I think they will solder alright, it is just that if the pads are too big they may short out through the iron shield.
But if you just put a tiny square of insulator underneath, you can solder on both legs any way. It's a bit of a bodge but will work if the footprint is wrong.

Just going to few about 12 house now, what a Saturday! But I will make the amends and catch up later.

Does that mean you've been drinking for 12 hours? Don't fall in the aquarium, I am picturing some sort of James Bond evil villain type of aquarium.

what those atlas guys say about remote monitoring, is so true.

Had a 5 minute look, I couldn't see any problems. You could spend the rest of your life adjusting the silkscreen, it looks good enough. Seeed has changed its ordering method, it is a bit more straight forward. Basically rename your drl file to .txt and zip it up. I did Steff4 last night which just fixed the disaster that was steff3( where I didn't really look at the Gerbers). It was $32 10x10 posted. (Green, I have lost my confidence).

I thing I found useful was a SMD resistor and capacitor book. It cost about $35 from Ladyada I think, but many people sell them, they hava all SM805 foot prints in useful ranges. I will look when I get the chance. I have finally convinced my boss to let me design a board for work. Solar battery charger just 20W. I better not stuff it up. He wants me to use MyroPCB, they do a fair bit of manufacture for us. I am not sure, if they will fit, as there is no footprint. If it is metal under those inductors then you will have to watch for shorting via the inductor sheild if the pads dont fit properly. But if the boards came back and were wrong it wouldn't be such a problem as you could put down the flimsiest layer of plastic or paper to insulate this and then solder around it. the little inductor under the ic might need to move closer to the edge, maybe half a pin pitch. just so it doesn't get blocked by the ic socket. I would put + and - on the silkscreen top layer for each connector. but outside of the footprint so that you can easily work out +- when you are plugging. Use the text tool in Kicad.

I added some drill holes, I will probably relay the board slightly once I have tested one working.

It is hard not to make a mistake on a board. You will definitely see improvements you will want.

the Gerber lines up properly, I don't know about the .pho file.

Just make sure the foot prints of the stuff you order matches the ones on your board.

In steff2 I ordered D2Pak fets but had footprints for DPak fets luckily this mistake wasn't critical as I could force them on.


If you want to use Seeed Studio then I would scrape off approximately 1.1mm of the height of your board. It will save you money.
If you use someone else it may not matter.

Height : 30.4 -> 82.5 = 51.1mm

Width: 123.2 -> 261.6 = 138.4mm

Seeed Studio use 50mm increments for pricing.

In kicad there is a yellow pcb edge layer, right click on the lines and edit the y value to make it 50mm long.

Follow the instructions from Seeed Studio to the letter, for you first spin use green it is cheaper.
If you cut the board down to 50mm x 138mm then you can buy the 5cm x 15cm option from seeed.



PCB Cart is another place.

But I think UK has a lot of manufacturers too.

thats an important cap. the digikey part will be fine. The only two sizes I used are 805 and 1206 so the big ones are 1206. I think all the caps are 805 but the resistors are mostly 1206. the 0.1ohm are 1206 I think. different board houses have different ordering requirements I have been using Seeed Studio (very very cheap but they have a few hoops to jump through). Make sure after plotting you view in Gerber viewer. It should be 6 layers + the drill file. Put them all in the gerber veiwer together and check alignment of the holes and tracks. Make sure everything sits on top of one another. the layers are. Copper top and bottom solder mask top and bottom. this layor covers most of the board except where you need to solder. silkscreen top and bottom. To plot gerbers press plot and use defaults. There is a tute on this out there. Basically I just use defaults. Dont invert the Y axis. Also put the board edges onto the plots. plotting is actually quite easy. I just sent off steff3 to Seeed. ( 2channels, 36V, 100mm x 50mm board, black) Hopefully this is the One. Steff2 was too big and too green. Hard to put into a light fitting. Edit : oh and dont forget to put mount holes in, like they did with raspberry pi. I use a model called '1pin' which has a drill hole of 1/8 inch. This will let you use 3mm screws and rivets.

