM0" (default serial settings are ok, otherwise, 115200n8)
- should have gdm-cdm login
- username root, password root
i2c
ZynqMP> i2c bus Bus 0: i2c@ff020000 ZynqMP> i2c dev 0 Setting bus to 0 ZynqMP> i2c probe Valid chip addresses: 33 4E 53 5B 6B 77 ZynqMP> i2c md 0x5b 0x98 0098: fc c2 3d 00 00 12 1a 6e 0a 90 85 04 94 10 08 50 ..=....n.......P
root@gdm-cdm:~# i2cdetect 0
Warning: Can't use SMBus Quick Write command, will skip some addresses
WARNING! This program can confuse your I2C bus, cause data loss and worse!
I will probe file /dev/i2c-0.
I will probe address range 0x03-0x77.
Continue? [Y/n]
0 1 2 3 4 5 6 7 8 9 a b c d e f
00:
10:
20:
30: -- -- -- 33 -- -- -- --
40:
50: -- -- -- 53 -- -- -- -- -- -- -- 5b -- -- -- --
60:
70:
root@gdm-cdm:~#
root@gdm-cdm:~# i2cdump 0 0x5b
No size specified (using byte-data access)
WARNING! This program can confuse your I2C bus, cause data loss and worse!
I will probe file /dev/i2c-0, address 0x5b, mode byte
Continue? [Y/n]
0 1 2 3 4 5 6 7 8 9 a b c d e f 0123456789abcdef
00: XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX XXXXXXXXXXXXXXXX
10: XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX XXXXXXXXXXXXXXXX
20: XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX XXXXXXXXXXXXXXXX
30: XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX XXXXXXXXXXXXXXXX
40: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ................
50: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ................
60: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ................
70: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ................
80: 0a 90 85 04 94 10 08 50 90 5b a0 00 a0 00 00 00 ???????P?[?.?...
90: 00 00 00 00 00 00 00 00 fc c2 3d 00 00 12 1a 6e ........??=..??n
a0: 0a 90 85 04 94 10 08 50 90 5b a0 00 a0 00 00 00 ???????P?[?.?...
b0: 00 00 00 00 00 00 00 00 fc c2 3d 00 00 12 1a 6e ........??=..??n
c0: XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX XXXXXXXXXXXXXXXX
d0: XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX XXXXXXXXXXXXXXXX
e0: XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX XXXXXXXXXXXXXXXX
f0: XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX XX XXXXXXXXXXXXXXXX
root@gdm-cdm:~#
read ethernet mac address from i2c
(this code is copied from uboot command line i2c code)
in uboot sources board/xilinx/common/board.c replace original function with this:
// special code to read ethernet MAC address from the DS-DM-Rev0 board. K.O. Sep-2022
int zynq_board_read_rom_ethaddr(unsigned char *ethaddr)
{
struct udevice *bus;
int ret;
int busnum = 0;
ret = uclass_get_device_by_seq(UCLASS_I2C, busnum, &bus);
if (ret) {
printf("%s: No bus %d\n", __func__, busnum);
return ret;
}
int chip_addr = 0x5B;
struct udevice *dev;
ret = i2c_get_chip(bus, chip_addr, 1, &dev);
if (ret) {
printf("%s: Bus %d no chip 0x%02x\n", __func__, busnum, chip_addr);
return ret;
}
int dev_addr = 0x98;
unsigned char data[8];
ret = dm_i2c_read(dev, dev_addr, data, 8);
if (ret) {
