CAI Webcontrol Temp Sensors

[apostolakisl]

I am trying to connect a remote thermometer to my CAI Webcontrol. I purchased several of the recommended DS1822 sensors. Hooked directly to the board, they work fine. But I would like to locate one about 40 feet away. I connected it with CAT5, rebooted, and the sensor doesn't show up in the menu. In fact, it caused the sensor still directly connected to the board to disappear from the menu as well. I ohmed out the wires and all is good. I removed the sensor directly connected to the board, and still nothing. I replaced a single sensor directly to the baord and it works again (the same sensor that wouldn't work at the end of the wire).

So, in short.

1) All sensors work when hooked directly to the board.
2) Neither sensor works at the end of 40 feet of wire.
3) When a sensor is attached at the end of 40 feet of wire, it also makes the sensor directly attached to the board vanish.
4) I used CAT5e wire and the individual strands ohmed out fine.
5) And yes, I double and triple checked that I had the right wires connected to the right things.
6) I did reboot the unit each time I changed something.

I am assuming that the CAI doesn't have enough juice to drive current down the wire to the sensor?

I am hoping someone knows what I need to do to get a temp sensor sending data from 40 feet away.

 

[AnthonyZ]

While I'm not sayin' that the cat5e is the issue (to my knowledge, many people use it), I have run all my OneWire over 18/2 shielded cable and have never had a problem.

 

[apostolakisl]

Thanks Anthony.

Do you power your stuff locally then if you are only running 2 strands?

CAI did respond to my email and said I should power the sensor locally from a 5v supply and only run the signal/ground to the cai unit.

 

[AnthonyZ]

Thanks Anthony.

Do you power your stuff locally then if you are only running 2 strands?

CAI did respond to my email and said I should power the sensor locally from a 5v supply and only run the signal/ground to the cai unit.

Thus far, I am running on parasitic power (via a Barionet 100). However, I only have three DS18B20s at the moment. Two weeks ago my wife bought me four more (Oh, how I love that woman!). Implementing those may make me rethink my current setup.

 

[fwd03]

I think it depends on the environment. If you are near high power source, it is better to have shield cable, since the long wire will act like antenna picking up noises ( like cell phone signals, TV signals, Saterlite signals) along the line. I think Anthony's shield cable is a wonderful idea - make sure the shield is grounded.

 

[Michael McSharry]

The shield also acts like a capacitor and that makes the digital waveform look pretty ugly and hard for the electronics to decode in the timeslots available. Typically the twisted pair is more effective for 1-wire applications. The CAI design cannot handle parasitic power and as they indicated the local power can only handle very local sensors. This is true even when the higer humidity sensor voltage is used as the power source. I was not aware the current generation CAI was able to use the DS1822. I have only had success with a limited selection of DS18B20. What I like for the CAI is the thermometer probe off of ebay from overseas. It is well encapsulated to handle many different install applications and the price was not that much more than the through-hole package version of the sensor from maxim-ic. The ones I purchased awhile ago look to be the same as http://cgi.ebay.com/3m-Digital-Ther...158?pt=LH_DefaultDomain_0&hash=item336945f3be

 

[Sooty]

I also recently acquired a webcontrol. DS18B20's work OK for me with a separate 5 volt supply, I've tried both direct from Maxim ones, and ebay 10 for $10 from China ones.

 

[apostolakisl]

I also recently acquired a webcontrol. DS18B20's work OK for me with a separate 5 volt supply, I've tried both direct from Maxim ones, and ebay 10 for $10 from China ones.

How close are the temp sensors to the unit (or in other words, how much wire between it and the cai)?

 

[fwd03]

For parasite temperature sensors to work, the controller must pull bus high all the time during temperature conversion. For 12bit digital sensor, that is at least 750ms -- very close to 1s. That means the 1-wire bus will not be able to talk to anybody else on the bus during that time period. There is no way to bypass that, since that bus providing power for the parasite part to do A/D conversion. For 8 sensors, you will have 8 seconds to update all the temperature if you lock up the bus because parasite power requirement. If you do not use parasite sensor on the bus, then the bus will be free up to talk to other devices.

