FEDEVEL
Platform forum

rdpdo , 05-19-2023, 05:34 AM

Hello everyone,

I have a project consisting of making an ADC main board with a raspeberry PI4B which will allow me to measure differential voltages from a MEMS sensor on remote micro-boards.

These boards are connected to my main ADC board with 4-wire AWG32 cables + shield. The wires transport differentials analog outputs from MEMS IC.

I wonder how to connect the shield on the side of the ADC card and on the side of my MEMS micro-cards.

Could you give me your opinion please?

Thank you !​

robertferanec , 05-20-2023, 01:07 AM

If you are not sure, the simple answer is: provide different options, test them and for production use the best one.

For example on number of boards I placed footprints between shield and board GND and tried to use capacitors or ferrite beads or 0R connections. In most cases, direct connection was the best, but, there were situations where I wanted to use something else.

rdpdo , 05-20-2023, 03:47 AM

Ok thanks you,

Some advice to place direct connection to GND only in one side (https://forum.allaboutcircuits.com/t...ielding.78211/). Do you think it is a good idea ?

qdrives , 05-20-2023, 09:30 AM

For motors we always had the best results connecting the shield at both ends.

I see that you are showing that shield connected to a 5th pin (as JST connectors are plastic). You would be creating a pig-tail. That way, shielding does not really help at higher frequencies (> 10MHz).

robertferanec , 05-22-2023, 01:10 AM

I am not super expert for this, but one side vs, two sides may only be important if both devices are powered from power main. if one of them is battery powered or if one is powering the other one what the effect of not connecting the other side would be?

What you may want to consider is:
- do you want to connect them directly or using a capacitor? For example, when connecting directly and both devices are powered from different power mains, you will have DC currents flowing through the shield. This may cause problems.
- can the cable behave like an antenna? If you have a noisy ground and if you connect this noisy ground to your cable, cable can create a very good antenna for some frequencies.

I have done some videos on shielding and emc:
- https://www.youtube.com/watch?v=F9tB6JtG7oQ
- https://youtu.be/gHF5JyJF-N4

rdpdo , 07-21-2023, 04:17 AM

Good morning,

I slowly continue my reflections on my project... Even if I am an amateur I would like to make a device as clean as possible. Could you give me your opinion on these ideas:

I specify that in my main card I will have 20MHz for the CLK of the SPI bus. I only have WiFI and FM/AM radio signals in my home. My sensor only sends a few 10ene of mV on the differential signals. The currents are of the order of 0.1mA.

- I plan to use an onboard battery (1.8V) in my sensors to power them rather than deporting the power supply from the main board. This way there would only be the differential signals to be transmitted which I could do on a 1 pair twisted shielded cable.

- I also plan to put my sensors in a metal shield and use an EMC metal gland to connect the cable to the shield. I plan to connect the mass of the battery of the sensor to the shield of the cable and also to the shield of the box.

- I also plan to use a battery for my main board so as not to pollute with the 50 Hz and the DC/DC choppers.

- I plan to connect the ground of the battery of my main board to the shield of the cable from the sensors but only with a capacitor so that there is no low frequency connection (and therefore avoid the ground loop) but an HF connection.

What do you think ?​

qdrives , 07-23-2023, 10:22 AM

There is no problem for EMC to have the power through the cable, if it is clear on the board.
Differential signals often need to be seen as two single ended signals and need/have a reference.
What do you want to "sense"? Your overview shows MEMS, they can handle SMPS.
Shield is used as a low impedance return to the (noise) source.

rdpdo , 07-23-2023, 12:32 PM

Thanks for reply,

My sensor is to sense acceleration.

My ADC have differential inputs so my idea is this but I dont know if it is correct ? It multiply the number of GND and so perhaps it is a bad idea ? (I have 32 sensors)



Else I can do that wich is less expensive by taking power from main board. In this case there is only one GND so perhaps it is better ?



Thanks you very much !

robertferanec , 07-23-2023, 06:14 PM

I am not expert for this, so this is just what I would probably do. I would buy evaluation board and try to connect the sensors by different ways. If it would not work well, I would design my own sensor boards with AD close to the sensor and transfer the data digitally.

rdpdo , 07-24-2023, 03:29 AM

Thks,

The fact I cannot use digital accelerometer is that I cannot synchronize them and so my vibration data will be delayed each with other... It is why I must use multiple analog souce which will then be all acquire in a synchronozed way.

