Date
1 - 20 of 28
QCX-mini C38
|
Hans Summers
Hi all I thought you may be interested to see this photo attached. I dug into the records of the assembly facility which soldered the PCBA in the QCX-mini, and managed to get the actual photo of the reel that was installed in the machine for the QCX-mini assembly. It's very nice that such photos are captured and retained. Note the manufacturer Samsung (should be OK) and the part number CL21A106KPFNNNE. You can find the datasheet for this capacitor here: https://datasheet.lcsc.com/szlcsc/1811151151_Samsung-Electro-Mechanics-CL21A106KOQNNNE_C1713.pdf At page 4 you can see an explanation of the part code from which we can see:
What should certainly be noted is the 10V rating of this capacitor. An X5R capacitor exhibits capacitance dependent on applied voltage, capacitance DECREASES as the applied voltage increases. This capacitor is already beyond its rated 10V and therefore the capacitance would be expected to be significantly under 10uF, and therefore inadequate to protect the '1117 voltage regulator. The manufacturing error was therefore at the assembly facility who did not follow my BOM specification (25V rated capacitor) and used this lower voltage capacitor reel. I also believe this further reinforces with a sound theoretical reasoning, the recommended solution of fitting a 10uF electrolytic or tantalum capacitor at the voltage regulator input. It remains the case that there is no evidence to suspect that the AMS1117-5.0 voltage regulator is at fault or is in need of replacement, and replacement is NOT recommended since it will risk damaging the PCB unless you are quite experienced in such matters. 73 Hans G0UPL http://qrp-labs.com |
|
|
|
Mike Besemer - WM4B <mwbesemer@...>
Hans,
You are DA MAN!
Now… GO TO BED!
Hope the COVID situation is under control in the lab and the QTH.
73,
Mike WM4B
From: QRPLabs@groups.io [mailto:QRPLabs@groups.io] On Behalf Of Hans Summers
Hi all
I thought you may be interested to see this photo attached. I dug into the records of the assembly facility which soldered the PCBA in the QCX-mini, and managed to get the actual photo of the reel that was installed in the machine for the QCX-mini assembly. It's very nice that such photos are captured and retained.
Note the manufacturer Samsung (should be OK) and the part number CL21A106KPFNNNE. You can find the datasheet for this capacitor here: https://datasheet.lcsc.com/szlcsc/1811151151_Samsung-Electro-Mechanics-CL21A106KOQNNNE_C1713.pdf
At page 4 you can see an explanation of the part code from which we can see:
What should certainly be noted is the 10V rating of this capacitor. An X5R capacitor exhibits capacitance dependent on applied voltage, capacitance DECREASES as the applied voltage increases. This capacitor is already beyond its rated 10V and therefore the capacitance would be expected to be significantly under 10uF, and therefore inadequate to protect the '1117 voltage regulator.
The manufacturing error was therefore at the assembly facility who did not follow my BOM specification (25V rated capacitor) and used this lower voltage capacitor reel.
I also believe this further reinforces with a sound theoretical reasoning, the recommended solution of fitting a 10uF electrolytic or tantalum capacitor at the voltage regulator input. It remains the case that there is no evidence to suspect that the AMS1117-5.0 voltage regulator is at fault or is in need of replacement, and replacement is NOT recommended since it will risk damaging the PCB unless you are quite experienced in such matters.
73 Hans G0UPL
|
|
|
|
Thanks Hans for posting that. I had my suspicions that it was an assembly facility using the “cheapest available part” error. Glad you uncovered that.
toggle quoted message
Show quoted text
I use a lot of 10 uF 25v 0805 ceramic caps in the 7805 regulator section of my TSW kits and buy them from either Digi-Key or Mouser so I’ll just change out both suspect caps on my Mini when it gets here in a couple days and not look back. This info simplifies the repair immensely for those of us who can do our own rework and have the right parts on hand. Nice to find out the regulator isn’t at fault. Thanks, Jim - W0EB On Dec 20, 2020, at 5:33 PM, Hans Summers <hans.summers@...> wrote:
|
|
|
|
Hans, What a relief! This issue must have been bugging you heavily to try and find an explanation for why the component would fail. So, once again, not a matter of poor components at ALL! Just a quality control issue with the SMD line. This is a very easy error, but should never happen. I have set up manufacturing lines where surface mount PCBAs were assembled and purchased quite a few from manufacturing facilities. Makes you wonder if it was a simple error or an intentional swap because the correct reel was not in stock. In any case, they should have systems in place to prevent it. All said and done, it is very nice that you dug into this as deeply as needed to make the root cause surface. You amaze me! Thank you for your commitment to excellence! 73 Michael KN6IZE |
|
|
|
Ryan Flowers
Nice job on the root cause analysis, Hans. Very nice. I hope your PCB manufacturer got an ear full over that one...