printf("%s: Bus %d chip 0x%02x read error %d\n", __func__, busnum, chip_addr, ret);
return ret;
}
printf("%s: Bus %d chip 0x%02x addr 0x%02x read: 0x %02x %02x %02x %02x %02x %02x %02x %02x\n", __func__, busnum, chip_addr, dev_addr, data[0], data[1], data[2], data[3], data[4], data[5], data[6], data[7]);
// see Atmel-8807-SEEPROM-AT24MAC402-602-Datasheet.pdf
if (data[0] == 0) {
// eiu-48 chip
ethaddr[0] = data[2];
ethaddr[1] = data[3];
ethaddr[2] = data[4];
ethaddr[3] = data[5];
ethaddr[4] = data[6];
ethaddr[5] = data[7];
} else {
// eiu-64 chip
ethaddr[0] = data[0];
ethaddr[1] = data[1];
ethaddr[2] = data[2];
ethaddr[3] = data[5];
ethaddr[4] = data[6];
ethaddr[5] = data[7];
}
printf("%s: ethaddr %02x:%02x:%02x:%02x:%02x:%02x\n", __func__, ethaddr[0], ethaddr[1], ethaddr[2], ethaddr[3], ethaddr[4], ethaddr[5]);
return ret;
}
also this should have worked if i2c_xxx() functions were enabled in uboot:
i2c_set_bus_num(0); i2c_probe(0x5b); i2c_read(0x5b, 0x9a, ethaddr, 6);
read ethernet mac address from i2c (SHOULD WORK)
#ethernet related setup setup_eth=run readmac buildmac #read mac address from eeprom readmac=i2c dev 0; i2c read 50 FA.1 6 $loadaddr #build the ethaddr variable #not very nice, but does the job buildmac=\ e=" "; sep=" " \ for i in 0 1 2 3 4 5 ; do\ setexpr x $loadaddr + $i\ setexpr.b b *$x\ e="$e$sep$b"\ sep=":"\ done &&\ setenv ethaddr $e
read ethernet mac address from i2c (DOES NOT WORK)
this method does not work: inside board/xilinx/common/board.c:int zynq_board_read_rom_ethaddr(unsigned char *ethaddr), uclass_get_device_by_ofnode(UCLASS_I2C_EEPROM, eeprom, &dev) returns ENODEV and read from i2c does not happen. K.O. Sep-2022
Read:
- https://support.xilinx.com/s/question/0D52E00006hpKlsSAE/uboot-and-ethernet-mac-address-from-eeprom?language=en_US (last answer)
- https://github.com/Xilinx/u-boot-xlnx/blob/master/board/xilinx/common/board.c (zynq_board_read_rom_ethaddr())
Note:
- 0x5B is the i2c chip address
- 0x9A is the data offset inside the chip, see datasheet or i2c read dump above.
Edit:
- emacs -nw PetaLinux_GDM_CDM/project-spec/meta-user/recipes-bsp/u-boot/files/platform-top.h
#include <configs/xilinx_zynqmp.h> #include <configs/platform-auto.h> //#define CONFIG_I2C_EEPROM //#define CONFIG_SYS_I2C_EEPROM_ADDR 0x5b //#define CONFIG_SYS_I2C_EEPROM_ADDR_OVERFLOW 0x0 #define CONFIG_ZYNQ_GEM_I2C_MAC_OFFSET 0x9A #error HERE!
- emacs -nw PetaLinux_GDM_CDM/project-spec/meta-user/recipes-bsp/device-tree/files/system-user.dtsi
/include/ "system-conf.dtsi"
/ {
chosen {
xlnx,eeprom = &eeprom;
};
};
&i2c0 {
eeprom: eeprom@5b { /* u88 */
compatible = "atmel,24mac402";
reg = <0x5b>;
};
};
- components/yocto/workspace/sources/u-boot-xlnx/configs/xilinx_zynqmp_virt_defconfig
CONFIG_ZYNQ_GEM_I2C_MAC_OFFSET=0x9A
enable VX clock
devmem 0x80010010 32 0x8; sleep 1; devmem 0x80010010 32 0x0; si5394-i2c-file /media/sd-mmcblk1p1/00_freerun.txt 0 0x6b
arm and linux benchmark
memory benchmark:
daq13$ arm-linux-gnueabi-gcc -o memcpy.armv7 memcpy.cc -march=armv7 -static -O2 scp memcpy.armv7 to ... root@gdm-cdm:~# ./memcpy.armv7 memcpy 1 KiBytes: 1288 MB/sec memcpy 2 KiBytes: 1924 MB/sec memcpy 4 KiBytes: 2554 MB/sec memcpy 8 KiBytes: 3054 MB/sec memcpy 16 KiBytes: 3262 MB/sec memcpy 32 KiBytes: 3250 MB/sec memcpy 64 KiBytes: 3456 MB/sec memcpy 128 KiBytes: 3556 MB/sec memcpy 256 KiBytes: 3780 MB/sec memcpy 512 KiBytes: 3795 MB/sec memcpy 1024 KiBytes: 3789 MB/sec memcpy 2048 KiBytes: 3729 MB/sec memcpy 4096 KiBytes: 3717 MB/sec memcpy 8192 KiBytes: 3687 MB/sec memcpy 16384 KiBytes: 3632 MB/sec memcpy 32768 KiBytes: 3529 MB/sec memcpy 65536 KiBytes: 3318 MB/sec memcpy 131072 KiBytes: 2893 MB/sec root@gdm-cdm:~#
ethernet receive:
daq13:bin$ ./ttcp -t -s -n 100000 10.0.0.24 ttcp-t: buflen=8192, nbuf=100000, align=16384/0, port=5001 tcp -> 10.0.0.24 ttcp-t: socket ttcp-t: connect ttcp-t: 819200000 bytes in 7.25 real seconds = 110358.39 KB/sec +++ ttcp-t: 100000 I/O calls, msec/call = 0.07, calls/sec = 13794.80 ttcp-t: 0.0user 0.2sys 0:07real 3% 0i+0d 760maxrss 0+2pf 1461+31csw daq13:bin$ root@gdm-cdm:~# ./ttcp.armv7 -r -s ttcp-r: buflen=8192, nbuf=2048, align=16384/0, port=5001 tcp ttcp-r: socket ttcp-r: accept from 10.0.0.25 ttcp-r: 819200000 bytes in 7.27 real seconds = 110098.22 KB/sec +++ ttcp-r: 212040 I/O calls, msec/call = 0.04, calls/sec = 29181.53 ttcp-r: 0.1user 5.7sys 0:07real 81% 0i+0d 584maxrss 0+2pf 125601+2699csw root@gdm-cdm:~#
ethernet transmit:
root@gdm-cdm:~# ./ttcp.armv7 -t -s -n 100000 10.0.0.25 ttcp-t: buflen=8192, nbuf=100000, align=16384/0, port=5001 tcp -> 10.0.0.25 ttcp-t: socket ttcp-t: connect ttcp-t: 819200000 bytes in 6.95 real seconds = 115078.69 KB/sec +++ ttcp-t: 100000 I/O calls, msec/call = 0.07, calls/sec = 14384.84 ttcp-t: 0.0user 0.7sys 0:06real 11% 0i+0d 584maxrss 0+2pf 1162+1017csw root@gdm-cdm:~# daq13:bin$ ./ttcp -r -s ttcp-r: buflen=8192, nbuf=2048, align=16384/0, port=5001 tcp ttcp-r: socket ttcp-r: accept from 10.0.0.24 ttcp-r: 819200000 bytes in 6.97 real seconds = 114841.84 KB/sec +++ ttcp-r: 161335 I/O calls, msec/call = 0.04, calls/sec = 23160.01 ttcp-r: 0.0user 1.9sys 0:06real 28% 0i+0d 760maxrss 0+2pf 80646+51csw daq13:bin$
Install Xilinx tools
- install Vivado 2020.2
login at https://www.xilinx.com/myprofile.html go to "Downloads" go to archive, find 2020.2 download Xilinx_Unified_2020.2_1118_1232_Lin64.bin sh ./Xilinx_Unified_2020.2_1118_1232_Lin64.bin banner window should open with spinner "downloading installation data" "a newer version is available" -> say "continue" next "select install type" window: provide email and password, select "download image" select directory /home/olchansk/Xilinx/Downloads/2020.2\ select "linux" and "full image" next download summary: space required 38.52 Gbytes download installation progress downloading spinner, 16 M/s 47 minutes... "download image has been created successfully". Ok. check contents of /home/olchansk/Xilinx/Downloads/2020.2 ls -l /home/olchansk/Xilinx/Downloads/2020.2 total 67 drwxr-xr-x 2 olchansk