If you edit the pad in the component editor it gives you the offset x and y and the size x and y of the pad. I found editing these text boxes to be the easiest method. The hardest bit is saving it into the correct library and loading the module back into the board, which you have done. I dont know if the inductors are any good. The bigger the inductor the less heat and higher efficiency, but higher cost and space, you have to weigh it up. Make sure it has a saturation current of at least 2amps. I managed to clear up all my errors, do you have an unconnected track? If you do the location of it will probably not be where you expect because the software doesn't know where exactly the net is separated, because it is separated everywhere. If you want I can probably fix it for you, but I need a zip with all main files + any special libraries and modules. Take your time buying your house, a bargain will come along.

Hi edit: sorry I probably wrecked the kelvin connections I did fix it up. I emailed you the brd and the gerbers. to save on my upload space. The problem was a few unconnected tracks that were actually connected but the DRC isn't smart enough to realise. there was only 3 but the the DRC thought there was heaps. The problems (gotchas would be more accurate) were caused by the grid settings, nets wont be connected unless they snap to the grid. if you change grid settings some of your pads will no longer snap. Also Kicad doesn't seem to remember your grid settings so you have to change it whenever you open up. Try to use the same one or two all the time. I had to go down to the very lowest to get one track to snap. The other problem is not actually a problem with the board but with Kicad. the DRC doesn't understand the copper areas and that everything inside is connected. So work around this I just join the unconnected ground pins on the underside of the board with thin tracks until the DRC says no errors. The actual copper pattern ends up being the same. Its like programming, get rid of the warnings so you can see important ones when they eventuate. Also a put vias into a lot of the ground pads, which I have only just learned how to do. Apparently my steff3, 2 channel, 36 volt, Black boards are finished manufacture and are returning to me.

No worries about helping, I know many many projects die a death and I don't want your project to die because of me pulling out, and advising you what to do is easier than doing anything myself. Sailing to work are you serious? I really thought you could almost copy my layout of the output stage. Assuming your connector requirements aren't too different. So layout the power stage first for one channel, including 2 x bjt and the sense resistors. With the best track widths, try to keep on one side if possible but use the 2nd layer for short runs if you need it.. Get this section laid out perfectly. Keep all the components close as is comfortable. If you want I can check at this stage, before you go further. I have a notification on this thread. Delete the components for the other two channels from the pcb editor. Then cut and paste the perfect layout to make 3 copies. then edit the component identifiers to connect the copied channels to the ratsnest. This should give the three channels fully connected to everything else in the same channel. After this layout the attiny and the caps and resistors, you could almost copy my layout. The only change I would make is the smd inductor I would put it on the top layer with the IC socket( use a socket). It should fit in mechanically, within the socket. Once you have these sorted and close to the attiny. give me another look. Lastly you need to route the pwm outputs and the current sense inputs. The Kelvin connections from the sense resistors are the thin 2/3 tracks together. Dont use vias on the Kelvin connections. Use Vias only where you really need to(which you will) firstly on the power and then the second preference is the the pwm. The kelvin connections only have less than 100mV max, and we are trying to detect changes in the micro volts. They are run together so that any noise present will effect each track similarly.

I've been playing around trying to do the track widths, do you place all the tracks yourself

I have done, I found it quicker now that I know what I am trying to acheive. If you had a lot of obvious connections then I think that the autoroute can do those fairly easily.

and if so how do you make sure you are connecting up correctly/ deal with vias?

The ratsnest will dissappear as each node is fully connected. If you are using vias for higher current paths then you can use multiple.

What widths do you think I should be aiming for, for my thin, medium and thick traces and what clearances would I need?

Default for thin, 8mm for thick and say 3mm if it is carrying some current. But reduce these widths just for segments where it is necessary.
Do your placement first obviously.

In terms of my layout, which is similar(ish) to yours can you see anything that will/may cause me a glaring issue?

I haven't looked have you changed anything since last time I looked?

Sorry Andrew, I was actually Sailing all weekend, it was fantastic.

Did you sort out how to change track widths, I know it is very cumbersome.

You probably found this out already but

assign a track to a track type in the digital rules.

When you lay it out it will take the track width of the track type.

