reef-pi :: An opensource reef tank controller based on Raspberry Pi.

[Sral]

Hi Rob, thank you for your answer.
I’ve done some test:
I’m running a board with multiple PCA9507 as I2C extender; if I connect the sensor with 20cm of cable everything works fine.
If I run the full cable (around 2mt) to the same connector the sensor doesn’t work.

I’ve then tested the sensor attached to the outlet expansion board (which is connected to the cpu board by a bus extended by P82B96D): the PCA9685 that drives the output works without a problem, SHT31D connected to the same bus with 20cm of cable doesn’t work…

I was testing everything on a prototype board with very long cable (about 5 meters) and everything was working fine so I tought that with shorter cable I’ll not have any problem… could it be something related to the pullman-up resistor value? On my board I have 2.2K pull-up…
Let me know if tou have any idea, I’ll keep as last chance the use of the ESP32 only to read this sensor…
Thank you!

2.2k actually sounds excessive, depending on the other components in the system that might also have internal pullups as well. I'm also unsure about the "multiple" PCA9507. I don't think you need multiple and they may in fact result in unwanted behavior. Adafruit was able to run an I2C line over 50m Ethernet cable with just ONE LTC4311.

I would recommend starting a dedicated thread and lay out the details of your setup: Power supplies, how they are connected to the PI, your I2C network and how that is powered/level shifted (if at all). That would help greatly in analysing the problem and helping you solve it.

 

[robsworld78]

Hi Rob, thank you for your answer.
I’ve done some test:
I’m running a board with multiple PCA9507 as I2C extender; if I connect the sensor with 20cm of cable everything works fine.
If I run the full cable (around 2mt) to the same connector the sensor doesn’t work.

I’ve then tested the sensor attached to the outlet expansion board (which is connected to the cpu board by a bus extended by P82B96D): the PCA9685 that drives the output works without a problem, SHT31D connected to the same bus with 20cm of cable doesn’t work…

I was testing everything on a prototype board with very long cable (about 5 meters) and everything was working fine so I tought that with shorter cable I’ll not have any problem… could it be something related to the pullman-up resistor value? On my board I have 2.2K pull-up…
Let me know if tou have any idea, I’ll keep as last chance the use of the ESP32 only to read this sensor…
Thank you!

No problem, and yeah something isn't right, the pca9507 is decent, I've used that part in the past and it definitely helps, 2 meters is doable. One thing to note which I forgot to mention is the Pi has 1.8k resistors already installed on the SDA and SCL pins so that may be enough. On my controller I rely on them before the 5v converter then I use 4.7k. With those 1.8k resistors and the 2.2.k you've installed you're down to 0.99k, then if the pca9507 has 2.2k now you have 3 pullups in parallel giving you 0.0022k which could definitely be a problem. As @Sral mentioned it's likely the extender and other devices on I2C have pullups so look for those and remove them if they have them.

 

[Sral]

No problem, and yeah something isn't right, the pca9507 is decent, I've used that part in the past and it definitely helps, 2 meters is doable. One thing to note which I forgot to mention is the Pi has 1.8k resistors already installed on the SDA and SCL pins so that may be enough. On my controller I rely on them before the 5v converter then I use 4.7k. With those 1.8k resistors and the 2.2.k you've installed you're down to 0.99k, then if the pca9507 has 2.2k now you have 3 pullups in parallel giving you 0.0022k which could definitely be a problem. As @Sral mentioned it's likely the extender and other devices on I2C have pullups so look for those and remove them if they have them.

My thinking as well, the 5V Converter might also have another 10k on both sides as well.
My Setup does absolutely fine So far using only internal pull-ups on:
3.3V line:
Pi Zero
Adafruit ISO1540 (isolation and level shifting)

5V line:
Adafruit ISO1540
Adafruit LTC4311 (dynamic pull-up)
RTC module
Adafruit SCD30 Module ( after 1m of shielded USB cable)

 

[Simonv92]

Thank you for all of the answers!
My setup is this one:
Custom PCB with 2.2K Pull-up on 3.3V I2C Bus Line.
On this bus I have:
- 1x P82B96D to extend bus to output board (works fine)
- 1x DS3231 RTC (seems to works fine)
- 4x PCA9507 to have 4 dedicated I2C output at 5V - each outup have his own 2.2K pull up
- 1x ADS1115 ADC (works fine)

Now I've removed the main 2.2K resistor (since raspberry has his own) and I'm trying to connect the SHT31D directly to the 3.3V bus and power.
I'm doing a reading every 30 sec and I'm having some numbers.. at some times the reading is 0 but it's a step forward.
Now I'll try again with PCA9507 as level converter...
@Sral do you have any photo of your setup?
thank you!