 

[apostolakisl]

For parasite temperature sensors to work, the controller must pull bus high all the time during temperature conversion. For 12bit digital sensor, that is at least 750ms -- very close to 1s. That means the 1-wire bus will not be able to talk to anybody else on the bus during that time period. There is no way to bypass that, since that bus providing power for the parasite part to do A/D conversion. For 8 sensors, you will have 8 seconds to update all the temperature if you lock up the bus because parasite power requirement. If you do not use parasite sensor on the bus, then the bus will be free up to talk to other devices.

This is the response I got from cai support last week when I asked a similar question.

"The only other I/O can slow down the processing is the temp sensors. Without temp sensor, PLC can actually get to very close to 10 mS or less resolution. With temp sensor, it will be in 100mS or less resolution. If the input pulse is narrow than 100 mS, then you may consider not install any temp sensor."

There was no mention of using a separate power supply to avoid the slow down.

If you use a separate power supply, how good does it need to be? Will a 5v wallwart be fine or do you need a voltage regulator?

 

[sda]

This is the response I got from cai support last week when I asked a similar question.

"The only other I/O can slow down the processing is the temp sensors. Without temp sensor, PLC can actually get to very close to 10 mS or less resolution. With temp sensor, it will be in 100mS or less resolution. If the input pulse is narrow than 100 mS, then you may consider not install any temp sensor."

There was no mention of using a separate power supply to avoid the slow down.

If you use a separate power supply, how good does it need to be? Will a 5v wallwart be fine or do you need a voltage regulator?

These comments are for 1-wire in general and not specifically for CAI Webcontrol since I have no experience with it.

I think a 5V supply increases sensor reliability and puts less demands on the 1-wire data bus and controlling hardware.
As far as the actual power supply, I use a powered hobby boards hub which has a 5V voltage regulator internally. The spec (DS18B20, DS18S20, DS1822) allows VDD to be 3.0 to 5.5V. I wouldn't use a wall wart unless it was regulated or run it through your own 7805.

It takes about 750ms for the sensors to do a conversion (12 bit resolution) whether they are parasite powered or have +5V power.

My experience has been that when the sensors are parasite powered, you can only poll one sensor at a time (because the sensor uses bus power when doing the conversion). Send "convert t", wait 750ms, read sensor. If you have 10 sensors, it will take you 7.5 seconds to read them all.

If you have 5V and are polling the sensors individually, it will will still take 750ms per sensor. Although custom software can take advantage of a trick described in the spec sheet to stop waiting when the conversion is completed.

If you have 5V and the software is written to take advantage of it, you can send "skip rom" followed by "convert t" and all sensors will do the conversion at the same time, then you can read the results from each sensor individually. If you have 10 sensors, it will take you 0.75 seconds to read them all. I haven't seen where the spec sheet doesn't specifically say you can't use "skip rom" with parasite power, but I think I tried it, ran into problems and switched everything to 5V powered. I never went back and did a controlled "skip rom" test with parasite power.

BTW, the DS1822s have lousy accuracy, +-2*C.

 

[apostolakisl]

thanks for the explaination sda.

So, are you saying that the cai device (or other similar device) polls the temps one at a time and hangs for 750ms waiting for the result to resolve?

The cai support says it will slow the cycle through polling of all the sensors to 100ms if the temp sensor is in place versus 10ms if no temp sensor is in place. Some trick perhaps that lets it not hang for 750ms?

I have a device that will be inputting data to the digital input at about 5 hertz pulsed 5v. If I'm dropping pulses I will know why.