If you want to temporarily make some tracks a different width such as for the ground plane, Where the Kelvin connections carry a tiny current and the power ground has high current you can just change the track width for the track type(eg. High Current) while you lay that track. Then set it back to normal.
Also sometime you want to run the track thin into the pins but run it fat for most of the length.
First run it thin, then change the track thickness for the track type and right click and adjust the track thickness on the segments that you want thicker.

for the thicknesses

Connections into the processor only need to be thin. Power tracks make thick. Tracks on the ouput stage make thick as you can but thin them down as they feed into the component legs. It basically comes down to the current through the track. 8mm wouldn't be amiss joining the inductor to the Fet and the diode, if you can fit that much.
I quite liked my posistioning of the components, to minimise track lengths for the high current tracks. But maybe you want the connectors to be organised differently.

As for the Kelvin connections run them as pairs or threes lik it did. The only place the ground of the kelvin connection touches ground is at the sense resistor.
the point is we want to measure the voltage of the ground at the sense resistor compared to ground at the micro, but using only the minimum current we can to do it. (Conceptually tricky). Also we want the minimum induced noise so we run the differential inputs right next to each other, for as long as posible.

Hope this helps

Hi You have managed a lot of the hard stuff, ie working with PCB layout is hard. Did you use autoroute? Looks a bit that way. The most obvious problem is the track widths. on the high current paths. Also C1 pin 2 is not connected. I think you could virtually copy my layout, apart from the step up transistors of course. Apparently component placement is the key to pcb design. In your Design rules you have only small track widths even for the power tracks. Also the kelvin connections aren't done. You are just sharing the ground plane. These are hard to explain, Technically the Gnd side of the sense resistors are shorted to earth. But the trick is that no short is a complete 0 milliohm. So for each sense resistor short the ground pin to the ground plane but as well run a pair of tracks to the attiny filter input circuit( 10k and 100nF) , Right next to each other one from each side of the sense resistors. These tracks only go to the filter they do not touch ground except at the sense resistor. Have a look at my layout it might be clearer. the idea is that only current that is required to measure the voltage travels along these tracks, and the loop area of the pair of tracks is minimised. I can try and explain again if you need but look up Kelvin connections if you have any doubt. You will make mistakes in the design but often they can be fixed with a cut and a join. My Fets had the wrong foot print, I incorrectly chose a smaller one but it wasn't a show stopper. The rest of my foot prints were right.

Hi

I think you should build something soon, but I have a different idea for you for control.

the netduino can talk SPI and the attiny can talk SPI. so you could drop the tlc5940s completely.

Just have the netduino set the dimming level direct to the attiny via SPI. The only trouble with this is you would loose one channel of PWM, giving you two outputs per attiny.
Although if you can handle a small blink while the communication is done you could have 3 channels.

Also I have never written any SPI code in AVR so you would be on your own there.

L1 will be ok, it just needs to fit mechanically and I think it will fit but it will be tight.
If it doesn't fit you can easily make it fit by soldering L1 just off centre or filing a small notch into the ic socket.
You understand you need to use a socket dont you? I am not sure if I made this clear.

I tried editing the pad offset, but every time I pressed OK, and went back in to the settings, it had gone back to zero?

Right click on the footprint and Open the footprint in the module editor. Set the working library to say 'Andrew' and put all your mods as you do them in here. Create a new part from what is in the editor and save it into your footprint library. At this point you can save it as you make changes. Once it is done then delete the old footprints and place the new footprints into your pcb. The tracks may be no longer connected, so reconnect them using your knowledge of grids and snap to.

One day when I get something looking good made I will post about my latest version, hopefully soon.

Let me know how you get on with your new boards, out of interest I know you have dropped a channel, but how small do you think I could realistically drop these boards to? I don't know if I definitely will, especially if these boards work, but I may lay it all out on one big board eventually with the TLC board built in and the power distribution for the 30 channels will be much more elegant and a lot of the connectors could be dropped. I could essentially get away with a GO! bus connector and a power plug, but this really is dreaming ahead!

Yes I think before the final board is built you could optimise, but it is so hard to get the perfect design from scratch. It is iterative and I think 2 iterations will be your minimum.
If you get this first board made in small quanties, turn it on then I think you will find unexpected things, good and bad, especially in regards to the lighting aspect of it.

first work out if it's m205 or 3ag. you can get fuses anywhere. get some 2A, 1.5,1A, 500mA and if possible 800mA. futurlec has cheap prices on these. if possible get some 100mA, just for checking if the board is ok. ie no shorts. Or get a current limiting supply. Have you got a voltage supply? quite clean (free of noise) would be good, this means with a wasteful linear regulator like the Netduino has. maybe order a couple of 1A LM317s and you could make something suitable. Remember you have 3 channels at 500mA and 3.15v. + the attiny power + drive circuit + losses. I would say you are looking at 1.5 A at 5V or thereabouts. the caps look good (they are being discontinued because of the pb. dont worry. PB free is difficult and not really worth the effort) the TVS diodes look good. for soldering you can do it by a normal iron. temperature controlled. or solder paste and hot air gun or solder paste and skillet use a eutectic solder if you can. this means it solidifies within 0.1 degree. so 63/37 is good. 60/40 sn/pb is bad. non eutectic as it solidifies in 3 degrees and it is easier to do bad joints. You can use this if you really want to but just make sure the joints are good. the small volume solder pastes are generally eutectic.

