DALI communications setup not working.....do you know why?

[ocset]

Hello,
Please help us with our lamp which sometimes fails to correctly  interpret DALI signals that are sent to it.

We have an offline, non isolated 150W LED lamp which is DALI dimmable.

In fact, whilst in the factory, before we ship the product, we send it a DALI signal which tells it what power level to start at, when  its  installed by the customer.
However,  in the factory, when the mains cable running to the lamp was long, the lamp  would not correctly  interpret the DALI signal.     When we shortened the mains cable, the lamp would correctly interpret the DALI signals. 

The attached pictures  show the  setup which does not work, and the setup which does work.

It is inconvenient for us to use the short mains cable. What can we best add to the “long mains cable” setup to make it work? 
I was thinking of  passing the mains cable through a high permeability ferrite torroid several times just before it enters the lamp. Would this be best? 

(  incidentally, the lamp  has a small 1W Buck converter bias supply, but  the LED drivers operate in linear mode. The Buck converter has a differential mode CLC filter before it. There is no common mode filtration. That is, no Y capacitors and no common mode choke in the lamp. In fact, on the lamp PCB, upstream of the mains rectifier bridge, there is no filtration components at all,  either diff mode or common mode.)
 

[ovnr]

First off, isn't the DALI PSU supposed to be connected to the USB to DALI adapter, not the lamp?

Also, you should stop trying to figure out how to fix DALI comms by improving the mains cable filtering in the factory, and fix this on the lamp design itself. If you have issues in the factory, just imagine what sort of weird-ass issues the customer will have!

[ocset]

First off, isn't the DALI PSU supposed to be connected to the USB to DALI adapter, not the lamp?

.
Thanks,  the DALI PSU simply provides the 16V rail which the USB-to-DALI converter pulls down in order to send the DALI signal to the lamp…..so the “DALI PSU” needs to be connected to both the USB-to-DALI converter and the lamp’s DALI input.
(there’s a current clamp at the output of the DALI PSU so that the USB-to-DALI converter can easily pull the 16V output  down when signalling to the lamp.

The DALI receiver circuit on the lamp PCB  is as on page 3 of AN1465 by microchip...
http://ww1.microchip.com/downloads/en/AppNotes/01465A.pdf

Also, you should stop trying to figure out how to fix DALI comms by improving the mains cable filtering in the factory, and fix this on the lamp design itself. If you have issues in the factory, just imagine what sort of weird-ass issues the customer will have!

Thanks,  we suspect  (but don’t know for sure), that the reason for the problem is that when the mains cable is long, (and therefore presenting a larger inductance back to the mains), the laptop power supply is drawing noise currents from the actual lamp itself….and thereby upsetting the DALI receive on the lamp PCB. We could potentially mitigate this by putting an X2 capacitor at the mains input to the lamp. However, the lamp PCB is surface mount only, and X2 capacitors are not available in surface mount. We do however, have a 220n, 630V ceramic capacitor downstream of the mains diode bridge in the lamp PCB. There are no filter capacitors or inductors, either diff mode or common mode, upstream of the rectifier bridge on the lamp PCB.

If you have issues in the factory, just imagine what sort of weird-ass issues the customer will have!

Thanks, but when we tell our engineering contractor this, he simply tells us that in the actual  customer installation, there will obviously not be a laptop with its power supply  there. So there will be nothing to draw the  noise currents that are upsetting the lamp’s DALI receive.
--------------------------------
Incidentally, when we tried to send the DALI signal to the lamp at a different location, (in the engineering office) where the mains cable was indeed short ( about 1 metre), the DALI signal was not received correctly if the lamp’s heatsink was not earthed. (the lamp driver PCB sits on an insulation pad on top of the heatsink)….but it was received properly when the heatsink was earthed.

[ovnr]

If you have issues in the factory, just imagine what sort of weird-ass issues the customer will have!

Thanks, but when we tell our engineering contractor this, he simply tells us that in the actual  customer installation, there will obviously not be a laptop with its power supply  there. So there will be nothing to draw the  noise currents that are upsetting the lamp’s DALI receive.
--------------------------------
Incidentally, when we tried to send the DALI signal to the lamp at a different location, (in the engineering office) where the mains cable was indeed short ( about 1 metre), the DALI signal was not received correctly if the lamp’s heatsink was not earthed. (the lamp driver PCB sits on an insulation pad on top of the heatsink)….but it was received properly when the heatsink was earthed.

