SG9665GC V1 teardown (and some EMI measurements/shielding)

[Hex]

Hi,
welcome to everyones favorite segment: Teardown!

Foreword
After swapping my car over a year ago I dind't install the SG9665GC again because back then I had issues with RFI. I could either use amateur-radio and listen to DAB+ radio or record the traffic and listen to music from the USB drive. Normal FM radio was also getting noisy when the SG9665GC was running. Back then the DAB aerial was mounted on the windscreen, approximately 50cm away from the cam. The USB power cable for the cam is DIY made of 2x 0,5mm² litz wire (AWG21 - AWG20) to fit around the windscreen and behind the airbags.

After the last RFI-thread I thought it would be a shame to let the SG9665GC rot in the basement. Because any warranty is long gone I could also do a teardown and some shielding attempts.
Now the SG9665GC is again mounted on the windscreen but the DAB aerial is mounted near the hatch, approximately 2m away from the cam.

Teardown
I couldn't find any teardown of the SG9665GC, so here we go. The Cam is V1.
The case is held together by four screws, two might be covered by labels.

The camera must point upwards a little bit to disassemble, also mind the cable for the speaker.

To take out the sensor/lens assembly remove two screws which hold the adjustment knobs.

Now you can access the main PCB

Remove the tape from the capacitors and mainboard. The PCB is marked with "B40 MB_V21_20150129"

Now you can see the Novatec processor NT96655BG with RAM HXB15H1C160BF (couldn't find a datasheet, probably its 1Gbit) on the right and FLASH 25L3206E (32Mbit) on top. Below there is a battery and a crystal for the RTC and also some pads, probably for a serial interface.

The connector for the FFC (flexible flat cable) towards the sensor/lens assembly is behind the caps

Lift the caps carefully, they are attached to the FFC and connector with adhesive tape

Lift the black clip on the connector upwards to disconnect the FFC. Don't forget to mark the FFC...

Below the FCC the main crystal/oscillator was hidden

Now you can continue with the sensor/lens assembly

The case consists of two plastic part holding together with four clips. The PCB is screwed to the front of the case

The two bigger screws in the middle of the PCB hold the lens assembly, the screws diagonally hold the PCB.

Unscrew the PCB to get out the sensor/lens assembly

As promised the lens and lens holder is made of metal. The lens is glued in with a good amount of glue which does not come of easily (in case you want to adjust the focus).

The sensor PCB is marked with "IMX122_V01 840_SEN 20141230" and contains probably the low noise LDOs (low dropout linear voltage regulator) requiered for the sensor

With the sensor/lens assembly removed you can unscrew the three screws holding the mainboard and remove it

Buttons

And LEDs

The LCD is also connected with a FFC to the mainboard and hold by the black platic frame clipped on the mainboard. Unclip carefully and lift off the LCD

Slide the black clip backwards to disconnect the FFC

Remove the LCD

Teardown finished!


RFI part in the following posts...

Regards
Hex

 

[flip9]

Thanks for the pics. Saves me from having to tear mine apart.

Can you include the GPS unit when you do the RFI tests?

 

[Hex]

Can you include the GPS unit when you do the RFI tests?

The measurements with the spectrum analyzer are already done, unfortunately without GPS receiver because it is already mounted inside of the car (and inside the lab there is no GPS reception). Maybe I'll do some measurements inside the car.

Because there are no more than 20 Pictures allowed, here the next few measurements.

Temperature measurements
Took some measurements with a thermal camera. Left is the infrared image (white = hot, blue/black = cold) and on the right the normal image. The temperature readings are the maximum of the box. The measurements are taken at room temperature (~23°C) and when the camera was running a few minutes.

Mainboard

The hotspots are the Novatec processor (Bx2) of course but also a diode (Bx1, SMD marking "SL")
Not sure if it is for reverse polarity protection or a Zener to limit the voltage across the caps. When off, I measured the same voltage across the diode as across the caps. Only the cathode is directly connected to + of the caps, anode is high impedance (a few kOhm or so).