Hi Andy We are just having the hottest day in Hobart for around 40 years, 41.6C. I am sitting with a wet cloth on my head trying to cool down. Did you buy the house? If you did I hope you have a dedicated Aquarium room. the voltage across RL should be 5volts (+/- say 5%) at low currents ( high load resistance) and up till the current limiter kicks in. The Attiny is happy with 3.3v up to 5.5 volts and 5.04 volts sounds just right. You need two current settings very low and normal. Use say 30mA to test for shorts and then say 100mA to see if the led works and then you can probably discard the current limiter for that board.

Cant get some of the links for digikey. the 0.1uf are ok.

goto go sorry back in a few days.

Thanks those caps got me down to about £70!

Lets face it was your diligence not mine. I normally just buy the first thing.
I wouldn't change the BJTS they are good quality and not so expensive and have been used in Steff2.

I think Atten 858D , (Dave does a video too, #167) do a budget heat gun (dont get ZHIAOXIN 858D and YiHUA 858D, and check earthing before using), but If you want to save money a small wedge shaped soldering tip would probably be good enough for the whole job..

Can you think of anything else that will be useful?

Copper braid.

I have had my own mcpcbs manufactured. mcpcb I think it was $30 for 600 peices + $170 for tooling postage and TT payment. <sigh>
You can mount them on FR4 but need thermal vias. Each led doesn't need a huge amount of Al and it depends on a few factors but you can definitely get away with 50mm x 50mm for each XTE or XPG led. Probably even less. Heat sink at the back mean you can place them closer together, but they make a lot of light.
I am using Steff2 in my workshop and it gives enough light, maybe a little more than a single fluoro. 21 leds @ maybe 300mA. I don't think heat will be a big problem.

my calc was 3 x 500mA x 3.15 volts = 4.725W
+ attiny and gate drivers 5v x 100ma = .5W
+ inefficiency = .5W

Total power for 3 channels on one board = 5.725W @ 5v = 1.1 amps.

I'll go to really plain terms now, so in terms of voltage drop in series this would be 3.15v per LED so you would need say 12v, but we get away with 5v as the channels run in parallel right?

? each led has its own channel, in this case there is no series or parrallel. In the 12v 24v, and 36v designs you have the led strings in series. You cant really parrallel the leds without individual power resistors to share the currents equally, this is wasteful so not considered.

Current i.e. the 1 or 1.5 amps we will supply can be larger than the required 700ma plus any other losses, as the circuit will only take what it needs right? Would this increase if say we needed 5 LEDs in a row to say 4A or does current not drop like voltages does in series hence it would still be 1 - 1.5A?

So here's where I get confused, where do Watts come in to the equation?

For example will my 5v 1A wall wart be able to power all 10 of my boards and I guess I would need to distribute the power in a parallel design between the 10 boards right?

If you run 10 boards or 30 channels of 1 led then you will need 1 x 11 amp 5 volt supply. Or you can use 5 x 2.2 amp supplies or 10 x 1.1amp supplies. Dont parrallel the supplies just join the negatives as common. Keep the the positives of the supplies separate to each other.

Try not to overload a supply, you can probably use standard 1amp wall warts but maybe turn the power down a smidge. you can do this in the attiny firmware so that the Netduino firmware cannot cause failure. Check the warmth of the Wall wart while at full load. You don't really need to use wall warts there are other more efficient supplies.


Also I would put a smoke detector in your lab especially while still in development.

Sorry those caps were overspecced.

Try these.
445-1369-1-ND

on the caps you can go down to 16v which still leaves a comfortable safety factor. None of the caps need accurate tolerances or temp coefficients.

the fets should be at least 20v.
same with the bjts.

I probably waste a bit of my money but then again we aren't just about to buy a house.

Yes use the 805s if they fit your footprints. Well spotted. I think I have used the radial ones on some jobs because I once bought 100 of them. No hurry with the Digikey order they only take about 4 days to deliver, so probably wait about 2 more weeks and see if something else you need comes up. £9+ for 30 is way too much, the only ones that are even a bit expensive are the 0.1 ohm. I think I got a syringe of solder from farnel it was about $20. You should keep it cold, so I wrap it in plastic bags and keep it in the freezer.