I don't know what the intended use for the lamp is, but I do know that there are a lot of horrifyingly noisy scenarios out there. Sure, I tend to overengineer things, but I'd never ship a product that I knew was vulnerable to something as simple as a long extension cord and a laptop.

Anyway: If you did indeed implement the DALI transciever as shown in AN1465, you should have a fully opto-isolated interface. Noise on the lamp mains should have an absolutely minimal effect. You may find some conducted emissions through the DALI PSU - have you tried other options, just for kicks?
As for the capacitors: There are plenty of X2-rated SMD ceramics. There's even Y2-rated film caps. And even if there weren't: I'd be shocked if your design met emissions without some filtering, and if that means hand-soldering a bent through-hole part onto each and every board, so be it.

If you just want to find a quick fix for the issue: Try using two long extension cords, one for the laptop and one for the lamp.

[ocset]

I'd be shocked if your design met emissions without some filtering,

Our next revision of this board will use inductors in the DC Buc positive and negative, (ie just downstream of the mains bridge)  and more ceramic capacitors just downstream of the mains diode bridge....and this board (not yet out) does indeed pass conducted diff and  common mode emissions...but that board is not yet ready.
This new version will not have Y capacitors though.
I suspect we may need  Y capacitors at the mains input simply to filter away noise like that from the laptop power supply. 

By the way, the boards we are currently doing are for installation at our investors' own private estate.

Anyway: If you did indeed implement the DALI transciever as shown in AN1465, you should have a fully opto-isolated interface. Noise on the lamp mains should have an absolutely minimal effect.

Thanks, but we are using non-shielded opto's in the DALI RX/TX circuit, and i wonder if we are therefore getting common mode transients going through the optocoupler?   
The following discusses  shielded optocouplers...
http://www.powerelectronics.com/power-management/shielded-optocouplers-prevent-line-switching-power-supply-upset

We are using simple TCLT1009 opto's which are not 'formally' shielded...   
https://www.vishay.com/docs/83515/tclt1000.pdf

There's even Y2-rated film caps.

Thanks,...they look good (SMD Y2 caps...
https://content.kemet.com/datasheets/KEM_F3018_SMP253_Y2_250.pdf
)
...however, the terminal is just 0.5mm wide, and so i suspect these will be susceptible to being knocked off the board if some kind of stabiliser is not also used to  support them on the PCB?
Another point about the  Kemet SMD Y2 capacitors, is that they have just 11.7mm between their footprint terminals, so that means that on the PCB, there will be just 11.7mm between  mains live  and mains earth…and that’s bare metal to bare metal because it’s the actual PCB  footprint’s pads….so that is surely illegal?......i mean, its supposed to be  2.5mm minimum spacing I  am sure? (regulations)......and that’s when there’s solder esist covering the pads…..so the  SMD Y2 cap depicted here is surely illegal?

[dmills]

The DALI receiver circuit on the lamp PCB  is as on page 3 of AN1465 by microchip...
http://ww1.microchip.com/downloads/en/AppNotes/01465A.pdf

Which probably fails utterly to meet the DALI spec for slew rate, you need to actually read the standard not just implement some dev board circuit as if it was a final product...

Unshielded opto plus 'isolated DALI' circuit mounted hard against a non earthed heatsink coupled to the mains capacitively, what could possibly go right? That DALI bus is probably the nearest thing to 'ground' around that board, so where is the common mode going to go if it has no other option.

I note from your diagrams that the one that works has the laptop running on battery, while the one that don't has it on mains, maybe a clue there!

You need to fix this on the lamp itself, installers are absolutely not above running lighting circuits in the same containment as the lift motor wiring or the booster pump inverter supply cables, and if you think a laptop is a problem....

I would strongly suggest hiring yourself a better engineer if  "There won't be a laptop" is the answer offered to a clearly serious immunity problem.
I would take discovering something like this on a prototype as a good thing, it will be orders of magnitude cheaper to fix now then if it is discovered on site after the decorators have been in!

Regards, Dan.

[ovnr]

Thanks,...they look good (SMD Y2 caps...
https://content.kemet.com/datasheets/KEM_F3018_SMP253_Y2_250.pdf
)
...however, the terminal is just 0.5mm wide, and so i suspect these will be susceptible to being knocked off the board if some kind of stabiliser is not also used to  support them on the PCB?
Another point about the  Kemet SMD Y2 capacitors, is that they have just 11.7mm between their footprint terminals, so that means that on the PCB, there will be just 11.7mm between  mains live  and mains earth…and that’s bare metal to bare metal because it’s the actual PCB  footprint’s pads….so that is surely illegal?......i mean, its supposed to be  2.5mm minimum spacing I  am sure? (regulations)......and that’s when there’s solder esist covering the pads…..so the  SMD Y2 cap depicted here is surely illegal?