-- Ryan Flowers - W7RLF MiscDotGeek - QRP and More |
|
|
|
Shane Justice <justshane@...>
Hans, Great forensic work! Seems the board house owes you a stack of boards and some compensation for the recovery effort necessary to make up for 1000+ boards that now need additional rework. 73, Shane KE7TR
|
|
|
|
Evan Hand
Hans,
toggle quoted message
Show quoted text
On Sun, Dec 20, 2020 at 05:33 PM, Hans Summers wrote: P = rated voltage 10VI am concerned that leaving C38 in the radio and adding a 10uF 25volt device will limit the maximum supply voltage to 10volts. Is this the case? To better understand the risk can you tell us the expected failure mode when the original C38 is stressed with 16volts dc, plus some ripple? Will the parallel 25volt rated device overcome that failure mode, or is there potential for another failure in the future? I have the part to add the 10uF 25volt capacitor to add to the mini. Should I remove C38 so that it will not fail in the future and cause the problem at that time? Is the best solution to remove C38 and replace it with a 20uF 25volt device? What is the correct specification for C38? I really appreciate your openness and diligent work in duplicating and then finding a solution to the problem. I am trying to make sure that this is truly a permanent solution and not just a patch that will fail in the future. 73 Evan AC9TU |
|
|
|
Ryan Flowers
Evan, read the opening post at https://groups.io/g/QRPLabs/topic/qcx_mini_critical/78967188?p=,,,20,0,0,0::recentpostdate%2Fsticky,,,20,2,20,78967188. I do believe it answers all your questions. In short: Add the cap, no other mods needed, no further issues.
-- Ryan Flowers - W7RLF MiscDotGeek - QRP and More |
|
|
|
Hans Summers
Hello Evan, all I have done some detailed testing of this 10uF 10V MLCC capacitor... First of all, this is a good quality MLCC 10uF capacitor, made by Samsung. I have used Samsung Galaxy phones for the last 5 years so I like Samsung. But this is merely coincidence. Anyway. Here's what I found. 1. Capacitance vs Voltage Refer to the attached graph which shows my measured capacitance vs voltage for each of the three capacitors under consideration:
Experimental method: I wired a tactile push-button switch in series with the cap, and a 100K resistor across the capacitor; then push the button to charge the capacitor (from a variable voltage supply) then measure the voltage decay on my oscilloscope and use the oscilloscope cursors to measure the time required for a 10% decay in voltage. Use the formula for capacitance voltage decay to calculate the capacitance. Repeat for various supply voltages and plot the graph. The notable result is that the 10uF electrolytic and tantalum capacitors measure very close to 10uF and show little if any variation with supply voltage. What slight variation there is, could be measurement error. The MLCC 0805 SMD 10uF 10V capacitor shows an enormous variation in capacitance with voltage; at 14V supply the capacitance is 1.3uF. I believe this is normal and the rated "10uF" value is the small-signal value. For this application of a capacitor at the supply voltage input to the voltage regulator, if an MLCC were to be used then the rated voltage ought to be many times higher than the operating voltage. A more suitable capacitor for this position in the circuit is electrolytic or tantalum. 2. 10uF 10V 0805 Samsung SMD capacitor failure mode and voltage I applied voltages up to 31.6V (the upper limit of my 30V 5A linear variable bench power supply), each time for several minutes (steady) and rapid pulsed on/off operation and I could not force this capacitor to fail at all. I also found in the capacitor datasheet some mention of "withstand voltage" testing at 2.5x the rated voltage. I do have two PSUs so theoretically I could wire them in series and continue up in voltage to try and force failure. But given 31.6V is so so far above the likely operating range of the radio I think this is unnecessary and would cause some considerable complication in my lab setup here (e.g. is it correct and safe to wire the two PSU outputs in series etc). It leads me to the unprovable conclusion that the capacitor is quite unlikely to fail at normal QCX-mini voltages in the 12-14V range (or even outside that). I know you will say that if it is operated beyond its specified rating then its lifetime will be reduced... I have no way of testing that in a reasonable timescale, but I think the fact it doesn't fail at 2.5x the voltages we want to use it at, provides a significant and sufficiently safety margin. 3. AMS1117 voltage regulator failure voltage During this testing I also had an AMS1117 voltage regulator IC connected to the same supply voltage. It was connected permanently, not downstream of the push-button that was used for the capacitor experiments. As I gradually ramped up the voltage for the 10uF 10V capacitor capacitance value measurement graph, I also observed the state of the AMS1117. At 28V, the AMS1117 failed. I imagine it would fail at lower voltages under pulse conditions such as power-on. The failure mode was as usual, a short-circuit which is nasty, but I don't think it is unique to the AMS variant of the 1117. 4. AMS1117-5.0 source I also investigated with the PCB A and supply chain and was informed that the AMS1117-5.0 was procured from a reliable and reputable source, and has been used many times before in various PCBA without issue. I do not believe the AMS1117 is at fault here, it is not a fake or a poor quality device. CONCLUSIONS
Executive summary: Just add a 10uF electrolytic or tantalum and BE HAPPY. 73 Hans G0UPL http://qrp-labs.com On Mon, Dec 21, 2020 at 9:40 AM Evan Hand <elhandjr@...> wrote: Hans, |
|
|
|
Dave VE3GSO
That is a fascinating graph, and leads me to think that if one were to use these capacitors in an audio low pass filter then the cut off frequency could be tuned by a suitable control voltage. Very much like a variable capacitance diode is used at RF.