The hot diode is blocked in this one by the capacitors

Novatec processor (Bx1)

Diode (Bx1), Novatec processor (Bx2) and sensor PCB (Bx3)

Diode is blocked again

Novatec processor (Bx1) and RAM (Bx2)

Sensor PCB

The temperature of the diode bothers me a little bit, already reaching 80°C at room temperature is not directly problematic but surely limits it's lifetime. Assumed we get 50°C environment temperature when the sun hits the cam and 60°C on the PCB the diode would get around 120°C hot. With typical limit of 150°C junction temperature there is not too much headroom.

 

[flip9]

Not sure if it is for reverse polarity protection or a Zener to limit the voltage across the caps. When off, I measured the same voltage across the diode as across the caps.

I'd expect it to be a schottky for backflow protection from the caps. Maybe its power rating is under rated?

I would expect it to heat up a bit from the inrush charging of the capacitor (if theres no resistor limiter) but since you say its been running for a few minutes its probably not the case.

Can you unsolder and check the Vf of the diode?

 

[thancam]

if memory card slot can move to other cool section, or diot and processor can rearrange, then the dashcam would be like mobius: can use almost any kind of memory card.
by colour, i think temp of memory slot has to be at least 50C. It's pretty high.

 

[jokiin]

if memory card slot can move to other cool section, or diot and processor can rearrange, then the dashcam would be like mobius: can use almost any kind of memory card.
by colour, i think temp of memory slot has to be at least 50C. It's pretty high.

there's only so many ways you can lay them out before the camera needs to change size and shape to accommodate the changes, this is one of the first V1 boards and there have been a couple of revisions and component changes since then so the results from the V2 and later versions may differ also, there's some thermal measurements of the Mobius on the site here somewhere that someone else did a few years ago, don't see it often as not too many people have access to thermal imaging cameras

 

[Hex]

if memory card slot can move to other cool section, or diot and processor can rearrange, then the dashcam would be like mobius: can use almost any kind of memory card.
by colour, i think temp of memory slot has to be at least 50C. It's pretty high.

Never had any issue with the micro-SD. At the moment Im using a SanDisk 128 GB SDSQUNC-128G-GN6MA.

Your guess was pretty good, here are the temperatures at the card connector

Don't let you fool by the "cold" case of the connector, in LWIR bare metal has a very low emissivity and acts like a mirror reflecting the temperature of the surrounding. A more accurate reading is the SD-card shining trough or the PCB nearby.

 

[Hex]

RFI measurements
Measurements were done with a H-Field probe connected to a Keysight signal analyzer. I probed around the guts of the camera to see what is working or emitting on which frequencies. Most of the captures are done with peak hold to capture glitches.

Wideband measurement up to 6 GHz
At first a reference shot with the SG9665GC turned off to identify other transmitters

Near marker 1: analog FM radio transmitters
Near marker 2: digital TV transmitters and GSM900
Between marker 2 & 3: LTE, GSM1800, ...
Marker 3: WiFi 2,4GHz
Marker 4: WiFi 5 GHz

And with the SG9665GC turned on and probing around

As you can see most stuff is going on under 1 GHz so we change the upper frequency to get a better resolution.
The Peaks near marker 3 & 4 can most likely ignored because they are emitted by WiFi devices nearby.

Measurement up to 1 GHz
At first also the reference shot

This time I could not guess what transmitters are causing these peaks.
Anyway, here with the SG9665GC turned on and probing around

The peak at 111 MHz is common at pretty much every measurement. The markers are sorted (mostly) by decreasing power.


Now for the individual components and locations
Power input

This one was done to see how much is radiated by the power cable and what potential ferrite needs to block.

SD-card slot

Here you can see the intense amount of noise generated. The FFC from the sensor is coupling probably as well...

FFC

The interface towards the sensor is probably the major contributor to the emitted RFI
Measurement in average-mode to see the more or less continous emission

Novatec processor

Of course these measurements are are not relating to a credible power level but are an indicator for possible RFI. As you can see most of the emitted peaks are near frequencies I need to have "clean".
111 MHz: analog FM radio
148 MHz: HAM radio
185 MHz: DAB radio
...
For getting credible values something like a GTEM cell is needed. It would be interesting to get the measurements (which maybe have been done for CE and FCC conformance tests) from the manufacturer. Especially the difference between V1 and V3 would be valuable because my V1 has a very old PCB. I would not hesitate to buy a SG9665GC V3 if I knew it would perform better in terms of RFI.