users 9 Sep 1 16:22 bin drwxr-xr-x 3 olchansk users 15 Sep 1 16:23 data drwxr-xr-x 4 olchansk users 4 Sep 1 16:22 lib drwxr-xr-x 2 olchansk users 644 Sep 1 16:22 payload drwxr-xr-x 2 olchansk users 7 Sep 1 16:22 scripts drwxr-xr-x 4 olchansk users 4 Sep 1 16:22 tps -rwxr-xr-x 1 olchansk users 3256 Nov 18 2020 xsetup daq13:2020.2$ ./xsetup spinned loading installation data xilinx design tools 2022.1 now available -> say continue "welcome" -> next "select product" -> vivado -> next -> vivado hl system edition -> next select devices: only zynq ultrascale+ mpsoc -> next select destination: /opt/Xilinx (as root, mkdir /opt/Xilinx, chmod olchansk.users /opt/Xilinx) install ... complete move /home/olchansk/Xilinx/Downloads/2020.2 to /daq/daqstore/olchansk/Xilinx/Downloads/
- install petalinux 2020.2
./xsetup "a newer version is available" -> say "continue" next "select product to install" -> select Petalinux (Linux only) -> next "select destination directory" -> select "/opt/Xilinx" (disk space required 2.64 GB) -> next "summary" -> install ... error about missing /tmp/tmp-something files "installation completed successfully" (hard to dismiss, "ok" button is partially cut-off) done? I think it failed, /opt/Xilinx/PetaLinux/2020.2/bin is empty except for petalinux-v2020.2-final-installer.run try to run it by hand, same error about /tmp/tmp-something files. strange... notice it complains about "truncate", which truncate finds ~/bin/truncate, get rid of it, try again now complains about missing texinfo and zlib1g:i386 apt install texinfo -> ok apt install zlib1g:i386 -> installs bunch of gcc stuff -> ok try again reports "already installed" -> delete /opt/Xilinx/.xinstall/PetaLinux_2020.2/, delete entries in ~/.Xilinx/registry/installedSW.xml try again success
- install vivado 2022.1 and petalinux 2022.1 - everything is pretty much the same
Petalinux
- cd PetaLinux_GDM_CDM
- petalinux-config
- enable i2c MAC address and DHCP
git clone https://bitbucket.org:/team-ds-dm/ds-dm-u-boot-xlnx.git cd ds-dm-u-boot-xlnx git checkout ds-dm-u-boot-xlnx linux-components -> uboot -> ext-local-src external u-boot local source -> ds-dm-u-boot-xlnx (path to the customized uboot git repository)
- enable DHCP
Subsystem AUTO Hardware Settings -> Ethernet Settings randomize MAC address -> NO ethernet mac address -> leave empty obtain ip address automatically -> YES
- set hostname and product names
Firmware Version Configuration -> Host name -> "ds-dm" Product name -> "Petalinux_GDM_CDM"
- configure linux kernel
petalinux-config -c kernel
- enable NFS-Root
petalinux-config Image Packaging Configuration > Root File System Type -> set to NFS Location of NFS root directory set to "/nfsroot" petalinux-config -c kernel Networking support > IP: kernel level configuration enable DHCP, BOOTP, RARP File systems > Network file systems > Root file systems on NFS
- manually fix linux kernel command line:
grep nfsroot PetaLinux_GDM_CDM/project-spec/configs/config edit CONFIG_SUBSYSTEM_BOOTARGS_GENERATED to read earlycon console=ttyPS0,115200 clk_ignore_unused panic=60 root=/dev/nfs nfsroot=/nfsroot/%s ip=dhcp rw
- check configuration in
- PetaLinux_GDM_CDM/project-spec/configs/config
- PetaLinux_GDM_CDM/project-spec/configs/rootfs_config
- PetaLinux_GDM_CDM/components/plnx_workspace/device-tree/device-tree/system-conf.dtsi
JTAG server
localhost:3121
Firmware registers
Register map
0 | ALL | ALL | RO | USR_ACCESSE2 see https://docs.xilinx.com/r/en-US/ug974-vivado-ultrascale-libraries/USR_ACCESSE2 1 | ALL | ALL | RW | read write scratch register 2 | ALL | CDM | ?? | MGT not used 3 | ALL | CDM | RO | MGT debug_data 4 | ALL | CDM | RW | clk_config_vec 5 | ALL | CDM | ?? | not used 6 | ALL | CDM | RO | CDM_link_data_processing:o_error_time 7 | ALL | CDM | RO | CDM_link_data_processing:o_error_count 255 | ALL | ALL | RW | control register
control register 255
31..28 | ALL | function mux, 4 bits: 0 - outputs driven by control register, 1 - not used, 2 - dsvslice production, 3, 4, 5 - LVDS loopback, see source code 27 | ALL | lemo_in(4) invert 26 | ALL | lemo_in(3) invert 25 | ALL | lemo_in(2) invert 24 | ALL | lemo_in(1) invert 23 | ALL | lemo_in(4) enable 22 | ALL | lemo_in(3) enable 21 | ALL | lemo_in(2) enable 20 | ALL | lemo_in(1) enable 19 18 17 16 | ALL | trigger_enable 15 14 13 | ALL | drive ext_out(2) LEMO output 12 | ALL | drive ext_out(1) LEMO output 11 | ALL | fp_led_out(3) front panel LEDs 10 | ALL | fp_led_out(2) 9 | ALL | fp_led_out(1) 8 | ALL | fp_led_out(0) 7 | ALL | not used 6 | ALL | vx2_tx_out(2) 5 | ALL | vx2_tx_out(1) 4 | ALL | vx2_tx_out(0) 3 | ALL | not used 2 | ALL | vx1_tx_out(2) 1 | ALL | vx1_tx_out(1) 0 | ALL | vx1_tx_out(0)
Power up sequence
- from README file:
busybox devmem 0x80010004 32 0x1; # link startup mode reset until phase-aligned busybox devmem 0x80010000 32 0x5; # reset link and link status registers sleep 1; # just holds reset high busybox devmem 0x80010000 32 0x0; # lower reset and let links lock sleep 2; # give links time to lock busybox devmem 0x80010008 32; # read status 0x00000003 means locked and phase aligned busybox devmem 0x8001000c 32 # read data, 0x1B1C 3CBC lower 16 bits are K-Codes K.28.1 (3C) and K.28.5 (BC) when locked and phase aligned
Build firmware
Build from git clone
- git clone git@edev-group.triumf.ca:fw/exp/darkside/gcdm.git
- #Makefile change VIVADO_SETTINGS_SCRIPT := /opt/Xilinx/Vivado/2022.1/settings64.sh
- #. /opt/Xilinx/Vivado/2022.1/settings64.sh
- . /opt/Xilinx/Vivado/2020.2/settings64.sh
- make clean
- make all_from_scratch
- . /opt/Xilinx/PetaLinux/2020.2/tool/settings.sh
- make petalinux_create
- make petalinux_rebuild_new_hw_des
- bomb out: The TMPDIR: /home/olchansk/git/ds-dm-gcdm/PetaLinux_GDM_CDM/build/tmp can't be located on nfs.
- mkdir /tmp/build_tmp
- rm -rf /home/olchansk/git/ds-dm-gcdm/PetaLinux_GDM_CDM/build/tmp/
- ln -s /tmp/build_tmp /home/olchansk/git/ds-dm-gcdm/PetaLinux_GDM_CDM/build/tmp
- try again
- grinds, loads a whole bunch of packages...