toggle quoted message
Show quoted text
Dave On Dec 21, 2020, at 08:25, Hans Summers <hans.summers@...> wrote:
|
|
|
|
I agree Jim,
When this issue first became known I thought the suggested electrolytic thru-hole part soldered in parallel was a decent solution but now I'm not feeling good about leaving the inadequate part there in parallel. If one of them fails in a bad way down the road it could still cause grief so I will remove the SMD caps from the board and replace them with the proper part. Another small order to Digikey....to save us the hassle of the parametric search would you mind to post the digikey PN for the 0805 part you referenced with a good ESR and other ratings? TNX Joe "I use a lot of 10 uF 25v 0805 ceramic caps in the 7805 regulator section of my TSW kits and buy them from either Digi-Key or Mouser so I’ll just change out both suspect caps on my Mini when it gets here in a couple days and not look back."
|
|
|
|
Dave your comment reminds me of discussions (sometimes heated!) in guitar tube amplifier forums where the topic was about how such voltage sensitive capacitors used in areas of old tube preamps could affect the tonal quality of a guitar amplified through such circuits, where on the one hand you had those highly knowledgeable engineers to whom a capacitor is a capacitor is a capacitor and all caps of a given value have the same frequency determining effects, and on the other hand the sworn testimony of those with no engineering knowledge stating having experienced a marked tonal change simply by putting a modern (voltage stable) cap of the same capacitance value in a vintage amplifier. Your idea certainly has merit. I'd be curious to hear about your findings if you ever decide to do some experiments....Joe
|
|
|
|
Evan Hand
Hans,
Thank you for the detailed analysis of the possible failure. I am very comfortable with doing the recommended fix now based on the data from Samsung and your tests. I feel that it is the best it can be and sorry that your supplier of the populated board let you down. Thank you also for the great kits that you supply. If you check your sales records you will see that I am an avid purchaser of your kits. All have exceeded expectations. I am most impressed with your attention to detail and the excellent manuals provided. Happy Holidays to you and yours. Now off to build the QCX-Mini. 73 Evan AC9TU |
|
|
|
Likely they used the same cap for input and output. Those LDOs are generally used for smaller voltage drops than the mini and they probably did what they always do with that regulator. They might not have realized it was a 12V-13.8V input. The 10V part on the output is perfectly fine. Probably no conspiracy there, honestly. Stuff happens. On Sun, Dec 20, 2020, 5:28 PM Michael Greene <kn6ize@...> wrote:
|
|
|
|
Jim Manley
Fred, Hans provided a BOM with very specific part values and they made a substitution for a part that they had no idea what the purpose of it was. The result was a disaster. When the layout files and BOM for the Original Flavor Raspberry Pi Model B were sent to a Chinese manufacturer, they made an unauthorized substitution for the Ethernet RJ-45 jack. A likely young, overzealous, inexperienced engineer looked at the specified part and saw that it was tens of cents more expensive than another part that appeared to be just yet-another RJ-45 jack. What he/she didn't know was that the part specified has integral inductors that isolate the Ethernet cable conductors from the PCB from high voltages and currents.that can destroy the circuitry, and potentially people. When the first 2,000 boards were received and samples were tested, everything seemed OK the first day in the lab. One engineer took a board home and connected it to and Ethernet port on his Internet-connected router. The board, which had worked perfectly in the lab, couldn't establish a connection to the home router, and when he went back to work the next morning, he related what he had experienced. A networking engineer asked him a number of questions, and one was how long the Ethernet cable was to the router, and because the router was in a remote location in the house, he had to use a 50-foot cable. The engineer went and found one of the same length and when they connected the Pi using it ... the board couldn't connect to the router in the lab, either. The networking engineer said he would be back in about an hour and took one of the pre-production prototype boards along with the production sample. When he got back, he placed two printed images on the workbench in front of the gathered team. He asked if anyone could see any differences in the images, which were magnified X-ray views of the Ethernet jack on each board. Someone soon said, "There are