Solar Thermal Pump Controller requiring EMC Test Pass

[ed1066]

I started a Thread on 9 Apr 2013 “EMC Testing” before attempting to get my control circuit based on a 20X2 through the UK EMC tests. People gave me helpful time and advice and I’m reporting back here after completion of the testing. This is the same material as in the main Forum thread:

http://www.picaxeforum.co.uk/showthread.php?25020-EMC-Testing-%96-my-experience

The April 2013 thread gives more background and is

http://www.picaxeforum.co.uk/showthread.php?23667-EMC-Testing

I am keen for people to benefit in particular from the EMC aspects of my experience and it was suggested that I re-post on this Forum.

The controller is for controlling the 12V pump in a solar thermal system. After 4 half-day visits to my test facility, I finally had a pass. There were three areas requiring work:

• Use of pwm
• Cables to DS18B20 temperature probes and separate OLED/LCD display
• Vulnerability to electric shocks in supply and signal leads

Firstly - The Original Circuit

Here is my original circuit, somewhat simplified (eg no download cable, non-relevant input removed and only one of three DS18B20 sensors shown).



Some notes:

The control circuit PCB is in a plastic box.

The pump has a brushless motor, it does not produce back emf. Following the April Forum discussion I removed its motor capacitor. The lead from the control box to the motor is a 20cm length of simple two-core speaker cable.

About 4m of the same speaker cable powers the system from a 12V switched mode power supply.

There are three DS18B20 temperature sensors, on 2m – 6m or so leads, simple 4 core ‘alarm cable’ with one core not connected to anything (3 core not available). One sensor shares a lead with the OLED for some of the way. One capacitor covered all the sensors (the OLED has its own)

The 27 ohm resistor covered power to all the temperature sensors & OLED and was intended to protect the circuit if a short occurs in a lead or connection.

The OLED is the AXE133Y, its lead is 3 – 5m long and its in a plastic box.

The frequency is reduced to 4 MHz which is that used by the ReadTemp command. As I need ReadTemp that seemed to be the lowest frequency possible.



Emissions and pwm

The first tests showed the motor was associated with noisy radiated emissions above the permitted limit.

· The problem only arose when the power to the pump was being reduced using pwmout.

· Using a linear power supply rather than a switched mode power supply actually made it worse

· Adding ferrites on the power line to the pump did not help enough

· Conducted Emissions were not a problem

The problem was solved by adding a 1K resistor to the connection from the output pin to the FET. I tried 2.2K as well and 1K was sufficient to reduce noise with only a slight increase in temperature (0.5 deg C) recorded on the FET heat sink, and a slight change in total power consumed. The 2.2K caused a higher temperature rise and a higher power consumption.

See the updated circuit diagram with this and the other changes.



Common Mode (conducted) RF Immunity - Cables to DS18B20 probes and OLED

Any cables longer than 3m needed testing, which confirmed there were problems with both the leads to the temperature probes and the OLED. Noise added to the leads with a heavy coil thingy knocked out the signals.

The book recommended to me by this Forum back in April was helpful. It’s by Tim Williams: “EMC for Product Designers”. The Fig 13.16 on p358, for example, suggests that capacitors are needed due to the high impedance at each end of the signal cables.

I did tests with a piezo gas lighter connected to a circuit to produce transient currents – see next post. It showed the vulnerability was reduced by a combination of resistors on the signal and capacitors to ground. The capacitor should be on the cable side, the resistor nearer the pin on the 20X2. The values for these components were eventually derived using trial and error (hence several visits to the EMC test facility). For the temperature probes with more than 3m cables a filter is needed at the PCB end only. The resistor is 330ohms and the capacitor 2.2nF.

For the OLED signal cable a filter is needed at the OLED end only. The resistor is 1K and the capacitor 10nF. The 5V supply also needed a 12 ohm resistor at the OLED end, to prevent loss of voltage I reduced the 27 ohm resistor beside the voltage regulator to 12 ohm.

I tried twisted wire pairs instead of simple alarm cable, but it was marginally worse with twisted pairs.

I added a code signal, the number 199, to the OLED prior to any other signal. If serin in the code controlling the OLED read a byte variable equal to 199 then it went on to read the data to display. So if the signal was corrupted, the display didn’t update rather than displaying random spurious messages.

The system only just passed these tests, I was told that to get better results the only option would be shielded cable and enclosures etc. For the final circuit, the vulnerability was greatest at the high frequency end, eg 30 to 80 MHz.


