I see no coupling capacitors on PCIE TX lanes.
Is it Safe to Remove Coupling Capacitors?
Or
Is it possible to make capacitors into the chip?
Add more photos：
\$\begingroup\$
\$\endgroup\$
11
-
\$\begingroup\$ Is it possible they expect them to be on the motherboard? \$\endgroup\$– user4574Oct 17, 2022 at 4:36
-
\$\begingroup\$ Why do you immediately assume there are no capacitors and they can be removed or left off when you don't see them on one board? What if the capacitors are on the other side of PCB? Or integrated to the chip? The specs require AC coupling. \$\endgroup\$– JustmeOct 17, 2022 at 6:22
-
\$\begingroup\$ I can't find any coupling capacitors after careful inspection (will add more photos to the question). And, according to the PCIE Specification, external coupling capacitors are assumed because the values required are too large to feasibly construct on-chip. \$\endgroup\$– Jue WANGOct 17, 2022 at 6:56
-
\$\begingroup\$ Too large to construct on-chip, maybe, but who says they have to be on-chip? What if they are simply in the same plastic package with the chip? You need the datasheet of the SanDisk SSD controller to know. \$\endgroup\$– JustmeOct 17, 2022 at 7:17
-
2\$\begingroup\$ Occham’s Razor. Correlation (this SSD doesn’t work with YOUR design, but works in everything else) does not equal causation (that WD have made a mistake). You’ve been given some options to consider as to where the caps might be; they might even be embedded in the PCB (‘embedded elements’)? WD are not going to avoid use of the coupling capacitors if there was any quantifiable probability of warranty liability blow-back, unless they’d really done their homework that in certain circumstances - like PCIe over the shorter runs of nVME instead of big PCIe connectors - they weren’t needed. \$\endgroup\$– TechydudeOct 17, 2022 at 9:56
|
Show 6 more comments
3 Answers
\$\begingroup\$
\$\endgroup\$
2
If you use a multimeter in diode mode between one of the differential IOs and ground, you should be able to know if there are capacitors in the chip. Most likely you will measure the ESD diode drop which means no capacitor.
They're 75-200 nF which is not possible to do on a chip, so they would have to be discrete MLCCs. The chip looks like a multi chip module (CPU+RAM) but it would be quite surprising to find discrete caps in the package.
The point of these caps is to allow the transmitter and receiver to use different DC bias on the lines (ie, different common mode voltages). So maybe in some circumstances they could be omitted if the designer is absolutely sure the DC bias will be the same on both ends, but... that doesn't seem like sound design.
-
\$\begingroup\$ This seems like a logical way to test this. One thing that comes to mind is the transformer like scheme seen in some MOSFET gate drivers, maybe they found another way to AC couple on chip \$\endgroup\$ Oct 17, 2022 at 10:49
-
\$\begingroup\$ GND -> PETp0/PETn0, The multmeter reads “OL” in diode mode. \$\endgroup\$– Jue WANGOct 20, 2022 at 5:35
\$\begingroup\$
\$\endgroup\$
5
My guess would be that WD engineers somehow concluded that capacitors are not strictly necessary, so they removed them for this SSD as it is a low cost model. I think removing them is purely to save cost. There may be small capacitors on the chip but there is no easy way to know.
As to why this may work well enough in a laptop or computer but not in your design, my guess would be higher capacitance. Both laptops and computer motherboards contain 4-10 layers, and the connector may be far away from the processor directly, meaning that the trace capacitance is not insignificant.
Here are two photos, one where the connector is close and the other where it is far. It may be that this SSD will work in one laptop but not in the other.
source
-
1\$\begingroup\$ You are just guessing. If a specification says the data lines must be AC coupled, do you think such a large and reputable manufacturer could get away with manufacturing products that are not within specifications and could potetially just not work properly or even cause damage? Unlikely. A few capacitors cost almost nothing compared to many other components so unlikely savings compared to complaints or even inability to pass required product compliance tests. \$\endgroup\$– JustmeOct 17, 2022 at 9:19
-
\$\begingroup\$ I will add a capacitors on the RX lane of the motherboard. Hopefully this will test whether the guess is correct or not. \$\endgroup\$– Jue WANGOct 17, 2022 at 9:29
-
-
\$\begingroup\$ @Justme I think it was quite clear that this is a guess. No one except those engineers know, and I very much doubt that they would break NDA to answer. Even if the spec says they must use capacitors, they may have concluded that they are unnecessary for whatever reason. Sound or not. \$\endgroup\$ Oct 17, 2022 at 10:45
-
\$\begingroup\$ @AnasMalas Not to mention that if it did not have AC coupling, it likely would have been already noticed by someone to be a non-compliant device, news spreading, causing product recalls, lost revenue and paid damages. The SSD just likely performs within standards and problem is elsewhere than SSD manufacturer omitting AC coupling. \$\endgroup\$– JustmeOct 17, 2022 at 10:54
\$\begingroup\$
\$\endgroup\$
1
After adding a pair of 100nf capacitors on the PCIE RX lane of the motherboard, everything is A OK now. I really think Anas Malas's guess is right. @AnasMalas thank you very much. Don't be superstitious about big companies.
-
\$\begingroup\$ Glad I could help, regardless of whether I was correct or not. Does the new motherboard work fine with the SSDs that do contain the capacitors? \$\endgroup\$ Nov 30, 2022 at 13:21