I just got my failed boards back yesterday, they make them in a few days but it takes about 3-4 weeks in the post. the black boards look great shame they don't work. You can see in the gerbers if the board outline is there. When you plot there is a checkbox to put on the board outline layer. the postage to Aus is just 7 dollars I made my Steff4 boards green (yuck) to save money,they are only $25, + 7 dollars postage. in other words just pay the postage might be the best option. my last order was Ordered: 29/10 shipped 6/11 arrived 18/11 thats actually better than expected. One more thing on your Gerbers, I think you should decrease the solder mask clearance. so Set preferences -> dimensions ->Pads Mask Clearance->Solder Mask clearance to 0.0508 mm This will mean the mask almost is exactly the pad. Then re plot the gerbers. Otherwise you can have pads close together that dont have a mask separating them, the solder can bridge these accidently quite easily. I fixed this in Steff4 but it was the case in steff3.

It will be fine without the solder mask changed. Just be aware that on certain pads there will be no solder mask between them. so check them visually before you power on. Check the close ones.
8mm ? 8mil big difference. Very confusing, if only the world would go metric.

I choose 6mil because it didn't cost any more. but I don't think you have anything will a very small pin pitch so it wont matter.
I just used a chip that was a DFN (QDFN) package. Then it starts to matter. 50mil between pins. unlike the dip which has 254 mil.

the fuse holders I used are from futurlec

futurlec fuses

the picture at the top with two separate holders. I cant remember if we used m205 (5mm) or 3AG(6.3mm) but obviously you need to get the ones that fit the footprints. Check the length of the mechanical outline, it will say what fuse size we used.

update: I checked my steff2 boards and they have m205 ( 20mm x 5mm) so I think you have the same.

I used no lense yet, it is actually what I am having the most trouble with. the leds can be very dazzling and they need some diffuser. The board doesn't need heatsinking but the leds probably do. I just riveted them onto aluminium. I didn't even use heat compound as it is too messy. I tend to run everything fairly cool (<500mA) so they almost get away with just the mcpcb.

Yes the pots will be good.

remember for testing you need a current limited supply.

You can make a current limiter with a pot on vero board. and maybe just use a wall wart.

something like this.
My link

Or maybe start with some very low fuses and if you see a bit of light on the leds then move in bigger fuses.

You will also need some code , which I must have posted before, that you need to adjust and upload. You will need AVR Studio 6 which integrates somehow with Visual Studio.
And a programmer board. to program the chips. I have an AVR dragon but I think a Bus Pirate would do the job, it also might be useful for your SPI, I2C stuff.
You can probably also use a netduino if your very clever. I don't recommend this path.
The coding part of it is harder than C# and VS. But you should be able to almost straight copy my code. Its in C.

Hi Just a bit of an update. I got my Steff4 boards back which use a similar design to yours. And guess what, no big mistakes. Thats a relief after what happened in the last version. They didn't print my silkscreen though. I have written some code that works for them but came up against some weirdness that took my a while to sort out. Basically we have changed the topology from my original design. Now we are measuring Fet current instead of load current. Looking at the circuit, The Fet current + the Diode current = the load current. At high duty cycle these will be close but at low duty cycle driving into a potential short circuit these will be way off. So we have to adjust for it in software. We know the Rail voltage and the Led load and the Fet current so we can work out the rest, and adjust accordingly. The only thing is that it wont be super accurate or super linear, unless you do some specific calibration and create a map. This circuit would not be good for colour balancing. This is a bit of a shame as one of the selling points of the 5940 was its accuracy, but really the Leds aren't that accurate anyway. When you want I can send some code. I still want to refine it for load short circuit detection. My C code looks like desktop code, it is not optimised yet at all. I wont optimise unless I have to. Almost no macros and passing everything on the stack. C can be so annoying, remember to cast (INT32) each operand up before any multiplication or division, it gets me every time.

1.5 amp is fine. any adjustable voltage reg is fine LT1117 is not doubt more expensive but they work in the same way. the transistor just needs to have enough current / power capability to handle the loads. attached is an LTSpice schematic. R2 sets the max current and R3/R4 set the output voltage. It will be a nice clean voltage coming in, from the regulator. I suggest R2 to be say 10 ohms to test the boards for shorts and then 1 ohm to protect the leds while developing/testing the code. Maybe a switch to switch one of the resistors in. Play around with R2 and R4 and RL and see what changes. You need a larger voltage on the wall wart for this because the current limiter and Vreg both use significant voltage. Once you are out of developement phase, you can go back to your 5V wall warts. Assuming your boards pass the test.