 

[Sral]

Have a look on page 6 of the PCA9507's datasheet. It specifies typicall pull-up values of 10k and gives some formulas on what minimum and maximum Pull-Ups to use. They do for example mention that the minimum Pull-Up Resistor value depends on the maximum Pull-down current of your device(s).
The P82B96's datasheet for example mentions a maximum output sink capability on I2C bus of 5.5mA, which equates to a minimum total pull-up resistance to 5V of 0.9k.
The SHT31D's datasheet mentions a sink current of 3mA, at 5V that euqates to a minimum total pull-up value of 1.37k . That's already pretty close to your 2.2k
The DS3231's datasheet mentions the same conditions, 1 to 3mA.

I would therefore caution against such strong pull-ups.

@robsworld78 1.8k on a PI sounds way too little resistance. I have looked it up: here under "voltage specifications" it mentions a pull up between 50 to 65k. That sounds more reasonable.


@Simonv92 You can find a few pictures of my setup in my build thread, like this post for example. I have made a kind of switchboard for the I2C line, which you can see here. The switchboard is the big block of 2x8 female headers, made up of 4 blocks of interconnected 2x2 female headers, one block for each of VCC, GND, SCL and SDA. It's simply there to power the 5V side of my I2C network and provide connection to all 4 "devices", e.g. one line in from the level shifter, one line out to the USB sockets and one line each to the LTC4311 dynamic pull-up and RTC module.
If you scroll a bit down to this post you can see the SCD30 Air Quality sensor that I have connected via a 1m shielded USB cable using USB-A sockets. The only strange thing is that I can't get a connection to the BMP280 air pressure sensor, that's connected identically next to the SCD30.
Downside to this is the ridiculuosly long power line that has way to many connectors with contact resistance, so I'll probably have to be carefull with current hungry devices

 

[Simonv92]

Thank you very much @Sral, I've already seen the PCA9507 datasheet and I've choosen 2.2K based on the informations right there...
All of the devices connected before the PCA9507 works fine (DS3231, P82B96). I've now connected a Ph board to the PCA9597 with 20cm of cable and it works fine (the same thing does SHT31D with 20cm of cable).
Now I'm able to run SHT31D with long cable only attaching it directly to the 3.3V bus which sounds a bit strange, since the PCA should improve the signal quality over long distances...
I don't know what to do... maybe I can try a setup like yours, with a ISO1540 as level shifter from 3.3 to 5V and a LTC4311 to improve signal quality (which I already have but in my setup does nothing...)
What do you think?

 

[Sral]

Thank you very much @Sral, I've already seen the PCA9507 datasheet and I've choosen 2.2K based on the informations right there...
All of the devices connected before the PCA9507 works fine (DS3231, P82B96). I've now connected a Ph board to the PCA9597 with 20cm of cable and it works fine (the same thing does SHT31D with 20cm of cable).
Now I'm able to run SHT31D with long cable only attaching it directly to the 3.3V bus which sounds a bit strange, since the PCA should improve the signal quality over long distances...
I don't know what to do... maybe I can try a setup like yours, with a ISO1540 as level shifter from 3.3 to 5V and a LTC4311 to improve signal quality (which I already have but in my setup does nothing...)
What do you think?

Hmm sounds like the pull up might still be the problem. 3.3V also means less current to sink to properly pull down the signal line, both through the pull-up and the lower charge on the cable capacitance.

I would therefore recommend trying a larger pull-up resistor.

how did you arrive at the 2.2k value ?

 

[robsworld78]

@robsworld78 1.8k on a PI sounds way too little resistance. I have looked it up: here under "voltage specifications" it mentions a pull up between 50 to 65k. That sounds more reasonable.

