Hermes-Lite 2.0 Build6 Group Buy Assembly Problems

[Steve Haynal]

Hi Group,

Elecrow informed me that roughly half of the units in the build6 group buy are not passing the screening test. They are failing in the same way that 3 units failed in the last build, with TX registering too low when measured by RX in full duplex. I ended up replacing the AD9866 on the last 3 failing units after numerous tests described in this group. My hypothesis was that the batch of AD9866 parts they obtained in China were low quality. So with the current build6 run, I ordered the AD9866 parts directly from the factory to ensure high quality. I now suspect there is some problem with Elecrow's handling or assembly of the AD9866 parts. We need to get to the bottom of this before I organize anymore group buys. Here is what I am doing. More ideas and input from the group are welcome.

* Review with Elecrow the process they used, ensuring the thermal pad is soldered, and no questionable steps occurred during handling and assembly.

* Once I have some more data from above, contact Analog Devices for their help.

* Elecrow will ship good units to Oceania, Asia and Europe customers first. North America customers, to whom shipping is the least for me, will have their units repaired and delayed.

* Elecrow will ship failing units to me. I will try again to see if I can figure out the problem.

* Purchase replacement AD9866s if needed with the insurance pool funds, have Zel Pro repair the boards, ship to North America customers.

73,

Steve

KF7O

[James Ahlstrom]

Hello Steve,

I would ask Elecrow for a high resolution photo of the stencil they are using. Check if it is correct for the new PCBs without the solder hole, or whether it is the old one.

If any of the defective HL2s are not claimed, I would like to purchase one. Do not repair it, send it to me unchanged. Maybe I can see something.

Jim

N2ADR

[martin...@gmail.com]

Hi all,
Does the rf switch, U10, switch along with K2?
The reason that ask is the Quisk Raspberry pi test image does activate K2 during the full duplex low power test.
I know some boards did pass and you would expect them all to fail the test if this were the case but maybe they are just on the edge of passing...

Cheers

73

Martin
VK7MA

[Steve Haynal]

Hi Group,

More information is coming in from Elecrow. It turns out that *no* units are passing the screening test. This is hard to believe and I hope there is something different that is affecting the screening test. The final failing test is not the best as it assumes a fairly fixed TX/RX isolation. It may be that this batch just has more TX/RX isolation, perhaps because the large thermal pad access holes have been removed. Elecrow is seeing smeter readings from about -90 to -105, when the passing range is -80 to -90. Still, with previous failing units there was a low signal registered via internal coupling within the AD9866 (no signal measured on the Tx DAC pins), so I want to be very careful. The new plan is that Elecrow will send me four units for testing. I requested units across the range of smeter readings they see. This will delay shipment to buyers by at least a week or two, but I'd rather ship known working units. Below is the correspondence with Elecrow.

73,

Steve

KF7O

Hi Lily,

Thanks for the answers to my questions. The weak signal reported by the test may be a problem with the test, a problem with U7 AD9866 or even some other problem. It is a good idea to ship units to me to test. I already paid to ship 4 units to me, so please ship 4 units as follows:

1. The unit that passed one time if you can still identify it. Otherwise a unit that has "Error: TX smeter" near -90.

2. A unit that has "Error: TX smeter" near -90.

3. A unit that has "Error: TX smeter" near -95.

4. A unit that has one of the lowest values, "Error: TX smeter" near -105.

Please store the remaining units until I update the test or let you know what the problem is.

Also, please e-mail:

1. A high resolution photo of the stencil that was used during assembly if still available.

2. High resolution photos of the front and back of 1 unit that will remain in your storage, and is not sent to me.

Best Regards,

Steve Haynal

On Tue, Apr 17, 2018 at 4:14 AM, wrote:

Hello, Steve,

Here is the response:

1.Not all the 25units failed in this step.Now we just have 24pcs completed boards because 1pc boards missing components supply by Digikey.1 of 24pcs test passed but when we test it in second time, the sucessful one also failed.

2.ERROR: TX smeter is not stable,sometime they are -90,sometimes are -95 and the maximum is about -105.The data will change  with the Net wire.We use 3 different net wire but the same boards will have different data.