- finishes with desire to copy things to /tftpboot
- make sdcard_cp_to wants to copy files from PetaLinux_GDM_CDM/images/linux/ to SD card
Rebuild
. /opt/Xilinx/Vivado/2020.2/settings64.sh . /opt/Xilinx/PetaLinux/2020.2/tool/settings.sh /usr/bin/time make vivado_rebuild_fw_hw_des # to see errors: more Vivado_GDM_XU8/GDM_XU8.runs/synth_1/runme.log # if successful, updates GDM_XU8_top.bit ls -ltr Vivado_GDM_XU8/GDM_XU8.runs/impl_1/GDM_XU8_top.bit #-rw-r--r-- 1 olchansk users 7797807 Sep 15 13:30 Vivado_GDM_XU8/GDM_XU8.runs/impl_1/GDM_XU8_top.bit # update sdcard files /usr/bin/time make petalinux_repackage # updates BOOT.BIN ls -l PetaLinux_GDM_CDM/images/linux/BOOT.BIN #-rw-r--r-- 1 olchansk users 9228720 Sep 15 13:34 PetaLinux_GDM_CDM/images/linux/BOOT.BIN
Rebuild branch develop_ko
make petalinux_build make petalinux_repackage
in a different window, become root, go to the gcdm tree, insert the sd card:
make copy # will mount the sd card, copy xilinx boot files, unmount, eject
move sd card to the DS-DM board, reset DS-DM board by SRST button
Build in a fresh account
ssh dsdmdev@daq13 mkdir git cd git git clone /home/olchansk/git/ds-dm-gcdm cd ds-dm-gcdm . /opt/Xilinx/Vivado/2020.2/settings64.sh . /opt/Xilinx/PetaLinux/2020.2/tool/settings.sh /usr/bin/time make all_from_scratch BOARD=CDM (sdcard_cp_to bombs out! FIXED!) #/usr/bin/time make petalinux_build BOARD=CDM #/usr/bin/time make petalinux_repackage BOARD=CDM
clean and build GDM:
rm -rf Vivado_GDM_XU8 /usr/bin/time make vivado_create_and_build BOARD=GDM /usr/bin/time make petalinux_repackage BOARD=GDM
rebuild GDM:
/usr/bin/time make vivado_rebuild_fw_hw_des BOARD=GDM /usr/bin/time make petalinux_repackage BOARD=GDM
copy to running GDM or CDM:
scp PetaLinux_GDM_CDM/images/linux/BOOT.BIN dsdaq@dsvslice: ssh dsdaq@dsvslice scp BOOT.BIN root@gdm0:/media/sd-mmcblk1p1/BOOT.BIN ssh root@gdm0 -> shutdown -r now
copy to SD card:
open a root shell format 16 GB Sd card per above cd /home/dsdmdev/git/ds-dm-gcdm make copy (cannot confirm it worked, DS-DM CDM sn 6 serial console is broken)
prepare bootable sd card
format the sd card
this only needs to be done once
- become root
- cd ~olchansk/git/ds-dm-gcdm
- use "lsblk" to identify the SD card (should show as 8/16/32 GB block device)/ /dev/sdd in this case
- make sdcard_format SDCARD_DEVICE=/dev/sdd
- disconnect sd card, reconnect the sd card (to detect new partition tables, etc)
copy CDM boot files
cd /home/dsdmdev/git/ds-dm-gcdm make copy
copy boot files to the sd card
- as root: identify partition labels, run "blkid", should say "BOOT", "rootfs" and "data"
- mount
mkdir /media/olchansk/BOOT mkdir /media/olchansk/rootfs mkdir /media/olchansk/data mount -L BOOT /media/olchansk/BOOT mount -L rootfs /media/olchansk/rootfs mount -L data /media/olchansk/data cp PetaLinux_GDM_CDM/images/linux/BOOT.BIN /media/olchansk/BOOT/ cp PetaLinux_GDM_CDM/images/linux/boot.scr /media/olchansk/BOOT/ cp PetaLinux_GDM_CDM/images/linux/image.ub /media/olchansk/BOOT/ umount /media/olchansk/BOOT umount /media/olchansk/rootfs umount /media/olchansk/data eject /dev/sdd
boot messages