I just lost my post when I searched for Digikey parts. here it is again.

these are suitable.
BC337-40RL1GOSCT-ND
AE10015-ND

Are these ok for the LV317?

I would use through hole stuff for the current limiter and put it on vero board.

Can you get a cheap resistor pack somewhere like Futurlec
Use ltspice to simulate, it is way easier to use than Kicad, you will learn heaps. Play with your current limited supply. When you get it right then you can order.

heat proof tape I haven't used it yet but from what I have heard it is useful.

Final question what exact IC holders did you get, because looking at the ones I have that don't have enough pins theres not a heap of space for the little inductor, I'm sure I've seen thinner versions before but thought its worth checking?

Inductor will be fine, I have done it this way twice now.

In terms of LED load replacement in the form of resistors what would you suggest and would you suggest changing my current sense to a lower value for initial testing i.e. to produce say 300ma instead of 700ma or just do it in the firmware? If so can you suggest some component characteristics for either/both?

Doesn't really matter, even if it is 50mA. If you have the leds fused properly then you can use the leds. Dont change the sense resistor. Just change a constant in the firmware.

The solar charger sounds cool, have you implemented any battery management?

I haven't actually got a battery or a solar panel yet. This was actually my one of my other work projects, but they wont finish it now. I am going to keep going for my own benefit. Hopefully I will sell the product back to them, at vastly inflated prices of course. (Heh Heh Heh)

I am monitoring the charger current and the load current because the battery current is actually the difference in these. If you charge too slowly you cause sulphation apparently. Many systems don't measure the load current, I also need to temperature compensate.
Previously one of our installs stuffed up because the ambient temp was 45C and we fried all the batteries.
the client was ropeable, all I could say is sorry our equipment is not suitable, and withdrew the solar unit from sale.

Watch a few vids and practice your heat gun before trying to do the full board. Maybe do the electrolytics with the iron.
One guy said there is no such thing as too little solder paste.
Try to roughly follow the suggested temperature profile. Warm up slowly, dont stay more than 30-60 seconds at full temp and cool down steadily. You know it's too hot if you burn the boards.

It's not so bad being retrenched, I already have a few small electrical jobs come up and I got a few months pay. I am enjoying the different time scale and being at home.
I am hoping to get more electronics than software for a change.

My only beef with losing my job is the fear they gave my main project, which was software, back to the consultants who will F**K it up.

A Little software saga about my work project (not actually a quote):

Project Design done with the consultants me and one other staff member. difficult software but good design outcome. Consultants did great job here.

The project was basically an Autocad plugin that talks to a SQL database which synchronises with our CRM Server which loads data from our Quickbooks.
Data is also held in the autocad design files that get updated as changes are made in CRM or in another Windows forms app.
Sales guys request designs through CRM and are returned the PDFs of the plotted designs through CRM.
Autocad designer does risk assessment and then design to meet the assessment. Autocad releases design back to DB which synchronises back to CRM.
Sounds complicated, it was and I was hoping CRM would come to the party and it actually did. When I first heard of what we were doing I thought this project has fail written all over it, but didn't say anything as I wanted the job, and this project was why I was hired.

Then we got the consultants to tender for the project, They tendered $45000 and said they could do it in 3 months. I said it wasn't possible for me to do it in 3 months. So I couldn't compete, I would just manage them. Which isn't my strongest suit.
So they get the tender and after 6 weeks I check progress and they haven't started. 3 months and the Winforms app is sent to me but it is just a hollow shell of an app they once sold to someone else. One year later and it doesn't work and the winforms app doesn't even make sense so I don't know how to use it. the Synchronizer has the code written but not tested so obviously it has at least a month of debugging to get it going.
The CRM modifications were done, but probably only at Alpha stage.

So at that point I took the app back because they were too busy to do it anyway and in 6-12 months it finally did work and was being used. We have done about 50 designs in the last couple of months.
Stage One was finished.
I manage to throw most of there code away. Not out of spite but anything I don't understand has to go, anything that is hard to debug also has to go, such as those annoying .Net Visually designed DataSets.

So now I am on the outer, but I am hoping my project doesn't die from neglect.
The other guy I worked with on this project is an extremely good BA but a dodgy coder, but I think he can handle it, if he gets it given back.

Anyway that's my rant. Apart from that life is very good.

Tell me when your boards arrive. And how they work out.