Film caps: Well, these are film caps. They can take shocks better than ceramic. I wouldn't be overly worried. Remember, there will be a generous solder fillet as well.

Also, I think you got your numbers mixed up a bit. 11 mm spacing is fine for anything mains. The smaller ceramic Y2s are a bit tight, but are likely meant for potted assemblies. They're not illegal - it's your responsibility as the engineer to design something safe.

Besides, weren't you skimping on the clearances earlier?

[Kalvin]

Have you certified your devices yet? I guess your lamp's pwm-dimmer is emitting some serious EMI and you will have to fix those issues as well, and those fixes will also solve some of the problems you are seeing here.

[ocset]

I note from your diagrams that the one that works has the laptop running on battery, while the one that don't has it on mains, maybe a clue there!

Thanks, ..sorry but no, the laptop is mains connected in both cases. Sorry...damned me for forgetting to draw the mains cable running to the laptop. 

a clearly serious immunity problem.

Thanks, so do you agree with us that this is a common mode susceptibility problem?, and that Y capacitors to the mains input earth wire and a common mode choke at the mains input to the lamp are the  remedy?

I guess your lamp's pwm-dimmer is emitting some serious EMI and you will have to fix those issues as well, and those fixes will also solve some of the problems you are seeing here.

Thanks, but the PWM dimming bit is just a square wave signal coming out of a pin of the micro in the lamp...and then it gets filtered to DC...where it acts as the DC reference to the led current error amplifiers. So i dont believe there will be much noise from this.

Which probably fails utterly to meet the DALI spec for slew rate, you need to actually read the standard not just implement some dev board circuit as if it was a final product...

The DALI RX/TX circuit on page 3 of AN1465  does indeed have a slew rate which is longer than  we would like…..but to be honest, the length of each pulse in the Manchester coded DALI stream is 416us, and so this is OK for that cheap opto circuit……and we hope that our software engineer has programmed the micro to read the DALI pulse in the middle of the high or low pulse, and not near the edges where it may still be slewing.

AN1465
http://ww1.microchip.com/downloads/en/AppNotes/01465A.pdf

Unshielded opto plus 'isolated DALI' circuit mounted hard against a non earthed heatsink coupled to the mains capacitively, what could possibly go right? That DALI bus is probably the nearest thing to 'ground' around that board, so where is the common mode going to go if it has no other option.

Thanks..It makes me glad our product is no installed on aeroplanes.....none of the electrical circuitry has an earth connection on a plane, and you'd wonder how we'd  manage then...since our DALI circuit wouldnt work without us connecting the heatsink to earth.

[Towger]

I know FA about DALI except is is the industrial equivalent of DMX, but looking at diagrams 'ground loop' comes into my head.  What happens if the laptop runs on batteries?

[dmills]

Uh Treez, DALI specifies a MAXIMUM slew rate as well as a minimum, I was saying that I don't think your transmitter likely meets it.

The point about the heatsink is that you are going to get stuff coupled onto it, it is your responsibility to engineer things such that that does not pose a problem in any of the likely use cases.
It is not the 'earth' that is important, it is making sure that incidental coupling is at a low enough level to not matter (And I have done sensitive instruments for aerial use, it is no harder then any other sort of electronics, well apart from the paperwork!). 

Seems to me you got a very cheap (and even better, repeatable) susceptibility fail, would that all my problems were so amenable to investigation, time to fire up the scope and differential probe and have a poke around.

Regards, Dan.

[Kjelt]

Treez I responded in other threads but take this advice, buy a couple of dali drivers from A-brand manufacturers and study their DALI circuit.
Why do they have around fourty components and you have ten?
DALI lines could be a couple of hundred metres long and even parallel to mains, heck could be part of a 5 wire cable and oh yeah those A-brand interface even survive if you put mains on them accidentally since that was in the early days a major cause of returns since installers are not always fully awake when they wire the devices 

Then also buy a few different DALI controllers/masters from A-brands and put some 50 metres of cable between them, mains and DALI, it should still work flawlessly.
The issues you describe will result in a major complaint and return issue from customers, better deal on forehand and make sure you are DALI compliant.