Xilinx Zynq MP First Stage Boot Loader Release 2020.2 Sep 24 2022 - 13:29:15 NOTICE: ATF running on XCZU4CG/silicon v4/RTL5.1 at 0xfffea000 NOTICE: BL31: v2.2(release):xlnx_rebase_v2.2_2020.3 NOTICE: BL31: Built : 18:02:46, Sep 28 2022 U-Boot 2020.01 (Sep 28 2022 - 18:03:39 +0000) Model: DarkSide 20k DM Board: Xilinx ZynqMP DRAM: 2 GiB usb dr_mode not found PMUFW: v1.1 EL Level: EL2 Chip ID: zu4 NAND: 0 MiB MMC: mmc@ff160000: 0, mmc@ff170000: 1 In: serial@ff000000 Out: serial@ff000000 Err: serial@ff000000 Bootmode: SD_MODE1 Reset reason: SOFT Net: ZYNQ GEM: ff0b0000, mdio bus ff0b0000, phyaddr -1, interface rgmii-id Warning: ethernet@ff0b0000 (eth0) using random MAC address - d6:62:5f:13:00:44 eth0: ethernet@ff0b0000 ZYNQ GEM: ff0e0000, mdio bus ff0e0000, phyaddr -1, interface rgmii-id Could not get PHY for eth1: addr -1 Hit any key to stop autoboot: 0 ZynqMP> CTRL-A Z for help | 115200 8N1 | NOR | Minicom 2.7.1 | VT102 | Online 122:2 | ttyACM0
Software
- ssh cdm0 # or gdm0
- git clone https://bitbucket.org/team-ds-dm/ds-dm-software
- cd ds-dm-software
- make
dsvslice integration
- GDM is gdm0
- CDM is cdm0
- connect 1st VX port of cdm0 to VX1, 2nd port of cdm0 to VX2
- power up
- start the CDM frontend:
#ssh dsdaq@dsvslice #scp fecdm.exe root@cdm0: #ssh root@cdm0 #./fecdm.exe -h 192.168.0.251:1176 ssh dsdaq@dsvslice ssh root@cdm0 /home/dsdaq/online/ds-dm-software/fecdm.exe -h 192.168.0.251:1176
- CDM frontend should enable the VX clock, disable the trigger
- start a run
- CDM frontend will enable the trigger
- stop a run
- CDM frontend will disable the trigger
DS-20K DAQ
Overview
DS-DM, GDM and CDM are key parts of the DS-20K DAQ system:
- common clock distribution from external clock (atomic clock, GPS) to GDM to per-quadrant CDMs to VX digitizers
- common trigger distribution from GDM internal algorithm or external input to all VX digitizers
- run control: GDM, CDM, VX all start recording data at the same time (clock and timestamp reset)
- collection of trigger data from VX digitizers to per-quadrant CDMs to GDM
Deliverables
- hardware and firmware for GDM to CDM clock distribution
- hardware and firmware for CDM to VX clock distribution
- hardware and firmware for GDM external clock input (atomic clock or GPS)
- hardware and firmware for CDM and VX serial communications (VX LVDS I/O connector)
- firmware for run control (timestamp reset and sync): GDM to CDM to VX
- firmware for common trigger distribution: GDM to CDM to VX
- firmware for trigger data flow: VX to CDM to GDM
- firmware for busy control: VX to CDM to GDM back to CDM to VX
- firmware for flow control: FEP to GDM MIDAS frontend to GDM to CDM to VX
- GDM MIDAS frontend: clock selector and monitoring, trigger and run control, busy and flow control, GDM housekeeping
- CDM MIDAS frontend: clock monitoring, CDM housekeeping
specific performance:
- GDM external clock: 10 MHz GPS clock
- GDM to CDM fiber link:
- clock XXX MHz
- link data rate: XXX Gbit/sec
- CDM recovered clock: XXX MHz
- CDM recovered clock jitter: XXX ns
- phase alignment between CDMs: XXX ns
- phase alignment between CDMs persists across reboots, power cycles, firmware updates
- phase alignment between CDMs should be easy to measure
- phase alignment between CDMs should be easy to recalibrate if hardware parts are replaced (DS-DM boards, fiber transceivers, fiber cables, etc)
- data packet bandwidth: XXX Mbytes/sec
- data packet latency: XXX clocks
- data packet skew between CDMs: XXX clocks
- CDM to VX clock:
- clock: XXX MHz
- jitter, all CDM clock outputs: XXX MHz
- phase alignment between all CDM clock outputs: XXX ns
- CDM to VX trigger:
- TBD (use the VX "sync" input or VX LVDS I/O line or VX serial link packet)
- CDM to VX serial link:
- clock: XXX MHz (TBD: VX external clock, or LVDS I/O line or link recovered clock)