50k to 65k sounds a little excessive. Never seen anything on I2C bus that high. I'm wondering if those values are for the other GPIO's. I get 1.8k from here and I've seen other specs saying the same.

I2C at Raspberry Pi GPIO Pinout

Raspberry Pi I2C pins

pinout.xyz

 

[Sral]

50k to 65k sounds a little excessive. Never seen anything on I2C bus that high. I'm wondering if those values are for the other GPIO's. I get 1.8k from here and I've seen other specs saying the same.

I2C at Raspberry Pi GPIO Pinout

Raspberry Pi I2C pins

pinout.xyz

Interesting. Well, 50-65k are not realy excessive, they provide some kind of pull-up, but won't overpower devices with very little pull-down capability, so it sounds like a compromise between having at leat A pull-up, while also having compatibility with weak devices. Maybe those 50-60k are the on-chip resistors and 1.8k are external on the board, that could make sense. Something to learn every day !

Thank you very much @Sral, I've already seen the PCA9507 datasheet and I've choosen 2.2K based on the informations right there...
All of the devices connected before the PCA9507 works fine (DS3231, P82B96). I've now connected a Ph board to the PCA9597 with 20cm of cable and it works fine (the same thing does SHT31D with 20cm of cable).
Now I'm able to run SHT31D with long cable only attaching it directly to the 3.3V bus which sounds a bit strange, since the PCA should improve the signal quality over long distances...
I don't know what to do... maybe I can try a setup like yours, with a ISO1540 as level shifter from 3.3 to 5V and a LTC4311 to improve signal quality (which I already have but in my setup does nothing...)
What do you think?

Another thing: could you provide a picture or circuit diagram of the important VCC, GND, SDA, SCL lines from the PI to the PCA9507 ? I just read in the PCA9507's datasheet (first page), that you can't combine several PCA9507's A sides together, because the voltage offsets would interfere with the I2C bus' operation.
Maybe that's making communication beyond the PCA9507 impossible, because as soon as the SHT31D pulls it's side low, the other side might get stuck, since none of the four PCA9507 knows anymore where the communication is comming from and keeps pulling the lines "Low", e.g below 0.5V.

 

[Simonv92]

Interesting. Well, 50-65k are not realy excessive, they provide some kind of pull-up, but won't overpower devices with very little pull-down capability, so it sounds like a compromise between having at leat A pull-up, while also having compatibility with weak devices. Maybe those 50-60k are the on-chip resistors and 1.8k are external on the board, that could make sense. Something to learn every day !


Another thing: could you provide a picture or circuit diagram of the important VCC, GND, SDA, SCL lines from the PI to the PCA9507 ? I just read in the PCA9507's datasheet (first page), that you can't combine several PCA9507's A sides together, because the voltage offsets would interfere with the I2C bus' operation.
Maybe that's making communication beyond the PCA9507 impossible, because as soon as the SHT31D pulls it's side low, the other side might get stuck, since none of the four PCA9507 knows anymore where the communication is comming from and keeps pulling the lines "Low", e.g below 0.5V.

Here is the schematic of the Bus expander (the 4th channel is missing but it's the same)
The main bus is connected on Port B on every expander.
Now I'm running the SHT31D directly on the 3.3V bus without any IC between and it seems to work. I don't understand why the PCA are blocking the signal...
If in this way it works I can remove all of the 9507 and add an ISO1540 as voltage converter and run everything without any bus expander...

 

[Simonv92]

Another question, How do I calibrate a single point calibration probe such as ORP?
If I set 225mV for both mid value and second point I always have 225 reading... I'm missing something

 

[Sral]

Here is the schematic of the Bus expander (the 4th channel is missing but it's the same)
The main bus is connected on Port B on every expander.
Now I'm running the SHT31D directly on the 3.3V bus without any IC between and it seems to work. I don't understand why the PCA are blocking the signal...
If in this way it works I can remove all of the 9507 and add an ISO1540 as voltage converter and run everything without any bus expander...

I realized something, that I didn’t notice before: you have several components that provide I2C extension. Maybe I should have read more carefully or you could have stressed this a bit more.
Could you provide a sketch of your I2C network, e.g. which components translate this network from and to which components ?

