SN74LVC245A: Guidance to improve the SPI data transmission from MEMS sensor over 2m distance

Hello Clemens,
Thanks for your quick response. LVDS surely seems like the perfect solution for our application, however per channel cost is beyond our budget for this project. Are there any cost-optimized LVDS solutions available?

The clock frequency we plan for our use case is 1MHz & 2MHz.

Said that, for a typical industrial use case with CAT-5 high-quality shielded twisted cable, and a maximum distance of 2m, can we do equally well with 2x Level Shifters like TXU0304-Q1, that will convert the MEMS's 3.3V SPI to 5V over the cables & back to 3.3V at the MCU?

Do suggest us some way to evaluate these two options.

I made that selection as it is Automotive Qualified, based on an assumption the -Q1 ICs will be more robust overall for a small delta in cost.

Yes, your suggestions for cables have been noted, thank you!

Any ideas for lower-cost optimized LVDS devices? Also any benchmarks/articles/application notes for signal integrity tests done by TI for SPI transmissions over longer cable distances? I feel this might be a very common use case for customers to extend SPI over 1-2m distance in Industrial environments, as often sensors & the MCU board are far apart.

Can we optimize the design such that, we use LVDS DS90LV011ATMFX & DS90LV012ATMFX for MOSI & MISO lines, while we use buffer for CLK & SS lines? Am I missing something major while making this assumption? If this configuration works, we may get to optimize the cost significantly.

The TXB has many restrictions. I would not recommend it except when the signals actually are bidirectional, and no direction signal exists. And then you have to ensure that your circuit respects its limitations.

Using higher voltages reduces noise sensitivity (this is the idea behind RS-232), but not by a large amount. Very noisy environments require differential transmission.

Noted.

Are there some benchmark studies/simulations available that can help us understand the improvement that 5V I2C has over 3.3V I2C signal? 

Thank you for introducing me to LVDS for SPI & sharing the low-cost LVDS ICs, I shall be using these moving forward for our noisy environment applications.

Is there any special consideration to be given for LVDS implementation? ESD & other protection measures?

Hello Emrys,

As for robust system design, would you suggest us some prominent protection measures (except Isolation which is costly & at times not required) for I2C, SPI, and USART buses?

So far we have planned to use Poly switches for over-current protection, TVS diodes, and Ferrite beads, do guide us with any additional protection measures that we should aim for system-level robustness. Also if there are some integrated ICs for these rather than installing multiple passive components on the board.

Neet said:

As for robust system design, would you suggest us some prominent protection measures (except Isolation which is costly & at times not required) for I2C, SPI, and USART buses?

I'm afraid that's not really my area of expertise. I can tell you that most ICs aren't designed with system-level ESD protection in mind -- usually they are only built to handle minor strikes during manufacturing (2kV HBM). Anything that's exposed to contact with people or equipment should have an IEC-61000-4-2 rating. There's an app note that covers more details here: https://www.ti.com/lit/sg/sszb130d/sszb130d.pdf 

Hello Emrys and Clemens,

As we conclude this discussion, I would like to provide a concise reference for anyone seeking to extend SPI over 2-3m distances in noisy industrial environments. Could you please share the key parameters to consider when selecting LVDS drivers, as well as any special considerations to keep in mind when using LVDS for SPI over longer distances?

- Propagation delay
- Slew rate
- Output Voltage Swing
- Common Mode Voltage
- Input Threshold Voltage

What additional pointers to look for that can directly impact the SPI signal integrity &/or maximum speeds?

Thank you for your valuable input, I am now much more confident about using SPI-based Sensors & MCU communication over our requirement use case.

Hello Clemens,

Can you also suggest the use of RS-422/485 IC for similar application? Can we also use RS-485/422 drivers for transmitting SPI Signals?

RS-4xx devices are more power hungry and, in general, more expensive.

When using many channels, see, e.g., the AM26C31/AM26C32 or AM26LV31/AM26LV32; for the back channel, use any 1+1 transceiver like the THVD1500 or THVD1400.