Yes

Seen plenty of similar errors myself.

the not connected, you just need to zoom in an check it is connected, I think the colour changes in the middle.

If you don't want to connect to a pin, you put a "NO Connect" symbol on blue cross. It's just a message to the DRC.
Also if you label a node, and want to use this node somewhere else you might have mistakenly give a different name.
There are also power markers, that you place like components, these may have the same name as labels but are different.
Eg. you have to label a node as GND if you want to use it in different places, but you also need to attach a power marker to it which will be called Gnd.

ErrType(3): Pin connected to some others pins but no pin to drive it
@ (1.6750 ",1.2000 "): Pin 15 (power_in) of component IC1 is not driven (Net 10).

You have to say where the power is coming from, you can either ignore it or edit a connectors pin, if it is coming off board.
Ignoring it is fine if you are sure what is providing power. I did manage to get my design pass DRC but it was a bit of frigging around.


My mount holes are stupidly big on my board. I think my power vias could be bigger.

I found the large part of my problems, good news its an easy fix. Some of y 805 10k resistors that I squeezed onto 1206 pads were not connected. At least 4 of 10 were not actually connected. They actually looked connected. I now have the dimmer inputs going. Still haven't proved the current sense. This may take longer as it is hard to know if it's working until the loop is closed and stable. But I wouldn't wait for me on this as the idea is well founded and it should be ok and also it will take me a while to write this code. Be aware the current sense connections need to be Kelvin connections. This means look at my layout for the current sense. The idea is that these wires go nowhere but to the pin, and therefore carry minimal current and have minimal voltage drop, it might be the same node as another connection but you cant share a track. Also I give the ground it's own input into the adc. It's a bit weird but I am trying to minimise common mode noise. The differential inputs should run together as closely as possible. This stuff is sort of new to me too, but I think I have it right. The dimmer connections are not as difficult.

It looks like 100u and 33u as you say, I don't know why the marking is inconsistent. I am doing a redesign of my Attiny dimmer. I have just about decided on a schematic, I will let you know when I have more. I have finally sorted out the key role of the inductor, when you need it and when it is just desirable. It was never really clear in my head, finally it is clear, after all I am an EE, I should know better, it is actually quite obvious. You don't need the inductor when you're LED string can handle the full rail voltage continuously. eg. 4 x 3.2 volts leds at 12 volts. If they can't handle the continuous current then you need the inductor and you need to switch the supply (as in a buck converter) so that the LEDs are not overloaded. The smallest inductor you can use will depend on the switching frequency, the input voltage and the LED strings"V vs I" graph. It is hard to calculate exactly, but you would want to allow for a reasonable margin anyway, so an exact calculation is not necessary, just use bigger. I am thinking of either 64khz - 256 khz and therefore probably for me up to 470uF, but for you running off 12 volts the 100uF will probably be plenty. I will post as I have more info.

I have decided on a rough schematic, I am using the negative input as common again.
The details are on my domestic lights project . It is based on the Attiny but with some improvements.
I hope this helps.<br>

Hope your boards work out fine, you will have to let me know how they go.

Finally I have my lights working.

I haven't used or really looked at the Go architecture yet, but I think it may suit your project.

I think the 2uF are just for decoupling high speed transients on the 5 volt output.
The 5 volt output is not meant to be used by you so don't use it. It is used internally to power the logic of the chip.
It needs a decoupling capactior so if something suddenly turns on or off your logic still has close enough to 5 volts. ( I think it is 5v).

The thing about decoupling capacitors is that they don't need to be accurate, they just need to be big enough and fast enough, Surface mount ceramic is fast enough and 2.2u or 4.7u (which is what I used) is big enough.

In short 2.2uF is fine.

Finally the 180k does the equation for this take in to account the RSense and if so do I need to calculate this myself for my RSense of 0.285? Is there a suggested power rating for these?

I think the 180k is only a ball park number, it is about the minimum off time. It doesn't really matter as long as it is fairly close to that number.
do the power calculations for the 180k resistor. if your maximum ciruit voltage is 12V then your maximum power in the 180k = 12 * 12 / 180000 = 5uW.
If you do more electronics you will start to see that a high resistance equates to low power dissipation.

So if I was using say a 5v supply for 1 LED and a 12v supply for 3 LEDs is there anything else that needs changing in the circuit?