I got my boards back for a Solar mppt charger based on the Attiny861a and I was checking the current measurement, it only uses 100mOhm sense resistor and the attiny firmware control loop, it is rock solid, doesn't move a bit. That is good as your boards have the exact same control loop.

When you need some firmware I can give you the basics.
I have one chip programmed to test each channel initially. So you just plug them in and check it on the Cro see if it's all ok.
Only then put the real firmware in.
Once that is solid you can make some descisions about comms from the Netduino, and what Netduino.

Hi I've been a bit preoccuppied too. Also I'm actually having trouble with the mechanical and light aspects of my light fittings. I would keep the little drivers as functional blocks and close together. ie. inductor, diode, fet, connector, sense resistor, and then the bjts and other resistors. Once you have perfected this as a block. I would copy it and manually assign the part id to each of the new parts. Once you do this the little node lines should dissappear to show that it matches your schematic. this should give you a perfect copy of part of your circuit but with different part ids. Do you use source control? I really consider it a must. If you do you can to the diff between your original file and your new file. For doing this sort of copying work, It might actually be easier to edit the pcb as a text file than in the pcb editor. Look into it. It will give you real insights into what the pcb editor does. You are probably having trouble changing track widths because you have DRC on and so all tracks will go to the default track width. In DRC I think you need set track widths for each node. create 3 types of track/node. I use thin, normal and thick. Set each of them to the thickness you want. The assign all nodes to the appropriate type of track. When you go to place a track it will select the correct thickness. The only trouble is making the super thick nodes go into a small pin. Here you temporarily change the thickness of the track type. -> lay the track -> reset back to the normal thickness. Flooding areas was a bit weird, I only recently worked it out, I cant remember off the top of my head though. You can leave it to last anyway. When I get the chance I will have a look. I would use Seeed Studio because they are sooo cheap but the turn around is a bit longer. Just keep each channel as a small perfected block and you should be able to move them around quite easily. I am going long and thin and black for the next spin of my board. I want to make it look a bit Steam Punk, so I can expose part of the board.

Hi

There are foot prints for this type of cap but none that are exact so I wanted to edit the component to make one for my cap, but couldn't workout how to do this in the editor. The lines didn't seem selectable and the couple of pdfs I read didn't seem to help much either?

Firstly, put your Kicad stuff including libs and mods into source control, if you are going to work effectively this is a must, being a programmer you probably already use source control, it is so easy to make a mistake by clicking the wrong button, that rollback will save you and give you clues into how it all works and tell you what has been changed.

Just to start by giving everything standard footprints. Doesn't matter if the footprints are wrong as long as the pins are the same.
Then in pcb editor right click and edit module on the components that need changing. Then on each pin you can edit by right clicking. The pads are normally simple shapes, so you can edit the them by just changing the sizes and offsets in the text boxes.
If you need a complicated pad shape you can make it as two or more separate pins but give them the same pin number. This is a strange but necessary work around and it does work fine.
Work out how to save these modified mods into your own mod library and then associate them.
Did you see a youtube video on Kicad, I found one in particular it quite useful.

the footprints for the caps don't need to be exact, if it roughly good enough it should solder ok, you are better off using a standard footprint because you might change caps anyway.
Also there is a difference on components between the actual footprint and the solder pad needed. the solder pad is normally a bit bigger around the edges of the component so you can stick the point of the soldering iron onto the pad itself.

Am I right in saying all the resistors should 1206 and all the other caps should be 0805?

that's what I did, the caps need to be close to the processor and the resistors are not so critical. I would go with that unless you have a specific reason. The bigger components are easier to work with.

ps. the reason I chose 4.7uF through hole is because I had 100 of them already.