I’m worried that the P82B96 might not play well with the PCA9507. It says something like this in the datasheet:
„slightly different logic low-voltage levels are used at Sx/Sy to avoid latching of this buffer. A standard I2C low applied at the Rx/Ry of a P82B96 is propagated to Sx/Sy as a buffered low with a slightly higher voltage level.“
On the PCA9507:
„The static level offset design of the port B I/O drivers prevent them from being connected to another device that has rise time accelerator“

So if both use Offsets and accelerators, they might latch up. You can test this by connecting the SHT31D in the original not working condition and measuring the DC voltage on the lines. If they are too low (e.g. below 1V) it probably means that they are latched up.
Another fault might be, that the voltage offset of LOW state on the PCA9507‘s B-side might not be properly recognized by the P82B96, or vice versa.

 

[Simonv92]

I realized something, that I didn’t notice before: you have several components that provide I2C extension. Maybe I should have read more carefully or you could have stressed this a bit more.
Could you provide a sketch of your I2C network, e.g. which components translate this network from and to which components ?

I’m worried that the P82B96 might not play well with the PCA9507. It says something like this in the datasheet:
„slightly different logic low-voltage levels are used at Sx/Sy to avoid latching of this buffer. A standard I2C low applied at the Rx/Ry of a P82B96 is propagated to Sx/Sy as a buffered low with a slightly higher voltage level.“
On the PCA9507:
„The static level offset design of the port B I/O drivers prevent them from being connected to another device that has rise time accelerator“

So if both use Offsets and accelerators, they might latch up. You can test this by connecting the SHT31D in the original not working condition and measuring the DC voltage on the lines. If they are too low (e.g. below 1V) it probably means that they are latched up.
Another fault might be, that the voltage offset of LOW state on the PCA9507‘s B-side might not be properly recognized by the P82B96, or vice versa.

Thank you @Sral for all your help!
I2C diagram is this one:

Raspberry Pi Zero W 2 -> 3.3V Bus
3.3V Bus -> P82B96 to expansions board (equipped with another P82B96 to decode bus)
3.3V Bus -> DS3231 RTC
3.3V Bus -> ADS1115 ADC
3.3V Bus -> PCA9507 -> 5.0V Bus 1 to external devices -> 4 way terminal block
3.3V Bus -> PCA9507 -> 5.0V Bus 2 to external devices -> 4 way terminal block
3.3V Bus -> PCA9507 -> 5.0V Bus 3 to external devices -> 4 way terminal block
3.3V Bus -> PCA9507 -> 5.0V Bus 4 to external devices -> 4 way terminal block

Let me know if it's clear otherwise I'll draw it graphically..

 

[Sral]

Thank you @Sral for all your help!
I2C diagram is this one:

Raspberry Pi Zero W 2 -> 3.3V Bus
3.3V Bus -> P82B96 to expansions board (equipped with another P82B96 to decode bus)
3.3V Bus -> DS3231 RTC
3.3V Bus -> ADS1115 ADC
3.3V Bus -> PCA9507 -> 5.0V Bus 1 to external devices -> 4 way terminal block
3.3V Bus -> PCA9507 -> 5.0V Bus 2 to external devices -> 4 way terminal block
3.3V Bus -> PCA9507 -> 5.0V Bus 3 to external devices -> 4 way terminal block
3.3V Bus -> PCA9507 -> 5.0V Bus 4 to external devices -> 4 way terminal block

Let me know if it's clear otherwise I'll draw it graphically..

That's clear enough, thanks.

Hmmmm, the datasheets are not clear enough for me to judge the compatibility of the P82B96 and PCA9507 in your situation with certainty. Maybe you can test it out, e.g. connect the SHT31D to the PC9507 (test several, in case the one used is broken) , remove or deactivate the P82B96 from the network and see if the connection works then.

Another thing I noticed on the PCA9507: the datasheet only mentions standard topographies of multiple A-Port together in a star-topography, or A and B ports in series. So maybe they don't play well with each other, when you connect multiple B-ports to each other ? To test that you would of course have to remove or deactivate 3 of the 4 PCA9507's and then connect the SHT31D to the remaining one.

I hope that is even possible for you and you can find a solution.

