SPI with a level converter in the other end of connection

Question

I have an Arduino MEGA hooked up with a NRF24L01, and working nicely, talking with another Arduino Uno. It works as described, power supply to 3.3V, and SPI lines (MISO,MOSI,CLK) connected directly to the Arduino;

However, when I hook up a level converter in the end (5V to 3V, bidirectional) (making the NRF module in the middle, like in a T connection), the module stops working, and I don't know why. How does the level converter module is interfering in the connection? Is there any way I can prevent this and use it?

I intend in adding in the future, modules that talk in 3.3V SPI;

This is the level converter I'm using:

_{(source: at-the.space)}

I suspect it is because of the pullups (5 think that these 10kΩ resistors are pull-ups, but I'm a noob in that part), but I really don't know if that is the problem, as the seller advertises it as being compatible with SPI, I2C, and UART. I have successfully used these for a UART.

I have checked Slaving multiple Arduinos via SPI and tried using an external 5V supply, and a 3.3V supply, but no dice.. However, the 3V side has no connections to it; But, being bi-directional, I don't think there is a problem with this...

I found that 74HCT245D works very well for this, but I'd like to know why with the level converter from JY-MCU this won't work.

Providing the library and connections just in case (the module works without the level converter):

• Library & sketch comes from Optimized High Speed NRF24L01+ Driver Class Documenation.

• Connection:

  • NRF2401 Pinout: NRF24l01 http://ralphfox.at-the.space/imglink/nrf24l01.jpg

  • My connection: Arduino with NRF24l01 and Level Converter http://ralphfox.at-the.space/imglink/nrf_bb.png

Answer 1

More than probably your level shifter is not fast enough. A level shifter uses a mosfet to copy signals from one voltage level to the other, but each mosfet has a latency or, if you prefer, a rise time. Your board probably doesn't use a fast switching mosfet and so the SPI protocol (which is a quite fast one) is not copied correctly.

On top of that you actually don't need a level switcher for the nRF24 modules: the power pin must be limited to 3.3V, but the signal pins are 5V tolerant. This means you only need a 3.3V voltage source and all the remaining connections can go directly (and safely) into your Arduino board.

UPDATE: the speed concept still applies, it doesn't matter which side of the connection you are using. A mosfet has a little capacitance which slows down the signal transmission: higher the capacitance, slower the maximum data rate. The cheap chinese level shifter is not designed for high speed communication, while the Texas Instrument device is clearly designed for high speed. If you look at the SN74HCT245 datasheet you'll notice the switching speed is reported for 50pF capacitance and ranges from 8ns to 16ns: that's the time needed to build enough charge in the switching circuit.

Answer 2

AVR's like all digital have their limits. In this case you're dealing with source resistances that need to charge multiple gate capacitances through series resistances (wires and metal contact connections. Specifically an Absolute MAX 40mA/pin, some milliohms in your copper wire and PCB traces, and pico to nano Farad capacitances, plus depending on how High Speed the SPI is (the spec does not list an upper limit), you will be also contending with frequency limiting parasitics. At MHz frequencies, your wire and PCB traces will have higher resistances due to "skin" effect, where the electrons migrate to the surface of the wire when transferring. Depending on frequency of the wire, adding just a single gate could degredate rise/fall times enough to corrupt data, as the SPI library (whichever you choose) and the SPI peripheral within the AVR probably won't be as sophisticated as a PC based one. The AVR & the SPI library lacks sophistication in that neither has provisions to detect and compensate for extra gate capacitance on the bus line, nor could they compensate for china jumper wires which are quite small (9x40ga) wiring. There are many factors which could corrupt your SPI project that are coming into play to limit performance. You should probably do some internet research on high speed bus anatomy if you intend on pushing the SPI bus on the arduino to it's limits.

I think the gate capacitances are combining with source resistance and skin effect and your 40mA/pin is not in fact 40mA per pin but more like 10-20mA is hitting all gate capacitances when skin effect is considered (it effectively raises the resistance of the wire sufficiently to then be a factor). If connecting the JY-MCU converter AND running SPI at minimum speed corrects communication, then the extra gate capacitances are the factor you should change to higher performance versions. There's the 74AS series with gate times down to 1.7nS and fMAX up to 40MHz!

Hope this helps.