Yes you could change it to 5 volts maximum. I have changed my input protection, to at TVS diode and a fuse. and got rid of the mosfet. You can do the same. If you wanted to make it a sort of 5-12 volt combo that probably would be best. It would give you the option to have more than one led per channel if you want it. I don't think the efficiency loss would be too bad. Just get rid of the 24-12 stuff, then you have virtually the same circuit as me. It may save a bit of testing. Alternatively you can move the fet down to 5 volts. you would need a logic level fet. and you could get rid of the bc547s and associated resistors. Just like AHellenes diagram. If you did this there would be no point in using a linear regulator to get the 5v for the mosfet drain, it would just be a waste of power, but the regulator for the 5volts to the micro would be still necessary. If I was you I think I would go the 5-12 volt combo. I am currently selecting Parts from Digikey for the boards, for the first time I will be able to give you a real BOM and even Digikey part numbers as well as Manufacturers part numbers. Do you have anyway of programming the Attinyx61? I am not sure whether to go with the Dip20 or the Soic20, (Actually I will go Dip) I think I am going with Murata 4 pin torroids for the inductors @ about $1.80 each. The only problem is their height which is about 15mm at a guess. And which connectors for the led sting, I used green phoenix connectors on the Steff shield. Have you got any good 2 way connectors say 3 amp max. Less than 15mm height. You can probably use my boards if you want, with a couple of hand modifications if you go 5-12, I will make sure that you don't have to cut tracks if you want. I will put in a jumper before the black regulator and you can wire that straight to 12 volts.

All my parts will be surface mount where possible. If you want to go dip you will have to find similar.

But the Attiny will be Dip, just so I am sure that I can program it, I have never done an ISP header before and a cock up here would be catastrophic, the big inductors will be dip, and the connectors and fuse holder.

I think the program will be very small, probably the tiny261 will be fine.

I have an avr ISP mk2 will that do the trick?

I assume so but I really don't know.

Those connectors look good (your favourite).

I think I want the screw in onthe flying lead and the L shaped black header pins, probably I will go for 5mm.
I will draw the pcb at 5mm.

Where is the 180mv from? I remember 200mv for the current sense? Also is this squared as 180 or as .18?

Sorry you are right, it is 200mV average, but it turns on at 180mv and off at 220mV.
Lets work in Ohms, volts and amps. forget the millis.

You need 0.285Ohms to give you 0.7A.
the current sense average voltage = 0.2volts

So the current sense power will be V * V / R for each resistor.
It wont be much but we need to check in case you go surface mount.

0.2 * 0.2 / R
worst case is 0.75 ohm

P = 0.2 * 0.2 / 0.75 = 0.053 watts, ok I think you are safe with three resistors.

Does the reisitor voltage matter I found some that are 250mWatt and 200V rated is this ok?

These resistors will have maximum voltage of 220mv on them, your are ok by a factor of 1000. You will be hard pressed to find a resistor whose maximum voltage wouldn't be enough. So don't worry about voltage and use at least 125mWatt resistors.

the other resistor from memory was quite difficult to calculate, but I used 180kOhm

The Steff Shield is working, I am running off 24 volts and have the TVS diodes in.
I haven't damaged one chip. I need more properly heatsinked leds before I can run at > 350mA.

SteffShield

The main thing with the 2uF and 4.7uf caps is that they respond fast, to do this they need minimal inline inductance.
You must use ceramic caps and you should place them as close as possible to the legs of the chip. Surface mount ceramics are even better.

Here you will have competing requirements as almost everything needs to be close.
This is the advantage of surface mount if you are going to route a board.
In order of proximity to the chip I would put
Caps, diode, inductor, sense resistor, other resistors.
the other importan thing is to work out the high current paths and put thick tracks where necessary. I think some of my power tracks are 3-4 mm wide.

You dont need to change the sense values depending on the number of leds, this is the beauty of your current driver.

You just need to be sure that you have enough voltage for the number of leds. ie. Voltage Supply > 3.1(vf) * number of leds.

As I mentioned before I would leave spots for at least 2 and preferably 3 resistors in parrallel for your current sense resistor. You could even make them switchable.
for 700mA you need R = V/I = 200/700 = 285mOhms. I found that even if your resistors are exact your current wont be so this is why it needs to be tweaked.

Also On Semi told me that you need to short pin 1 and 2 externally.

pcb, does it make a difference the run of wire between the leds and the drivers?

I guess the length of the run for the leds has a few issues.

more inductance, but in our case this is a good thing.
more resistance therefore more losses which may not be that important for you
more emi noise generated. the NCL30160 will generate noise in two bands:
one at the hysteric switch frequency, which will vary and be anywhere from 100kHz to 1.4 Mhz, the inductor will be continuous at this frequency so it will not be and issue. the current change will be about +- 25% from 700mA.