[ocset]

Uh Treez, DALI specifies a MAXIMUM slew rate as well as a minimum, I was saying that I don't think your transmitter likely meets it.

Thanks, but  it seems to be widely accepted that received DALI signals can very easily be bit banged by a micro. We can read the input once every 50us and reconstruct the received DALI signal from that...the DALI RX/TX circuit on page 3 of AN1465 is easily amenable to that kind of bit banging surely?

It seems very surprising that Microchip's headline app note (AN1465) on DALI is giving a wrong circuit. Aside from that, we already have other units on other products in the field which are working with the page 3 circuit.

This thread discusses just bit-bashing received DALI commands, all agree that it is fine to do this...
http://www.edaboard.com/showthread.php?t=371919&p=1593089#post1593089

Then also buy a few different DALI controllers/masters from A-brands and put some 50 metres of cable between them, mains and DALI, it should still work flawlessly.

Thanks, i dont see why our system would be affected  by mains and DALI wires running alongside each other for several 100 metres. In the described problem of this thread....the problem is thought to be a laptop power supply drawing noise from the lamp input, due to the impedance seen looking back to the mains being high due to long mains cable...in the actual installation, as you know,  there will not be a laptop power supply up with the lamp.

The issues you describe will result in a major complaint and return issue from customers, better deal on forehand and make sure you are DALI compliant.

Thanks, as you know, our DALI TX/RX circuitry is perfectly capable of receiving the 2 byte dimming signals. If you look at the DALI pulse of the thread pasted in this post, you can see that its not a hard job for a half decent softy to be able to read that in to a micro, surely?

time to fire up the scope and differential probe and have a poke around.

Thanks, thats what i told our contract engineer....but he said its high frequency noise, and that its affecting the bus's inside the micro, and that we won't see anything at the micro's DALI RX pin. Do you believe that if we scope the micro's DALI RX pin when using the long mains cable, that we will see the DALI pulse  all distorted?

[mikeselectricstuff]

It seems very surprising that Microchip's headline app note (AN1465) on DALI is giving a wrong circuit.

Microchip write app notes to show how their MCUs can be integrated into products that maufacturers have the field-specfic expertise to design , i.e. "Hey DALI people, you can use a PIC in your design - here's a starting point", not "Hey lighting manufacturer, here's everything you need to know to create a  DALI interface. 

Do you believe that if we scope the micro's DALI RX pin when using the long mains cable, that we will see the DALI pulse  all distorted?

Are you seriously saying that you haven't done this already   

[wraper]

Do you believe that if we scope the micro's DALI RX pin when using the long mains cable, that we will see the DALI pulse  all distorted?

Are you seriously saying that you haven't done this already   

In some of the previous treads it was said they did not even have a scope  .

[ocset]

Are you seriously saying that you haven't done this already

Thanks, we have done it, and  the DALI pulses at the micro's DALI RX pin looks perfectly normal.
As you know, what we cannot scope, is in the myriad of  bus's and circuitry within the micro itself.         

Microchip write app notes to show how their MCUs can be integrated into products that maufacturers have the field-specfic expertise to design , i.e. "Hey DALI people, you can use a PIC in your design - here's a starting point", not "Hey lighting manufacturer, here's everything you need to know to create a  DALI interface. 

Thanks, attached to this post is one pulse of the DALI signal that we are receiving...and this is what it looks like when the DALI comms is **not** working...as you can see, its a very amenable pulse     .......and its produced by the circuit on page 3 of AN1465......we can think of no earthly reason why a softy couldnt write code to  correctly receive that pulse in....even despite its obvioulsy fairly slow rise time.

I know FA about DALI except is is the industrial equivalent of DMX, but looking at diagrams 'ground loop' comes into my head.  What happens if the laptop runs on batteries?

