Simple PCIe FMC carrier SPEC issueshttps://ohwr.org/project/spec/issues2019-02-12T09:18:39Zhttps://ohwr.org/project/spec/issues/99PCB copper planes not equal. Board may warp.2019-02-12T09:18:39ZErik van der BijPCB copper planes not equal. Board may warp.As the copper planes of are not homogenous, the V1 version of the PCB
may warp. This may pose problems with the assembly. Verify if this
problem is serious enough to improve on a V2 version of the PCB.https://ohwr.org/project/spec/issues/97GN4124 EEPROM size and footprint2019-02-12T09:18:38ZProjectsGN4124 EEPROM size and footprintI2C EEPROM for GN4124 configuration must be smaller than 2Kb.
The footprint might be changed as well.
Currently a QFN package is used, but there is enough space for a TSSOP.
2Kbit QFN replacement =\> 24LC024T-I/MC
2Kbit TSSOP replacement =\> 24LC024-I/SThttps://ohwr.org/project/spec/issues/96SANYO 20SP47M capacitor is at end-of-life2019-02-12T09:18:37ZErik van der BijSANYO 20SP47M capacitor is at end-of-lifeThe SANYO 20SP47M capacitor is at end-of-life and should not be used for
new designs.
Anyway, the SH version is longer life (5000h vs 1000h), but also at
end-of-life.
Replace type for next version of schematics and PCB.
Erik will find out replacement part.
### Files
* [SEQP.pdf](/uploads/a1229263dc3702da57b1937a691191b4/SEQP.pdf)https://ohwr.org/project/spec/issues/95OSC3 MULTICOMP MCOT7250005V00000RA wrong2019-02-12T09:18:36ZErik van der BijOSC3 MULTICOMP MCOT7250005V00000RA wrongOSC3 is powered by 3V3. In the BOM it is marked as MCOT7250005V00000RA,
which is a 5V oscillator.
MCOT7250003V300000RA exists (e.g.
http://fr.farnell.com/jsp/search/productdetail.jsp?SKU=1787242), which
is 3V3.
- Verify that need the 3V3 version
- Correct schematics (may need request for new symbol)Grzegorz KasprowiczGrzegorz Kasprowiczhttps://ohwr.org/project/spec/issues/93Board needs cut-out at front for ease of installing FMC2019-02-12T09:18:35ZErik van der BijBoard needs cut-out at front for ease of installing FMCThe board needs a cut-out at front for ease of installing FMC cards.
Without it, the PCIe front-panel should be removed for installing an FMC
card.
A prototype with a relatively large cutout has been made, including a
Gerber file for it. See attached file.
### Files
* [PFC3_Board_Shape.pdf](/uploads/1c28746fc3285d505cb94cee78bb52b4/PFC3_Board_Shape.pdf)https://ohwr.org/project/spec/issues/92SATA connector2019-02-12T09:18:34ZGrzegorz KasprowiczSATA connectorThe SATA connector flange is placed too close to the board edge - it is
still possible to install it, but should be shifted about 0.5mm outside
the PCB.Grzegorz KasprowiczGrzegorz Kasprowiczhttps://ohwr.org/project/spec/issues/90Termination resistors on SPI clock line2019-02-12T09:18:33ZTomasz WlostowskiTermination resistors on SPI clock lineR145 and R146 (both 100 Ohm) should be removed, as IC7 can't properly
drive a 50 ohm load. Also, the length of the trace between IC7.B2 and
IC8.C (flash mem) is approx. 15 mm which (given the max. SPI lock
frequency of ~30 MHz) doesn't require any termination.https://ohwr.org/project/spec/issues/89Resistor values2019-02-12T09:18:32ZTomasz WlostowskiResistor valuesResistors R193 and R241 which limit the current of front panel LEDs have
too big values and LEDs are not bright enough.https://ohwr.org/project/spec/issues/84PCB component spacing2019-02-12T09:18:29ZTomasz WlostowskiPCB component spacingSpacing between SMD small components (0402 passives) need to be
increased to avoid short circuits between adjacent components.https://ohwr.org/project/spec/issues/76Replace DS18B20U+ (1-wire thermo + ID) by MCP9801-M/MS2019-02-12T09:18:23ZProjectsReplace DS18B20U+ (1-wire thermo + ID) by MCP9801-M/MSWe don't need the unique ID feature of the DS18B20U+.
GN4124 EEPROM can be used to store board ID data if needed.
Note that GN4124 boot configuration stops when reading 0xFFFF.