The other one is the worry. When you are dimming you have to supply the NCL30160 with a square wave at 0-20khz. This will switch the enable pin of the converter. At this point you are switching 0-700mA at say 20khz. This is switched along the length of the cable run.I
I don't know if this will be a problem, but I can imagine it is not any worse than a lot of other things in your life. To minimise the interferance you could use cat 5 cable as it is twisted pair which will reduced the noise a lot. The cat 5 have quite a decent conductor cross section too. I would say fine for 700mA.
If you went with the cat 5 (at 24awg not 26) at a distance of 3m, or 6m return trip.
your cable resistance is 84.22mOhm/ metre. which equals 505mOhms.
voltage drop for your single led circuit with 24 AWG = 505mOhms * 700mA = 353 mVolts.

Vsense = 180mV
Vfet = 700mA * 42mOhm = 29mV
Vind = 700mA * 36mOhm = 25.2mV
Vled = 3volts?

As a ball park electrical efficiency calc. Your led drive output voltage = Vled + Vind(dc) + Vsense + Vcable + Vfet. = 3587 millivolts.
eff = Vled/Vdrive = 83.6%
Though most of the heat goes into the 6 metres of cable.

alternatively 1mm2 cable at same length is 34.6mVolts.
giving drive voltage of 3.269 volts.
giving efficiency of 91.7%

there is also switching losses but these will be smaller in comparison, and hard to work out.

I think I would try the thicker wire, which will be hard to get as twisted pair, and see if you get any noise problems particularly in your sound equipment.

the other point worth noting is if you have leds in series where possible, the losses are only per string, not per led, this is why I want to run large strings of leds
I will obviously be running up against the same issues in my Led stuff.

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.

Hi,

I have it down to 5 mosfets, which do you think would be best for the job? I have tried to fit your suggestions above:

MTD3055VLCT-ND
MTD3055VLDKR-ND
497-4335-1-ND
497-4335-6-ND
IRLR014PBF-ND

Finally will these do for the inductors? 811-1222-ND or These?

Many thanks,

Andy

I have assembled the board, I had one big mistake, one of my Black regulator transistors was around the wrong way, And I also put the wrong package on for my mosfets.
Luckily they just squeeze in.
the Black regulator is working.
I need to write the firmware to test the actual dirvers, and especially the current sense part. I should be able to get some results this weekend. Things have been a bit hectic lately.
I'll look at the Fets now.

I just checked the channel independence and all is good. No visible effects of one channel effecting another. I did an efficiency check 3 channels at 20.5 volts input and 1.23 amps input gives 25.2 watts. V1 20.27v I1 430mA p1 8.72W (not actually switching, because it is turned fully on) V2 20.3v I2 350mA P2 7.11W (barely switching) V3 17.56v I3 440mA P3 7,73W (only 6 leds in series not 7 that is why it is different) Pout 23.56W efficiency 93.5 % Obviously this figure has some error, but very hard to say what it is, but I think 93.5 % is in the ball park, nothing is dissipating heat bar the leds. Theoretically when the channel is fully on I will get losses in the inductor 40mOhm and the sense resistor 100mOhm and Rds 40mOhm. At 440mA this should give me a voltage drop of 0.0792 volts coming from 24 volts which is 99% - minus the gate drivers. Still I wanted over 90% and I am sure I have that. the light seems very solid now it's been running for hours, cant fault it at the moment.

Hi Andrew Good news I have fixed the init problem by setting the brown out fuse to 4.3 volts. I have three channels current sensing. Controlled by the dimmer. the differential sensing works and so I can just touch the slider and the output will correspondingly change, the output is quite stable. I will do more accurate tests later. All channels are behaving the same way. I have almost no flickering, maybe the tiniest bit at very low light settings. I have tons of scope to tighten the feedback loop, this will be the major factor of any flickering. I have written it in C. But have been told that assembler is better for this job. I am using Amtel Studio 6, which should be familiar if you us a netduino. I will give you the project now but I will probably change things before you are ready to program. the only thing not planned is I am using the 2.56 vref for the dimmer as well as the current sense. there is a good chance it will work with 5v as planned. Also I don't think we need Attiny461a as the code and ram requirements are very small, you can probably save your cents and get the 261a instead. I can give you help in the layout. not that I am an expert I've only done 3 boards but no abject failures yet. Apart from the silkscreen layers. Have you sorted out your tlc5940 stuff yet? NB I haven't tested the 3 channels independently yet, I am worried the current sense may be subject to cross talk, hence the strange track layout.