Hello Clemens,

Can you suggest a reference design that will guide us to evaluate the use of RS-4xx for our SPI transmission over the differential bus?

This seems more promising as we have RS-4xx drivers available with local suppliers while LVDS we need to source via import.

Also, in our application the power budget is not an issue, we have a stable power supply available.

Is the following assumption a valid use case?

- I've added Buffers for Level translation & Schmitt trigger for improved noise immunity from potential cable issues.

- RS-4xx diff drivers can be AM26C31/AM26C32 or AM26LV31/AM26LV32;

- Can the same be used for I2C with P82B96 in between RS-4xx driver & the I2C Slave/Master

Kindly guide us in the correct direction.

Hello Clemens,

We want to interface a MEMS-based Sensor to our MCU via SPI, and the following Application Note from TI suggests the use of a configuration as follows:

Before we implement it, I wish to have your suggestion on suitability and if we are missing anything fundamental in simplex communication from the Sensor to MCU?

We plan to use 2 x DS8921 on both sides, and this configuration reduces the cost per sensor-SPI-MCU drastically over using 2 x LVDS.

Regards.

That application note has an isolator (which increases propagation delay) and runs at high speed. Without an isolator and running at 10 MHz, clock feedback should not be necessary.

How many SPI signals, in each direction, do you actually need?

Ohh, we have 2 MEMS sensors on 2 separate SPI interfaces on the MCU, each MEMS sensor has 3 & 4 wire configuration for SPI.

We wish to minimize the number of SPI pins, and I am struggling to evaluate if at all we need SC/CS - Chip Select pin if it is always going to be Data from Sensor > MCU. The cost of the solution & reliability of data transmission is crucial in our use case with a maximum distance of up to 3m with the use of STP cable 24AWG with SPIclock of maximum 10Mhz for the signal transmission.

The IIS3DWB datasheet shows in figure 2 that the falling/rising edges of the /CS signals are needed.

You always need to write the register address. And the 3-wire mode uses a bidirectional line and thus cannot be used with RS-422. So you need to use 4-wire mode.

Are the sensors on the same board, so that they can share SPI signals?

Thank you for the feedback, we have 2 MEMS sensors shall be installed on 2 different PCBs which shall be mounted on different parts of a Machine to be monitored for vibration levels.

These 2 MEMS sensors shall be interfaced with 2 different SPI masters on the same MCU via separate SPI lines using 4-wires.

So to confirm before final use, are we missing something, when selecting 1 x AM26C31 (4 Driver Channels) & 1 x AM26C32 (4 Receiver Channels) such that we have:

Master Output (Drivers)
- MOSI
- SCK
- CS/SS 

Master Inputs (Receiver)
- MISO 
- Interrupt 1
- Interrupt 2

This configuration is within our budget & suitable for the board space when compared with DSLVDS047/048 ICs. Also, I often wonder why is there a huge price difference for same IC from TI? For Instance, AM26C31IDR is USD 0.23 while AM26C31IDBR is USD 0.416. Are there any functional differences between these other than apparent package differences?

Hello Clemens,

One final query on this topic, can we use AM26C32IDR drivers with AM26C32IPWR receivers?

Is there anything obvious we may be missing here? Can we interface any AM26C31xxx device with any AM26C32xxx device?

Hello Clemens Ladisch,

I am seeking your guidance on a specific configuration involving the use of AM26C31/32 Drivers for SPI communication in an industrial environment. Our objective is to interface multiple MEMS sensor slaves with a single MCU master over extended cable distances on a single SPI bus.

Could you please confirm the validity of the following configurations:

• Daisy Chain Topology: Connecting the slaves in a sequential manner, with proper termination at both ends, adhering to the EIA-422 standard.
• Star Topology: Connecting the slaves in a star configuration, if possible within the constraints of the EIA-422 standard.

Your insights into the compatibility, potential challenges, and best practices for implementing these topologies with AM26C31/32 Drivers would be highly valuable.

Thank you for your expertise and assistance.