Susceptibility to Voltage Shocks

In the EMC tests, this is the items ‘Fast Transient Immunity’ and ‘Surges’. The shocks were applied to the 240V supply to the switched mode power supply and to the signal cables (to DS18B20 probes and OLED)

The shocks knocked out the controller and it restarted after the test. This was a pass, but it showed vulnerability and I was strongly advised to address it.

Firstly, I worked on the circuit layout on my PCB, based on the Tim Williams book. I made it into 3 zones: (a) 12V supply, the voltage regulator and pump power control, (b) interface for temperature & OLED leads and (c) the inner protected circuit around the 20X2 chip. Each zone had its solid ground plane on the underside of the PCB and they were connected together at one central point only. All the signals & 12V/5V lines were on the upper plane of the PCB.

I then added a 220ohm resistor in the 5V supply to the 20X2 chip. Tests with the piezo gas lighter showed the vulnerability was resolved without problems in voltage to the chip.

 

[ed1066]

There is a photo limit of 2. I've added two here, one is the gas lighter device I mentioned and the other shows the OLED display. Its the lower of the 2 display boxes.

 

[BeanieBots]

Thanks for sharing.
So many people think you can just knock up a circuit and put it out there. There are so many regulations these days!
I'm doing a similar project myself and have the added 'issue' of controlling the power to a 3kW immersion. Rather than use phase angle control which would be almost impossible to suppress, I opted for burst fire which emmits almost nothing.

I love your idea of using a peizo lighter for testing. I'll certainly give that one a go.
On the subject of DS18B20, there is a device which offers bi-directional suppression. I enquired here about that very problem many years and Technical gave details about the device but I cannot find the thread now.

As you have found and as warning to others, beware the OLED!
I think it must be the on-board negative rail generator but they put out bucket loads of crud. So much so that any analogue devices need very careful attention.

Switchmodes are well known for line to output transition problems. It can be easily fixed with ferrites (or even a decent inductor) and caps between output and supply to circuit. (keep close to PSU).

Anyway, congratulations and thanks.

 

[ed1066]

Thanks BeanieBots, appreciated

I'm doing a similar project myself and have the added 'issue' of controlling the power to a 3kW immersion. Rather than use phase angle control which would be almost impossible to suppress, I opted for burst fire which emmits almost nothing.

I am also developing a system to control a 3kW immersion when there is a solar heat input. It's a great challenge. I have not considered EMC while developing and testing ideas. I use a large relay rated to 50A (3kW is about 13A at 230V) :

http://cpc.farnell.com/jsp/search/productdetail.jsp?SKU=SW04023

It gives out a nice flash when it switches - I imagine it will fail EMC without shielding, but then a 13A power socket switch presumably passes on emmissions somehow. The other night I switched off my PC at the socket in a dark room and saw a tiny flash through the gap between the switch and the socket housing. The PC was shut down so the power draw would be I guess 20W-30W only.

I found the immersion switching produced transients which caused the PICAXE chip to restart, but the measures I discussed above solved it easily enough.... and it's still on a stripboard with no ground plane.

I would like to hear more on your method. I don't even know what the terms 'phase angle control' and 'burst fire' mean.

As you have found and as warning to others, beware the OLED!
I think it must be the on-board negative rail generator but they put out bucket loads of crud. So much so that any analogue devices need very careful attention.

The only OLED problem for my EMC testing was its vulnerability to corruption in the incoming signals. There were no emmissions problems. Again you are further on than me and I don't understand 'rail generator' etc, so I'm not sure if we are saying the same thing or opposite things.

Cheers, Ed

 

[BeanieBots]

From your excellent writeup I got the impression you did this for a living!
For my sins, I've been earning money from electronic design for many decades now so I've lived through "whatever works" up to the current "nanny state" regarding design and legislation.

There's another term you will need to know and understand. "Zero Crossing".

Zero Crossing is what it implies, when the mains voltage crosses through zero volts.
This is the best time to switch things on and off because no current flows and hence there is no surge, spike, flyback or other nasties to cause arcing and/or EMC.

Yopu can get readymade solid-state relays that will do it for you or you can make your own using thyristor and a couple of trannies.