Therefore user data can be stored after 0xFFFF
The ALERT pin of MCP9801-M/MS must be connected to the FPGA.
This is to generate the carrier over-heat interrupt.https://ohwr.org/project/spec/issues/75V1.1 - Change 20Mhz oscillator net label2019-02-12T09:18:22ZProjectsV1.1 - Change 20Mhz oscillator net labelOn page "Clock oscillators & distribution" OSC2 is 20MHz and the label
says "CLK25\_VCXO".https://ohwr.org/project/spec/issues/74V1.1 - SFP control lines are mixed up on the top level schematics2019-02-12T09:18:22ZProjectsV1.1 - SFP control lines are mixed up on the top level schematicsThis leads to name mismatch between FPGA bank page and SFP page.https://ohwr.org/project/spec/issues/73V1.1 - Change "AUX" I2C bus name2019-02-12T09:18:21ZProjectsV1.1 - Change "AUX" I2C bus nameThe "AUX" I2C bus is only connected to Si570x, so it should be renamed.https://ohwr.org/project/spec/issues/72V1.1 - Mercury oscillator has the wrong footprint2019-02-12T09:18:20ZProjectsV1.1 - Mercury oscillator has the wrong footprintThe reference VM53S3 corresponds to 5x3.2mm package,
but the the footprint has been modified for a 7x5mm.
The right reference for a 7x5mm package is VM57S3.
Which is the package we want?https://ohwr.org/project/spec/issues/70Test pins should not be mounted.2019-02-12T09:18:17ZErik van der BijTest pins should not be mounted.In the schematics and BOM of V2, the 8 test pins "Black bead PCB
terminal test pin" are shown to be mounted.
They should be not mounted as they will not be used in production and it
will add to the production cost.
(B2,B3,B5,B7,B17-B19,B22)https://ohwr.org/project/spec/issues/69NUMONYX M25P32-VMF6G obsolete. Replace by M25P32-VMF6P.2019-02-12T09:18:15ZErik van der BijNUMONYX M25P32-VMF6G obsolete. Replace by M25P32-VMF6P.In the schematics and BOM of V2, IC8: NUMONYX M25P32-VMF6G is obsolete.
Replace by M25P32-VMF6P.
Tom suggested: *"also put a footprint for the VQFN (M25P32-VME6G)
version of this chip in parallel with the SOIC one"*. Check with others
before doing this as it would need renumbering and a more serious change
of schematics.https://ohwr.org/project/spec/issues/68V2: JP6 jumper: wrong reference2019-02-12T09:18:15ZErik van der BijV2: JP6 jumper: wrong referenceIn the schematics and BOM of V2, JP6: KONTEK COMATEL 386.0358.1.02.440
jumper is not correct (too long pins, and somewhat difficult to get).
Replace by Farnell/ 1022247 (HARWIN - M20-9990246).https://ohwr.org/project/spec/issues/64V2 - Typo in the schematics2019-02-12T09:18:12ZProjectsV2 - Typo in the schematicsIn the schematics, signal P2L\_FRAME should be called P2L\_DFRAME.https://ohwr.org/project/spec/issues/58V3: Move SFP cage further out2019-02-12T09:18:08ZErik van der BijV3: Move SFP cage further outIt has been reported that a user is unable to install an SFP connector
into the SFP cage because the
PC mechanics (on a commodity PC) pushes it back outwards. See attached
file.
Indeed there seems to be something really on the limit. We can improve
this in a future design by moving the SFP cage just 2 mm more to the
edge of the board.
### Files
* [sfp_too_near.jpg](/uploads/2900765fda0c34b0eda5d7cbcb5ed156/sfp_too_near.jpg)https://ohwr.org/project/spec/issues/46New firmware load can freeze PC. SW problem?2019-02-12T09:18:03ZErik van der BijNew firmware load can freeze PC. SW problem?Several developers have seen a lockup of the PCIe bus when downloading
(lm32) firmware.
See the WR dev mailing list:
http://lists.ohwr.org/sympa/arc/white-rabbit-dev/2012-04/msg00036.html
(you may have to click "I'm not a spammer" and then go again to this
specific link).
It may have been solved as Martin B. wrote:
??We had the same issue. And as far as I can remember it was the old
rawrabbit.ko module. Did you use the new spec.ko (spec-sw repository)?
Then read the dmesg output and see if the correct files where loaded
(check the bytes loaded counter twice).??
- To be followed up as it may hide a certain problem and is not yet
understood.Javier SerranoJavier Serrano