 

[robsworld78]

Interesting. Well, 50-65k are not realy excessive, they provide some kind of pull-up, but won't overpower devices with very little pull-down capability, so it sounds like a compromise between having at leat A pull-up, while also having compatibility with weak devices. Maybe those 50-60k are the on-chip resistors and 1.8k are external on the board, that could make sense. Something to learn every day !

Yeah maybe 50+k is better than 1.8k for I2C, that's above my pay grade. I've seen others complain about them being so low as well so still thinking they are 1.8k but can't be certain now, thanks. I searched that page you linked which is official and no mention of 1.8k or any other values.

 

[robsworld78]

Thank you @Sral for all your help!
I2C diagram is this one:

Raspberry Pi Zero W 2 -> 3.3V Bus
3.3V Bus -> P82B96 to expansions board (equipped with another P82B96 to decode bus)
3.3V Bus -> DS3231 RTC
3.3V Bus -> ADS1115 ADC
3.3V Bus -> PCA9507 -> 5.0V Bus 1 to external devices -> 4 way terminal block
3.3V Bus -> PCA9507 -> 5.0V Bus 2 to external devices -> 4 way terminal block
3.3V Bus -> PCA9507 -> 5.0V Bus 3 to external devices -> 4 way terminal block
3.3V Bus -> PCA9507 -> 5.0V Bus 4 to external devices -> 4 way terminal block

Let me know if it's clear otherwise I'll draw it graphically..

I'm not trying to bring you down but It looks like you're working on a fairly large project, if you're planning to run 18 I2C devices you might need to rethink things. A few years ago I wanted my controller to expand to large levels and went down the road you are heading on and for me it didn't end well. I attempted to make it work for months, even went to the expense of getting lots of PCB's made. I tried a few buffers, the pca9507 was the best. Running multiple pca9507 like you are I was able to successfully run 22-24 devices with about 100 feet USB cable total but it wasn't reliable. Even with only a few devices connected it wasn't perfect. I hate to say it and I hated to read it when I was doing this, it's just not possible if you need reliability. For any chance of it working you would have to code a strong CRC check and figure out how to handle the bus locking up.

Bottom line is you need to look at other protocols such as RS-485 or CAN bus. If you really want to keep going look at the PCA9615 to split the I2C signals into differential signals, personally I haven't used this as I finally came to accept this isn't the route a person should be taking but this does sound like the best bet.

Qwiic Differential I2C Bus Extender (PCA9615) Hookup Guide - SparkFun Learn

learn.sparkfun.com

Here's a good article explaining how differential signals works.

An Introduction To Differential I²C

A few weeks back, we talked about the no-nos of running I²C over long wires. For prototyping? Yes! But for a bulletproof production environment, this practice just won’t make the cut. This mo…

hackaday.com

Here's a pic with most of the stuff connected, I had 8 AC power bars connected as well at one point. All that and more went in the trash.

 

[Simonv92]

That's clear enough, thanks.

Hmmmm, the datasheets are not clear enough for me to judge the compatibility of the P82B96 and PCA9507 in your situation with certainty. Maybe you can test it out, e.g. connect the SHT31D to the PC9507 (test several, in case the one used is broken) , remove or deactivate the P82B96 from the network and see if the connection works then.

Another thing I noticed on the PCA9507: the datasheet only mentions standard topographies of multiple A-Port together in a star-topography, or A and B ports in series. So maybe they don't play well with each other, when you connect multiple B-ports to each other ? To test that you would of course have to remove or deactivate 3 of the 4 PCA9507's and then connect the SHT31D to the remaining one.

I hope that is even possible for you and you can find a solution.

Thank you again @Sral, I'll do some more test and I'll try to find out a reliable way to make everything works...

 

[Simonv92]

I'm not trying to bring you down but It looks like you're working on a fairly large project, if you're planning to run 18 I2C devices you might need to rethink things. A few years ago I wanted my controller to expand to large levels and went down the road you are heading on and for me it didn't end well. I attempted to make it work for months, even went to the expense of getting lots of PCB's made. I tried a few buffers, the pca9507 was the best. Running multiple pca9507 like you are I was able to successfully run 22-24 devices with about 100 feet USB cable total but it wasn't reliable. Even with only a few devices connected it wasn't perfect. I hate to say it and I hated to read it when I was doing this, it's just not possible if you need reliability. For any chance of it working you would have to code a strong CRC check and figure out how to handle the bus locking up.