 

[Hex]

RFI shielding
At the moment most shielding is done floating. While not shielding as good as galvanically connected shields it does provide a lower capacitance and thus not interfering with the high speed signals.
The general approach is: Cover parts and the PCB with capton tape to insulate, then add adhesive copper foil.

At first the back of the LCD. Apply copper foil...

...and capton tape

Here the capton tape is put on top of the copper foil so it is not shorting anything on the PCB.

Apply some copper foil to the case

Make sure to cut out parts of the reset-button and the SD-card slot

Put the mainbord back inside the case

Insulate the components with capton tape

And apply copper foil

And connect to the shield on the side of the case

Here the processor and RAM are not covered with capton tape because I want to use the copper foil as additional heatsink.
For something becoming a heatsink you have normaly three ways to get the heat away:
1. Radiation
2. Convection
3. Forced air or liquid
Forced air is clearly not possible because we do not want a humming fan inside the cam.
Convection is restricted because the PCB is mounted horizontally and thus the major part is radiation.
Because the bare copper does not radiate heat very well it needs to be covered with something that has a high emissivity. You can find a bit explanation here: https://en.wikipedia.org/wiki/Leslie_cube

So to increase the emissivity capton tape is applied on top

The usual taping continues at the sensor PCB. First capton

then copper

The FFC has already some tape, so just add the copper foil

Connect to the sensor and put in some bends

Install the sensor in the case

At this point I put the cam inside the car and did a test drive to check the DAB reception. Unfortunately it did not improve much. Because more is better I improved the shielding further.

Connected the floating shield from the FFC with the mainboard (yes, the wire should be as short as possible but unfortunately it needs to move so I can adjust the angel of the lens)

Add some capacitors directly across the power from USB plug (100 nF and 1µF MLCC in parallel) and add more kapton for insualtion. You know where this is going...

Add more copper foil

Somehow this picture is a bit buggy. Try i.imgur.com/AQLNDpA.jpg


This is the current stage of shielding. It did improve a bit as you can see from the measurements coming next. Unfortunately when the cam is running the DAB signal level is still reduced by ~30%. So when driving further away from the transmission tower there are still locations where no reception with running cam is possible.

 

[Hex]

RFI measurements after shielding

I don't know why, but it seems like I had captured in averaging-mode again. So the peaks are not as high as with max hold from the previous measurements.

USB port

SD-card slot/FFC

Sensor Board/FFC

Processor/RAM

LCD

Conclusion
The current shielding is helping a little bit, unfortunately the cam is still not as "clean" as I want it to be. The next steps are probably:

• Shielding the mainboard completely (except for the connectors and LEDs)
  
• Solder the shield directly to GND
• Use a more flexible shielding (braid) for the FFC
• Shield the sensor PCB completely and connect it to GND
  
• Solder the braid directly to the shield from the mainboard and sensor
• Add a filter to the power (and maybe GPS) connector

These points should cause the guts of the cam to be completely shielded like in a Faraday cage.
If this still won't help it might to be taken as given...

Regards
Hax

 

[SawMaster]

Top-notch work Hax

I think the braid and shield-to-ground changes will be what gives the most improvement, though I admit I know almost nothing of image sensors. And indeed I see several frequency peaks where they could be a problem for a lot of folks I don't know if those could be easily shifted to less troublesome frequencies as some of this must be inherent in the workings of the processors etc so that the can manufacturer is stuck with them. Some changes would be welcome though

I do know that with today's trend toward 'smart' devices with wireless connectivity we are going to see far more RFI issues in the future with everything We are doing this to ourselves, and often without any real need I fear that until everyone has big problems with it that nothing will change to keep it from happening, and I do worry what those changes will be.

Many 73,
Phil

 

[Hex]

I almost forgot some screenshots taken with a SDRplay software defined radio (SDR).
The SDR had a 2m λ/4 fake Nagoya antenna and a FM radio blocking filter installed, the cam was 1m away from the antenna. Unfortunately I forgot to note down the gain values before shielding, so the screenshots before and after shielding are not 100% comparable but give a little indication how the signal looks.