- bit rate: XXX bits/sec
- latency: XXX link clocks
- maximum skew between VXes: XXX ns
- VX to CDM serial link:
- clock: XXX MHz (TBD: VX external clock, or LVDS I/O line or link recovered clock)
- bit rate: XXX bits/sec
- latency: XXX link clocks
- maximum skew between VXes: XXX ns
- timestamp reset:
- maximum skew between VXes: XXX ns
- busy round trip time: XXX ns (VX to CDM to GDM back to CDM to VX)
- flow control latency: XXX ns (FEP software to GDM MIDAS frontend to GDM to CDM to VX)
Technical risk items
this refers to unexpected behaviour and performance of system components, causes big difficulty in implementing the system, prevents delivery of deliverables, and prevents or negatively affects operation of the DS-20K DAQ or of the whole experiment.
(14-sep-2022, list is not sorted by any criterial: severity, probability, ease of investigation)
(stability of course is long term stability, across hours, days, weeks, months, years)
- stability of Enclustra FPGA modules (crashes/year, failures to boot/year, flash corruption/year)
- stability of GDM external clock PLL (lock loss/year)
- stability of CDM recovered clock (lock loss/year, unexpected phase drifts, etc)
- unexpected failures or bit error rates in GDM-CDM fiber links
- stability of CDM VX clock outputs (stability of clock cleaner chip)
- stability of VX internal clock distribution (VX PLL lock loss events)
- stability of VX CAEN base firmware (different versions of CAEN base firmware have different clock distribution behaviour)
- strange things in CAEN base firmware (unexpected clocking of LVDS I/O, unexpected phase shifts between clocks, etc)
- DS-DM and VX hardware problems (incompatible LVSD I/O, incompatible clock signals, etc)
Milestones
(14-sep-2022: at this stage of the project, priority must be given to identifying and retiring (so called) technical risk factors. it is not good to build the complete system only to discover that (for example) some Enclustra FPGA modules require 5 attempts to boot and erase their flash memory contents once a month. Both example are real-life actual problems that caused big difficulties in GRIFFIN/TIGRESS and ALPHA-g experiments).
Development and testing milestones in time reversed order:
- full DAQ data challenge: all VXes, CDM, GDM, network, FEP, TSP, MIDAS operate as designed
- one quadrant data challenge: 1 VME crate of VX, CDM, GDM, network, FEP, TSP, MIDAS operate as designed
- vertical slice data challenge: 1 VME crate, 2 VX, 2 CDM (1 VX per CDM), GDM, etc operate as designed
- GDM-CDM link finalized (data rate frozen, data packet format frozen, data content permitted to change)
- CDM-VX serial link finalized (data rate frozen, data packet format frozen, data content permitted to change)
- run control (timestamp reset) and trigger distribution design agreed upon, frozen (list of possible triggers permitted to change)
- VX to CDM to GDM data flow design agreed upon, frozen (data contents permitted to change)
- major technical risk items retired (all hardware and firmware is working as expecred without mysteries and surprises, all problems are identified, investigated, resolved, solutions tested)
- stable operation of CDM-VX serial links in vertical slice system
- stable operation of GDM to CDM clock in vertical slice system
- stable operation of CDM to VX clock in vertical slice system
- vertical slice system assembled (1 VME crate, 2 VX, 2 CDM, 1 GDM, network, FEP, TSP, MIDAS)
ZZZ
ZZZ