Phase-angle fire is when you switch on when the voltage is at a particular phase in its 50/60 Hz cycle.
This is how most lamp dimmer switches work. It's an easy circuit to implement using just a pot an RC and a tthyristor because the RC (variable via the pot) produces the different phase angle. Where in the pahse you switch on determines how much power from each cycle you get and hence how bright the lamp is. As far as EMC is concerned, it the absolute pits because at about 50% power, it will be switching at peak voltage and producing maximum possible radiation.

Burst-fire is when you switch the device on/off for a fixed number of mains cycles. Zero-crossing is used to syncronise the switching with mains so that every on/off happens at zero volts. It is very effective for heater control but cannot be used for lighting because the relatively long period of several full mains cycles introduces flicker.

As I'm sure you are aware, EMC regulations cover both immunity and radiation.
A switch inherently passes but the device it switches might not. In particular, large inductive loads might need to have suppressors fitted. As already mentioned, switching power supplies are notorious for passing mains spikes out onto the low voltage side. If your micro design fails when say a domestic vacuum cleaner is switched on/off, then it is failing EMC immunity. (it might be the vacuum cleaner putting out too much but unlikely as it will have been certified).

It's a huge topic and the solutions are varied but there are many best known practices which will solve most problems.

PS.
The "rail generator" I refered to is a circuit onboard the OLED that is used to produce all the different voltage rails needed by the OLED. These are typically switch-mode supplies and they produce not only spikes on the power rail but the open-core inductors also emmit large amounts of air-borne radiation. This can catch people out because their circuit will work fine when laid out on the bench but as soon as they put it in a box with the OLED sitting just over their circuit, suddenly all the analogue levels/readings become erratic.

Good luck with your project and keep us posted on any new issues.
Taking something to market in a fully compliant manner is no small task and I'm glad you have shared your experiences.

 

[hippy]

I found this book to be a good read on EMC and associated issues including some of the case details which lie behind EMC regulations -

http://www.ee-techs.com/emc/autoemc-book.pdf

In my limited experience of EMC it was usually cables into and out of equipment which were most prone to interference or caused it. With digital signals there are often high-speed square waves and sharp digital switching with multiple harmonics. I imagine the low value resistors which improved things in ed1066's project work by 'rounding off' those square waves enough to reduce harmonics but not preventing them being read.

A previous company I worked for manufactured theatrical and disco lighting dimmers so the problems of mains switching were common though not something I was involved in solving. We never had any unexpected EMC issue in that respect which I believe came down to having designed robust and high-quality filtering around the triacs. High amperage phase-controlled mains switching requires quite substantial filters and heavy coils. Most of the physical weight and size of a lighting dimmer comes from those coils.

 

[BeanieBots]

Absolutely hippy! Older domestic dimmers did not have any inductors and produced all manner of undesirables that would muck up AM and LW radio reception. Newer ones incorporate large inductors but still produce enough EMC to be heard on an AM radio placed anywhere near the cables going to the lamps.

I recall a large theatrical lighting system that had problems when it was expanded. Some of the dimmers on the expansion unit would cause some of the lamps on the original unit to sudenly go full brightness when they were adjusted around the mid position. This was because the spikes generated enough noise to actually trigger the (thyristors in those days) original units!
I'd hazard a guess that neither units would pass today's regs.

 

[ed1066]

Thanks BeanieBots. This is very useful and a perfect launching point for when I need to return to EMC if the immersion heater control works well in my trials. Especially:

Zero Crossing is what it implies, when the mains voltage crosses through zero volts.
This is the best time to switch things on and off because no current flows and hence there is no surge, spike, flyback or other nasties to cause arcing and/or EMC.

You can get readymade solid-state relays that will do it for you or you can make your own using thyristor and a couple of trannies.

That's if the large mechanical relay is not going to pass. I'd guess that modern thermostats in immersion heaters are still relays and they are performing the exact same function as my controller. When you adjust them you hear a gentle click as they pass the switching temperature. I will get a feel for the emmissions from the thermostat in the immersion and compare it against my relay. I'll use a radio at shortwave and longwave, etc.

It seems this on/off switching is a breeze compared to the 'dimming' of 3kW, or the earlier 15W pump 'dimming' in the project I started this thread with.

Thanks also to hippy. I've downloaded the book ready. This is exactly what I hoped I'd be achieving on my pump 'dimming' and emmissions:

..... high-speed square waves and sharp digital switching with multiple harmonics. I imagine the low value resistors which improved things in ed1066's project work by 'rounding off' those square waves enough to reduce harmonics ............

Thanks , you've put it perfectly