Minimum input voltage to the NCL30160 = 6.3 volts.
Just run them all off 12 volts. You wont loose much.

Are you spinning your own board or using vero board?
I have bought 100 leds and designed my own mcpcb, they should arrive this week, I hope the mcpcbs don't have any errors.

There is no way I would go from mains to led. you would have to do it for each channel.
Better to bring it down to some lowish dc voltage and use a led driver for each channel from there.

I was talking on another forum about my troubles with the NCL30160 and a couple of things were suggested.

1. run at a lower voltage ie. 12 or 24 even though the chip is rated at 40V
2. use a transient suppression diode across the vin. one that suppresses at a voltage just above the vin you want to use.
3. bread boarding high frequency power circuits is fraught with difficulty, and could have been the cause of some of the problems

Also I said before you may get away without the inductors but I think you need them unless your nominal led string voltage is near your maximum dc voltage.

eg. 3 leds at 3 volts = 9 volts running from 12v probably doesn't need the inductor.

but 1 led at 3 volts = 3 volts running from 12v probably does need the inductor because the instantaneous voltage across the led may cause a large instantaneous current that is too much for the led.

I am drawing a schematic to lay out a board, but it will be for 24-48 volts input. You don't need the black regulator it is only to drop the voltage down from 24v to 12. My understanding is you will be using 12v input. I have actually made up a 3 transistor black regulator on a bread board and it did work as expected. I will be using a 2 transistor regulator to drive the 12volt rail with no current limiting. I only need a small amount of power at 12v. Then from 12volts I use a regulator to bring it to 5 volts. Obviously this step loses 7/12 of the power. As far as the bom goes just look on the schematics. I am wondering if down the track I should use the attiny25 (8 pin) and just do one channel per micro, this may suit you better, but for this version I am using the attinyX61a 20 pin. Maybe what you should do is a block diagram of your entire system. Maybe you have. This can really help you see things once its gets complicated. Then add a couple vero board prototypes to a couple of channels and see what works best. I will post a schematic with micro attached soon, I want to send it to the manufacturers soon. Hopefully 1 week.

Hi I have all but finished the board for the Steff2, attiny version. I also have a BOM, I will post soon, I am at work now. I will probably send off to seeed studio or similar this week. The board picture is attached, the inductors haven't got a 3d model, they are vertically mounted torroids. Either 100uH or 220uH, probably I will try both.

Hi Andrew Just giving you an update, I am having real trouble with the firmware, not sure why it's been so hard. Good news is I can run the power stage and it runs well, beautiful sharp edges with a .2 u second rise time. This is as fast as we need, and the .2 seconds limitation is probably not actually driver anyway. It might be the 555 timer I was using for testing. I had to go back to Avr Studio 4 before I could even get a square wave out of the Attiny. Not sure why. I am currently running the driver from the Attiny so that side looks good. I haven't tested the current sense arrangements yet. I haven't managed to get the feedback loop going. But the fet driver looks good. Both are pretty standard so It is really only mistakes that make a problem. When I get the current sense going I will let you know.

I sent the cad files, you will have to dig up some of the libraries or give me a list of what is missing. I tried to send one more lib directory but it zipped to 87MB. It might be the 3d models that take all the space. In kicad speak libs are the schematic graphics and mods are the pcb footprints. You need to associate these, there is a good video tutorial on Kicad, and also a pdf. the inductors are worth a try. just as long as you can still use the footprint if they are no good. It's almost certain they use a standard footprint.

Here is the final schematic for the first spin and the BOM. The 0805 resistors I didn't bother getting the part numbers for. Same with some of the others which I already have. BOM was wrong, it has been corrected.

I know I am jumping in a little late here, but I just looked at the picture with the inductors. Aren't the ones on the left actually 33mH instead of 33uH?

I looked at this link coilcraft series

I must admit I don't know.

The bit in the bottom right corner is the black regulator right?

that's right.

ps. I changed the logo, it will come out on the silkscreen.

I almost mounted the torroid footprints upside down, that would've been a disaster.
I just sent the gerbers off to seeed studio last night.
If you want I can send you some, I will test them first though.

Hi Andrew How is the aquarium going? I have some of the NCL30160's but I have blown 4/5 of them. They are static sensitive. But seem to work as specified. I used the circuit straight out of the datasheet. I may soon try and rejig my attiny version for my domestic lighting project. I have a book on PID control and and some better surface mount fets. But I would recommend you continue with something like the NCL30160, to keep the part count low and to interface nicely to your tlc5940s.