I will use these vibration data to get frequency response for extracting modal frquency with specialized softwaez (using OMA - Operation Modal Analysys). If my data are not synchronized the extraction will be biased...

Also, is it a problem that differential signals are not referenced by the GND in the main board ?

qdrives , 07-24-2023, 02:17 PM

1) DigiKey: "This product is no longer manufactured and will no longer be stocked once stock is depleted." Octopart only mentions Digikey as a distributor, not even Chinese ones.
2) The datasheet shows DC blocking caps, so there is no Gnd reference. You could connect one to it to make is single ended.
3) The ADS131M08 is a delta-sigma ADC. How do you want to synchronize with that?
4) What is the length of the cable between main and sensor boards?

My idea:
Add a "sync" line from the main board to the sensor board.
Have the ADC done on the sensor board.
Use something like 3.3V from main to sensor board and have a local regulator to say 1.8V.

rdpdo , 07-25-2023, 04:03 AM

Hi,
1) Yes I need to speed up my project ! All analog output accelerometer with price 1-15EUR have only 500Hz in Z axis ; VA1210 have 1kHz it is the only one I found and price is only few EUR ...
2) I prefer to use differential ended so there will be less noise ; The VA1200 can do single ended but I think I will have better results with VA1210
3) The ADS131M08 simultaneously convert analog inputs and store them before next sampling time. I will read all the the values from inputs ADC before next sampling time (i will use only 4Khz sampling rate)
4) I would like length to be 1 meter

This is a very good idea thanks you ... but If I make the ADC on sensor board it will be a lot more expensive but I will check the price for ADC 24bits with 1 chanel perhaps it is not too expensive... seems to be about 10EUR (ADS1291).
Also I will need to transport SPI 20 MHz data on 1 meter and i need a light flexible cable to not induce to much low frequency damping to the accelerometer ; what kind of cable can I use ? I am thinking using an ultra thin HDMI cable (2mm outer diameter) but it's 20EUR do you know if I can have it less expensive ?

qdrives , 07-25-2023, 01:40 PM

Digikey has it as a microphone, not so much as an accelerometer.
Again, it is a delta sigma ADC, not SAR or flash/pipeline.

"to not induce to much low frequency damping to the accelerometer" -- you mean the mass of the cable versus the mass of the vibrating object?

rdpdo , 07-25-2023, 03:53 PM

Thanks for the cable reference, indeed digikey classifies the va1210 as a microphone but it is indeed an accelerometer.

Concerning the adc sigma delta, the books that I read on the analysis of the vibrations conweillznt to use this thpe of converter... but you seem to say that it will pose problem? why ?

Otherwise I saw that TI offers LVDS drivers to transport SPI signals at kk 10ene MHz on twisted pairs.

Thanks again !

qdrives , 07-26-2023, 12:56 PM

What frequency of vibrations do you want to measure?
What does the signal look like?
The 4kHz sample frequency with a delta signal is the average during the period (0.25ms).
With digital (MEMS) you can do multiple samples too and 'merge' them in the CPU to get your average.

rdpdo , 07-27-2023, 04:27 AM

Good morning,

Thanks you for your help. I want measure vibration of about 5Hz-500Hz (if I can do 1kHz it would be better).

But finally after a lot of research I found digital accelerometers at TDK with 4kHz bandwidth which can have an external clock for the ODR and therefore I can synchronize them all in this way.


So I'm going to completely change my point of view and make my sensors boards with digital accelerometers.


I will be using LVDS circuits to transfer my SPI signals to sensor boards from my main Raspeberry PI board. All boards will have the same SPI clock and same ODR clock.

On the other hand, I will have to learn to use the simulation tools for the line impedances of the PCBs, the routing of the differential pairs, the PDM etc... and since I don't know anything about it, I have a lot of work to do!

Anyway I think it will be simpler than analog eventually.

It now that I find the cables not too expensive to adapt to about 20 MHz for my SPI. I'm in the process of finding out if the flat cables you told me about are suitable (the LVDS normally sends on trosaded pairs but I can perhaps use two tracks of the flat cables...)

Thank you all!​