Thanks, I see what you mean, -because the laptop, the DALI PSU, and the lamp itself all have connections to earth, and these can all “see each other”…and even the secondary of the DALI PSU will be able to “see” earth …(ie through the Y capacitors which must exist inside it)……..however, if a “Badboy” earth loop is causing our problem, then you’d think the problem could be solved by disconnecting the earth from the lamp and then it would work…..but, this is not the case, and , in fact,  disconnecting the lamp’s earth connection is one thing that makes its DALI receive not work.   
I suppose we can console ourselves that in the real installation, there will not be a laptop installed up with the product.   
I must admit I am  in some small way surprised that none of the other contractors in our company that we have spoken to about our problem  have said that our lack of any AC mains filter in our lamp could be causing the problem………I mean, surely since the problem is  most likely caused by the laptop drawing noisy pulses from the lamp’s input (due to the high mains impedance looking back to the mains), then I would have thought a filter at the AC input to our lamp would be the obvious solution?       …and at that, a common mode filter instead of a diff mode filter.
Also, getting the softy to receive the DALI pulses by consecutive 50us reads and then reconstructing the DALI pulse train, seems like a good idea….but again, the other contractors here have just brushed it aside. 

[ovnr]

Dear treez: Please find competent contractors. This whole project (how many threads has it spawned?) seems like an utter clusterfuck.

[dmills]

Amen, DALI is just not that hard, but a reliable DALI interface that survives all the shit that installers pull is going to take someone used to asking not "How will this work?" but "How can this fail?".

That is where you tend to find high parts counts in real DALI interfaces and all sorts of carefully considered error recovery code.

Providing your decoupling and input conditioning is reasonable it is unlikely to be interference induced into the chip the geometry does not favour it unless you have done something insane like used a pic with multiple ground pins and only connected some of them (or worse, which I have seen, used two ground pins on a processor to bridge a ground net, worked about as well as you would expect).

You need an embedded guy who is used to debugging with a scope (And need to be examining that rising edge around transition with a much higher sampling rate), seriously for the right consultant this is just not that hard.

Regards, Dan.

[ocset]

You need an embedded guy who is used to debugging with a scope (And need to be examining that rising edge around transition with a much higher sampling rate), seriously for the right consultant this is just not that hard.

Thanks, this is indeed part of the problem, our software consultant lives over 1000 miles away in another country. I have no idea how he is reading in these particular DALI pulses. The DALI pulses here are ones that we send to the product in production to configure the lamp to  start its life at a certain power level.

I do not know if he is using the DALI libraries to do the DALI receive, or whether he just bit-bashed the pulses in.

We cannot speak to him, he works most of the time for his other clients. When we have spoken to him, he has told us that almost all of the problems that we have had are due (in his opinion) to the fact that our product does not have an AC filter. (ie, does not have the standard common mode and diff mode filter that one expects to see at the AC input to an offline LED driver.
We do not have this because our PCB is all surface mount and common mode chokes etc, certainly with the low height profile of our enclosure, and X2 capacitors, are through hole components. Our investors categorically will not pay for the inclusion of an AC filter...certainly not until we come to an absolute grinding total halt......But since we made the DALI problem go away by reducing the length of the mains cable in production, and earthing the heatsink, they are saying its all hunky dory now and thats that.

Come to think of it, none of the kind responders here have ever alluded that  the lack of an AC filter in our product might be the cause of the problem.

[Kalvin]

I tried to hint that you may have a problem with the lamp's PWM driver which may emit or conduct enough noise into the AC line which will make the system fail. Also, without proper AC line filtering you may not be able to pass the required certification. Without proper certification you may not be able to sell your products.

Just build/hack a proper AC filter for one or two devices and see what happens.

[dmills]

Well you should probably be able to design this without a mains filter, but it is often easier to make a certain type of immunity problem go away by including one rather then doing all the engineering to make the bulk of the doings immune (Which may actually cost more then a mains filter!).

Filtering goes in because it is often the simplest and cheapest way to solve a problem, but it is seldom the only way.

Consultants many time zones away, it can work but that is not the way to bet.

Find someone local, this feels like a £10K, 1 month sort of problem and you have blown that on missed opportunity cost in all probability.

Regards, Dan.

[ocset]

Just build/hack a proper AC filter for one or two devices and see what happens.

Thanks, we will try that, the problem is that its probably a common mode noise problem, and to get proper common mode filtering, as you know, one really needs the common mode  filter circuitry to be 'tight' with low loop area, and indeed, built on the pcb to facilitate low area current loops.....and not just wired into the mains cable upstream of the product....this will mean a new PCB just for the purpose of testing to see if adding an AC filter helps...and our investors wont want to pay for such a thing....unless we can come up with something to persuade them that  mains connected products that dont contain AC filters are definetely more likely to suffer noise problems.

The problem is that we have many non DALI lamps installed, which dont have micros in them but do have control integrated circuits to control led current, and out of 25000 installed, there are precious few problems.....so our investors are just saying that an AC filter looks like not being needed.