3.We assembly your boards with machine with 

4. The pad on bottom of AD9866 assemblying well because we removed the chips and then reassembly it again and found everything went well.

5.During assemblying ,our normal temperateu is about 235degree to 240 degree.

6.Our assemblying worker must wear anti-static clothes and cap during assemblying.

7.All project must follow the ISO 9001 certificate.

8.Yes,we always keep  AD9866 dry and we do not  open the package when we recieved the parcel.The IC chipes will be assembly within 1 min in the boards after open the package.They do not baked to remove moisture.
 

9. Failing units are powered for less than 5 minutes before programming and test was executed

10.Before programming and test, power do not  applied to the units


I am wondering,could we ship 2 or more boards to you firstly so that you can check why it happen and then tell me ?So I will ask our engineer to repair them,is it ok?

Thanks in advance.
 

Best Wishes,

lily


Elecrow Electronic store
Wholesale, Arduino, Raspberry pi, Robotic, PCB vendor, PCB assembly
Web: www.elecrow.com  
Phone: +86 0755 23204330
Add: F-building 8 floor, Fusen industry parkGushu Hangcheng road, Bao'an Ave,Shenzhen city,China 518126
 

 

--------------Original Message-------------
From：
To：Lily 
Cc：
Subject：Re: order#87976 test problem
Date：2018-04-14 

 

Hi Lily,

 

I reviewed the pictures and movies you sent and Elecrow is doing the programming and test properly and using the most recent Raspberry Pi image.

 

***Exactly how many of the 25 units are failing this final step?

 

At the end of the test in the movie, there is the message "ERROR: TX smeter is -98."  Passing boards must see a reading between -80 and -90 here. 

***Do all of the boards fail with a large error close to -98, or do some boards fail with an error close to -90, such as -91,-92 or -93?

 

This failure is the same failure that we saw with 3 units from the last build of 10 units. With the 3 failing units from before, there was damage to U7 AD9866BCPZ and I had to replace these ICs. This time I ordered the AD9866BCPZ directly from the factory to be sure they were good. I would like to discuss these failures with the AD9866 factory, so please provide details of your assembly process, including answers to these questions:

 

***What process was used to assemble the boards? (Hand, machine, stencil and paste, wave soldering, etc.)

 

***Was the exposed thermal pad on the bottom of the AD9866 soldered to the PCB pad?

 

***What was the highest temperature the AD9866 parts were exposed to, and for how long?

 

***Was any rework required for the AD9866 parts, and if so, what was the highest temperature used and for how long?

 

***Were ESD protection guidelines followed throughout?

 

***Recently Elecrow was  ISO9001:2015 certified. Did the assembly of the Hermes-Lite 2.0 units adhere to these requirements?

 

***When the AD9866 parts arrived, they were sealed to keep them dry. When they were opened, did the indicator still show that they were dry, or did they have to be baked to remove moisture?

 

***During programming and test, were the failing units powered for less than 5 minutes before programming and test was executed?

 

***Before programming and test, was power ever applied to the units?

 

Best Regards,

 

Steve Haynal

 

 

 

 

On Fri, Apr 13, 2018 at 3:24 AM, Lily wrote:

Dear Steve,

Could you log in this websie and then down load the test process please?

https://www.elecrow.com/download/87976-test-process.zip 

There are an half of boards have such problem and we do not know why,some boards test passed but some of them will fail in the second test,what is the problem caused?

Could you help me to check it please?

 

Best Wishes,

lily

[Steve Haynal]

Hi Jim,

I requested the photos you wanted. If I can't easily figure out what is wrong with the four units Elecrow is sending me, I will send you one for your help.

73,

Steve

KF7O

[Steve Haynal]

Hi Martin,

Separate lines are used for the RF switch. The RF switch should always be to the low power output when in low power mode. I hope the problem is similar to what you found where the test parameters were to blame.