Bottom line is you need to look at other protocols such as RS-485 or CAN bus. If you really want to keep going look at the PCA9615 to split the I2C signals into differential signals, personally I haven't used this as I finally came to accept this isn't the route a person should be taking but this does sound like the best bet.

Qwiic Differential I2C Bus Extender (PCA9615) Hookup Guide - SparkFun Learn

learn.sparkfun.com

Here's a good article explaining how differential signals works.

An Introduction To Differential I²C

A few weeks back, we talked about the no-nos of running I²C over long wires. For prototyping? Yes! But for a bulletproof production environment, this practice just won’t make the cut. This mo…

hackaday.com

Here's a pic with most of the stuff connected, I had 8 AC power bars connected as well at one point. All that and more went in the trash.

Hi Rob,
thank you for your message, yes I was following your old project so I understand what you're saying... It was a very complex project but, in my opinion, it had one of the best looking GUI I've ever seen in a reef controller (and not only that)...
Regarding my project.. Is not that complicated, I think I've complicated the things by myself...
My goal is to run 3 separate PCBs in 1 single enclosure (to minimize cable run an footprint) the boards are these:
CPU Board - with Raspberry Pi Zero, and all the circuit for input side, 0-10V out from Rpi, DS18B20, ADC and I2C
Powerbar - connected to the previous board by a CAT5 0.15cm cable. It runs 8 AC outlet and 8 DC outlet by a PCA9685.
Light/Dosing - to design yet, it'll have another PCA9685 connected by the same CAT5 cable and it'll control 3/4 dosing pump and 3/4 lights channel.

I was trying to keep I2C signal very short since I know that long cable can cause many troubles.
Attached to the CPU board (but in the same box) I'll have 2 Ph Board (the one made by Roberto Buti) and the SHT31D sensor which is the only device connected by a long cable to reach my sump and measure temp/humidity to control an exhaust fan.

I was also thinking about changing the SHT31D to a DHT22 (as you do in your new software if I rember correctly) to be able to have long cable without problem, but I don't know how to add it to Rpi...

 

[Ranjib]

I'm not trying to bring you down but It looks like you're working on a fairly large project, if you're planning to run 18 I2C devices you might need to rethink things. A few years ago I wanted my controller to expand to large levels and went down the road you are heading on and for me it didn't end well. I attempted to make it work for months, even went to the expense of getting lots of PCB's made. I tried a few buffers, the pca9507 was the best. Running multiple pca9507 like you are I was able to successfully run 22-24 devices with about 100 feet USB cable total but it wasn't reliable. Even with only a few devices connected it wasn't perfect. I hate to say it and I hated to read it when I was doing this, it's just not possible if you need reliability. For any chance of it working you would have to code a strong CRC check and figure out how to handle the bus locking up.

Bottom line is you need to look at other protocols such as RS-485 or CAN bus. If you really want to keep going look at the PCA9615 to split the I2C signals into differential signals, personally I haven't used this as I finally came to accept this isn't the route a person should be taking but this does sound like the best bet.

Qwiic Differential I2C Bus Extender (PCA9615) Hookup Guide - SparkFun Learn

learn.sparkfun.com

Here's a good article explaining how differential signals works.

An Introduction To Differential I²C

A few weeks back, we talked about the no-nos of running I²C over long wires. For prototyping? Yes! But for a bulletproof production environment, this practice just won’t make the cut. This mo…

hackaday.com

Here's a pic with most of the stuff connected, I had 8 AC power bars connected as well at one point. All that and more went in the trash.

yes, at that scale go plc/scada for industrial setup, or home lab type setup with x96 and cat5/6.


thank you for sharing. this is a fun read

 

[Ranjib]

Some pics from the current tanks

I love palys . I think they look super contrasting , the palette with zoa colors pop .

Got some time to do some more work on the new pico stand. This one shows my last two years of wood working maturity . The entire thing is built with 3/4” walnut vineered plywood , with domino as joinery.

Doors on two sides and the front pane will allow hooking up common control such as light mode, disable ato etc. the color choices (white and gold ) is to keep it in tune with rest of home furniture (wify dictates those ) . Hardwares are all from

Hopefully this one will blend in well and also make things easier to operate .