Had no idea about the temp sensors. I just read that it supported the two types and typed the first number into ebay and there it was so I bought it. It is plugged direct into the cai and the temp it reports is consistent with what it feels like (sadly that number is presently 108.1 f), but "what it feels like" isn't exactly the most accurate. I am pretty sure I can tell 4 degrees f off, but it could easily be 1 or 2 degrees off and my skin would not know it.

 

[Sooty]

How close are the temp sensors to the unit (or in other words, how much wire between it and the cai)?

I tested connected to the terminal block - works but a bad idea since (without forced air cooling) the board gets hot and that heat is conducted to the 18B20 which will read several degrees higher than ambient, about 12" of hand twisted triple, and about 12 feet of cat 5.

 

[sda]

So, are you saying that the cai device (or other similar device) polls the temps one at a time and hangs for 750ms waiting for the result to resolve?

That would be the simplest way to implement it.

Knowing how the sensors work can be useful. The "convert t" command tells the sensor to do the temperature conversion and it stores the result in its internal register. The control device doesn't need to monitor the conversion and can do other stuff besides sit and spin. 750ms later, the new temperature sample is available. The control device can read that register immediately after the conversion or at any time in the future (until the bus is reset).

Ideally the control device would have a processing loop, set a flag when the temp conversion command is sent to the sensor, continue doing other stuff, then when 750ms came around the sensor would be read and its data utilized. That requires fairly tight coupling with the application, which may not be possible with a generic controller.

Only the control device designers can tell you what they actually implemented.

The cai support says it will slow the cycle through polling of all the sensors to 100ms if the temp sensor is in place versus 10ms if no temp sensor is in place. Some trick perhaps that lets it not hang for 750ms?

I wonder if they are setting the sensors to 9 bit resolution (1/2 degree C) instead of 12 bit resolution (1/16 degree C). That reduces the conversion time from 750ms max to 93.75ms max.

If you're curious, write a PLC program to sample the multiple temp sensors and toggle an output connected to an LED before the start of each loop. If the LED blinks at 1/10th second intervals they are doing a simultaneous temperature conversions ("skip rom"). If the LED blinks at longer intervals corresponding to the number of temp sensors, they are doing individual sensor conversions. It would be useful to know that fact if you're planning on adding additional temperature sensors.

Had no idea about the temp sensors. I just read that it supported the two types and typed the first number into ebay and there it was so I bought it. It is plugged direct into the cai and the temp it reports is consistent with what it feels like (sadly that number is presently 108.1 f), but "what it feels like" isn't exactly the most accurate. I am pretty sure I can tell 4 degrees f off, but it could easily be 1 or 2 degrees off and my skin would not know it.

The original sensor is the DS1820. There were some manufacturing problems with that one.

The DS18S20 was the replacement. It has an accuracy of +-0.5*C. 9 bit direct resolution, 12 bit resolution if you do some calculations. I use this sensor and have good results.

The DS18B20 is a full featured update of the DS1820. It uses a different data format and isn't directly compatible with the DS1820 and DS18S20. It has an accuracy of +-0.5*C.

The DS1822 is an economy mode sensor and is directly compatible with the DS18B20. Its description is "Econo 1-wire Digital Thermometer", thus the +-2*C accuracy.

 

[sda]

This came from the CAI webcontrol FAQ:
http://www.redhillnetworks.com/products/webcontrol/webcontrol-faq.html

3-1A: WebControl™ can simutanously connect 8 optional DS18B20/DS1822 based digital temperature sensors, as long as the temp sensor is 12bit and non-parasite part. DS18S20 is a 9bit part that only having temperature resolution > 0.5C. DS1822 and DS18B20 are 12 bit temperature sensors. They should work fine. WebControl does not support parasite parts.

So it doesn't support parasite powered sensors. Per the docs, it looks like the temp sensors are hooked up to a 3 position connector that supplies 3.3V. Maybe try using the 5V output from the humidity sensor connector. You may be getting too much voltage drop on the cable at 3.3V.