73,

Steve

KF7O

[Jim Amos]

Hi Steve,

An issue that I've ran into with low volume manufacturing and SMT devices is moisture contamination causing package failure during reflow.   You mentioned that you bought them, we're they delivered on a sealed try?   If not, were they baked prior to assembly?   If the vacuum packaging was opened before manufacturing (there's really only a relatively small window for moisture contamination) and not baked prior to assembly this could account for the high failure rate of the parts.   I've seen this happen with WiFi SIP modules that were bought through distribution.  They had been opened, and repackage for delivery because they were being bought in quantities that were not an integer number of the tray size.   They were then not baked prior to assembly and failed.

Once we get the pictures we should check for package damage on these parts, the damage from moisture could be visible.

73's
Jim N8CAH

[Steve Haynal]

Hi Jim,

Yes, moisture is a concern. I asked about this and they said the packaging was opened just before assembly. The parts were directly from the factory, sealed and with moisture indicator inside.

73,

Steve

KF7O

[Steve Haynal]

Hi Group,

Below is a link to high resolution photos from Elecrow. They are on Google Drive due to size. Photos of the stencil are promised for later this week.

https://drive.google.com/open?id=10CdE5zsx8l5FgGtj_KKmqyQp6E2O1o1N

73,

Steve

KF7O

[Steve Haynal]

Hi Group,

Elecrow now has at least 12 boards passing the screening test. They found a sensitivity between the length of the ethernet cable connecting the HL2 to the RaspberryPi and the screening test. There appears to be more TX/RX coupling with a longer ethernet cable. Elecrow wants to start shipping passing units but I asked them to wait a few more days. The four units they have sent to me will arrive on Monday. I want to do a thorough testing of these units and provide Elecrow with a more robust screening test. Given all the confusion, there still may be some new weird issue with these units and I'd rather catch it before shipping the units. Thanks for your patience!

73,

Steve

KF7O

[Steve Haynal]

Hello Group,

Good news! My 4 Hermes-Lite 2.0 Build6 units actually arrived early, today instead of Monday. Three of them passed the screening test here without issue when using a 2M ethernet cable. I did try the shortest ethernet cable I have (30cm) and saw tests fail. The fourth unit was not supplying power to the opamp TX preamp. Although it looked fine under a microscope, pin 5 of U12 wasn't connected. Resoldering that pin fixed the problem.

I did try the short ethernet cable with one of my older boards, but found no failure. The old board has the relay installed plus RF connectors, so I suspect there is more metal to provide coupling. Also, the RX/TX isolation on the new boards just seems to be higher such that the coupling is low enough to see changes due to ethernet cable length. Even with the long ethernet cable, the smeter reading of the coupling is low on the new boards compared to the new boards. I will run more tests this weekend, but do not suspect any problems. 

I will relax the test parameters for Elecrow, plus add an additional test. I am worried about the same AD9866 problem we saw last time and want to be able to catch that. With that problem, changing the TxDAC gain had no effect. I will add a test that changes the TxDAC gain and must see a reasonable change in the TX output to pass. Elecrow should have this new test by Monday morning when they start work, so boards should start shipping on Monday.

73,

Steve

KF7O

On Friday, April 20, 2018 at 9:25:20 AM UTC-7, Steve Haynal wrote:

Hi Group,

[Steve Haynal]

Hi Group,

I've updated the Raspberry Pi tests described on the wiki. The Raspberry Pi image is updated. The test is relaxed. An additional test to detect change in Tx amplitude when Tx levels are adjusted is added. This added test is for the AD9866 problem we saw in 3 units from the beta5 build. There is now a "Retest" icon to only run the test step and skip the programming step. Finally, a short 10cm temporary antenna is required to make test results more consistent. See the Retest section of the wiki page for details on this temporary antenna.

I completed one unit and have run RX and TX over the weekend. TX output is at the expected levels and clean. RX is seen in past units. The AD9866 is running a little cooler but is still the hottest component on the board. I couldn't find anything out of the ordinary. Elecrow will start shipping units passing the new test as soon as possible.

73,

Steve

KF7O 

[Steve Haynal]

Hi Group,

All the units are now passing with the new test! This batch just has better RX/TX isolation. Elecrow should start shipping units soon. Elecrow did add an unexpected charge of ~$6.50 per unit due to exceeding some amount limit for import/export. They have to use and pay for an import/export agency to get around this limit. This will be taken from the insurance pool.

73,

Steve

KF7O