DAB radio before shielding

DAB radio after shielding

2m amateur radio band before shielding

The "waves" are most likely artefacts from out of band interference

2m amateur radio band after shielding

And the interference at some of the peak-frequencies measured with the spectrum analyzer
111 MHz before shielding

111 MHz after shielding

185 MHz before shielding

185 MHz after shielding

As you can see the "frequency zoo" is a little bit more contained with shielding and the SNR at DAB radio improved from ~30dB to ~40dB.

 

[DT MI]

What I'm most impressed about with all this is that you've got some really nice toys to play with.

 

[jokiin]

what is the serial number of your camera, I can tell it is a very early one due to components used but helpful to know which production run this was from

 

[sp5it]

Isn't RFi coming from power supply of the camera?

 

[jokiin]

Isn't RFi coming from power supply of the camera?

the power supply he would have is a very early one, we changed from using that one over 2 years ago to a better one

 

[Hex]

Extreme shielding
As promised, this is and was my last attempt to make the cam "clean".
Good news first: IT IS CLEAN (at least with the first checks)
Bad news: It took nearly 5 hours to do so

I started with the mainboard again. Cover with capton and scrape of the solder stop to access the groundplane

Cut away the capton near the buttons and where you want to solder

If you want the white LED to stay white (and not some capton colored yellow) you can also replace the capton above the LED with some clear plastic film.

Cover with copper foil and solder the segments together and to the ground plane

Connect the LCD, don't forget to add copper foil over the FFC from the LCD.

Next is the Sensor/Lens assembly. Cover with capton

Put the copper foil on top and solder it to the ground plane.

Get some copper braid (mine was from a shielded cable) and stretch it until the FFC fits into it.

Put the FFC into the braid

Connect the FFC to the sensor PCB and solder the braid to the copper foil

Plug in the FFC, solder some of the braid to the copper foil and stuff the rest under the capacitors.

As final step the rest of the braid was connected with copper foil because I didn't want to solder near the plastic connectors.

Fin!

 

[Tobi@s]

Haha good job
Appreciate your detailed posting

 

[Hex]

RFI after extreme shielding
Here are the results from the SDR. Sometimes I needed to double check if I turned it on

DAB
Cam off

Cam on

You can see a faint dotted line appear near 175,00 MHz and 180,25 MHz when the cam is on.

Before shielding with braid

2m Amateur radio
Cam off

Cam on

You can see the peaks between 148,3125 MHz and 148,75 MHz get stronger. Not sure if this is caused by the cam or the cable acting as an antenna.

Before shielding with braid

111 MHz
Cam off

Cam on

A little bit more noise.. maybe.

Before shielding with braid

185 MHz
Cam off

Cam on

Oh, look at the peaks appearing near 185,500 MHz

Before shielding with braid

And here are the measurements with the spectrum analyzer and H-field probe. These are the only spots where I could measure anything above normal ambient noise level.

Outside directly the SD-Card

On the other side near the reset-button

And near the LCD

 

[Hex]

I'd expect it to be a schottky for backflow protection from the caps. Maybe its power rating is under rated?
I would expect it to heat up a bit from the inrush charging of the capacitor (if theres no resistor limiter) but since you say its been running for a few minutes its probably not the case.
Can you unsolder and check the Vf of the diode?

Unfortunately I couldn't desolder it because I wanted to finish the shielding.
Now I measured 4,37 V across the diode when the cam is off, the same voltage can also be measured across the two caps. There is the same diode also near the GPS-connector, so probably acting as a OR-ing circuit for the power.

what is the serial number of your camera, I can tell it is a very early one due to components used but helpful to know which production run this was from

Serial is SGGC15250393. Can you tell me when it was produced?

Isn't RFi coming from power supply of the camera?

RFI can be generated by virtually everything (even corroded drain pipes can cause RFI because it acted as a diode and was modulated by strong transmitters nearby). Mostly it's generated by switch mode power supplies though.
This is the post that started my interest in the cam again: https://dashcamtalk.com/forum/threads/rfi.28662/#post-340538
Here you can see the original power supply (probably an old one) caused RFI. After switching to a good one (Anker 24W, bought a few years ago) the second source, the SG9665GC, caused also RFI when the image sensor was active.

Haha good job
Appreciate your detailed posting

Thanks, this was my goal. I also appreciate when others show some more details.