Hi

I have just been working on it now,

Finally your last point was about 10K resistors and 39pf caps, I can see some 10Ks on the BOM but do I need to add the caps and if so will these do? CC1206JRNPO9BN390

Sorry the cap was 3.9nF. No it is for the black regulator.

I haven't fully tested the boards, only the black regulator.
I have just modified my original code to support the ADC differences, I am using differential inputs on the current sense, and one sided for the dimmers. You will put your TLC5940 square wave into the dimmer inputs, Share the grounds between the chips. the dimmer input must be 0-5v. PWM Square wave is fine, use a frequency just about 20khz or so.

For firmware I am using Amtel Studio 6 and an AVR Dragon, it seems to download ok, if I have any problems I will go back to version 4 which did work.
Last thing I need to do is rescale the current sense, which is just a constant.
Then turn it on and see what happens. I have been really busy but will try to get lights tomorrow.

Have you seen Mikes electric stuff he has a good video on small scale production, there is a link on hanzibals audio project that I posted.

I have found some indutors on ebay that are really reasonable don't know what you think? SMT DS3316P

They are quite good but I personally wouldn't use them as their DC resistance is a bit high at 334mOhms for 330uH. The smaller the package the higher the resistance.
But that may be fine for you as super efficiency is not your main priority. The only way to reduce the resistance for the inductance is to go bigger.
You will have to weigh it up for your situation.

I will also look at connectors as well at some point what did you go for in the end for power, LEDs and control signal?

I just went with cheapie chinese phoenix connectors, for power and leds, but 0.1 inch pitch pcb pins for the dimmer and the ISP header. (both 2x3)

On the BOM I have made a few comments and corrections based on your advice I think I am almost ready to order now but just want to make sure I have got everything on there.

497-5927-2-ND is wrong, use 497-5938-1-ND it is 5v operating voltage tvs.
the little smd inductor is fine use that.
power inductor: see above it is ok but you might want to use bigger.
39pf dont use
the fet is still wrong, see previous post. If you are using 5 volts on the fet gate you need a logic level fet.

I did a search, I think these ones are better. For up to 3 leds us the 30Vds for a longer string use the 60Vds. Either will work at Vds of 5v. And your driver will be fine. I still think 12 volts is better, but I didn't want to confuse things too much. So I didn't really talk about the limitations of logic level fets. Fet Characterisitcs: Logic level fets have a lower Vgs turn on or Vthevinen. This is not the absolute max Vgs rating. Vth is not really a voltage but a curve, The curve is Rds vs Vgs They have the curve in the data sheet. Basically by driving the FET further away from Vth you are closer to an ideal switch. When Rds is either 0 or infinity it is not using power. When Rds is in between at say 1 ohm it is using power. At a higher Vth the fet can turn off faster when Vgs is zero and when Vgs is 10v the fet can turn on more fully. You notice when you turn on, you will be supplying around 5 volts to the gate (Vgs) but the measurements for Rds on the digikey search are measured at 10v. You actually want a low rds at 5v which may not be the same thing, although it should be close because it is a curve. Hope this makes sense, it's a complicated business, don't be scared, driving the fets at 5v will still work but driving the fets at 12v will work more efficiently. Your choice. ps. I also forgot to order the 10k resistors so I had to put down 805 instead of 1206 and I didn't have any 3.9nF caps. So for the first board I had to improvise a bit. I did the reflow in my hacked microwave with a fan heaters guts. It worked pretty well for my first board. I circled the Vds max voltage because you need to allow about a 10 volt margin to what you actually want to run.

I guess you need to say are you making this on vero board or designing your own pcb?

If you go veroboard then you need mostly through hole components.

if you go pcb then surface mount is a good option.

Last night I put together some of my Steffshield project, and I must say it looks really good on the pcb, but I have at least 4 errors but luckily no showstoppers.

I have a pile of things going on right now, but I will give you part numbers for the bits you want.

Remember to give the tracks between high current components short paths, NCL, diode, inductor, sense resistor.
caps need to be right next to the pins.
TVS needs to be betwen the ground and Vin on the only path before the Vin Cap, with very short leads.

Rsense = 200mv/700ma = 290ohms

= 285 milliohms, better get a new calc.
Leave room for 2 resistors here, so you can use parallel to tweak the current.

When i get the chance I will draw something. (within a couple of days)