eight black blobs in this one, but not the other one." The networking engineer said, "Yep, those are integral isolation inductors that are used to eliminate the need for discrete ones taking up real estate on the PCB, and when the Chinese substituted the jack without them, the impedance using short cables was close enough to allow the board to connect to the router. However, when the long cable was used, the lack of those inductors caused a significant enough impedance mismatch to interfere with the network connection. The manufacturer ate the costs to ship the 2,000 boards back to China and then to the UK via express services, reworking replacement of the incorrect jacks or building new boards with the correct boards, and contract penalties for failure to deliver promised volumes on schedule. I don't think we have to wonder much what happened to that inexperienced engineer - they were getting reeducated in one of the Westernmost provinces ... and have mysteriously succumbed to a lack of water, food, and medical care. Stuff happens, but these are intentional acts that violate contracts. Does anyone else see what our future is going to be when we can't even get the Chinese manufacturers to adhere to a BOM? The only good news is that manufacturing jobs sent overseas to places like China are being reshored as robotics are being widely fielded in domestic manufacturing, and installation, programming, power and lubrication are applied, support for robots is needed. Even a slave laborer in Xianjing Province can't compete economically with robots, because the former need copious amounts of food and water, work breaks, sleep, medical care, vacations and pregnancy leave, endless training, etc., and the latter don't. Anywaaaaay ... Jim KJ7JHE On Mon, Dec 21, 2020 at 8:23 PM Fred Spinner <fred.spinner@...> wrote:
|
|
|
|
And even worse, they measure 8.4 uF. Both of them. Just did it. Cheap meter but my box of electrolytics measure close enough to what they are marked so I am sticking to it. So yeah they are in the +/-20% rating but just. -16% if I calculated correctly. Not big enough. At least in my case the SMD parts and reflow look just about perfect on my board under the inspection microscope. Big difference from many of the pictures on this group. No evidence of cold joints either. I think 0805 or 1210s would have been more appropriate for bypass caps. Then chip tantalums could have fit and this problem never would have happened. Overall it is a pretty sexy kit though. We will get there. On Mon, Dec 21, 2020, 9:58 PM Jim Manley <jim.manley@...> wrote:
|
|
|
|
FYI, I happened to stop in at their store in Ocala, FL yesterday and found this. Yor mileage may vary, very small & I think it will work fine. Vakits.com Paul K2DB On Mon, Dec 21, 2020 at 9:55 AM VE3VXO <ve3vxo@...> wrote: Dave your comment reminds me of discussions (sometimes heated!) in guitar tube amplifier forums where the topic was about how such voltage sensitive capacitors used in areas of old tube preamps could affect the tonal quality of a guitar amplified through such circuits, where on the one hand you had those highly knowledgeable engineers to whom a capacitor is a capacitor is a capacitor and all caps of a given value have the same frequency determining effects, and on the other hand the sworn testimony of those with no engineering knowledge stating having experienced a marked tonal change simply by putting a modern (voltage stable) cap of the same capacitance value in a vintage amplifier. Your idea certainly has merit. I'd be curious to hear about your findings if you ever decide to do some experiments....Joe |
|
|
|
Can not say for certain this is the issue here, but some of the assy houses in the far east will make part substitutions as part of a cost saving( aka increase profits) effort. They purchase components in very high volumes for the cost savings then try to fit their inventory into BOMs they are given. All contracts are different, but for very high volume manufacturers, they may even share the cost savings with the design firm. In some cases for smaller volumes, the substitution goes unannounced to the design house. It is likely this is the case here. What is amazing is I have seen this happen to large companies who have taken months to troubleshoot. Hans deserves a big thumbs up for his open and rapid diagnosis.
|
|
|
|
g4edg
As the original C38 is rated at just 10v it is still under stress, wouldn't it be a good idea to remove it after fitting the new C38?
Steve G4EDG |
|
|
|
On Thu, Dec 24, 2020 at 12:02 PM, g4edg wrote:
As the original C38 is rated at just 10v it is still under stress, wouldn't it be a good idea to remove it after fitting the new C38?See Hans' past at the bottom of this thread:
|
|
|