Wow that board layout was fast. Looks really compact!
Yeah its 100x54x40mm which is less than half the size of the last version.
I am expecting it to do 350A at 16S and 200A at 18S.
Boards shipped today so will get them in 2-3 days.
I ordered a lot of parts from LCSC to try them out. The prices on there are ridiculous, like 4x cheaper than mouser for MLCC. The STM32 chips are $5 instead of £9! Which almost halves the unit cost in components.
Really want to be able to sell these ones for a lot less. As new VESC based controllers are all pretty expensive, there is a decent amount of competition from china now with the likes of FlipSky and Maytech.
Talking probably around $300 if I can get the bom cost down enough. Last time the heatsink was stupidly expensive (£90 each!) and those HSOF8 fets (£3 each) and stacked 2220 MLCC (£2-3 each) are not cheap. With nothing but those three parts it was looking like £200 without anything else!
So I have now switched almost every part over to coming from LCSC and substituted lots of parts for cheaper alternatives. Final BOM cost comes in around £110 including taxes! Plus pcb costs, shipping and assembly, heatsink, thermal pad, bolts, etc. So looks like $300-350 RRP is going to be easily achievable.
Got the cad sorted out now and ordered a new heatsink from protolabs.
Neat new design. The modular construction will help for development and troubleshooting. Also repairs get easier.
Great that the TVS Diodes are still there.I like the test points as well.
But there is no CAN transceiver anymore?
Also will the final design feature a hardware overcurrent protection?
That would be a good one. Especially for the low inductance Motors, the currents like to shoot through the roof.
In addition some pads for low pass hall sensor signal filtering would be great.
Keep up the good work!
There is still CAN, the connector just doesn't have a 3d model.
I decided hardware overcurrent used too much space and was expensive. The solution was to design a board that didn't care what the current was, this will do 1000A spikes at 16S.
I will have a look at the hall sensor pads as that is easy to add.
Boards arrived and my god... ITS SO CUTE!
Looking good! And nice work getting the cost down as well. There is not really anything like it on the market aside Benjamin's one.
Are you still using bus bars?
No bus bars, or any large copper anything. I got fed up having to spend 5min heating boards up to put on bus bars, and once they are on they are almost impossible to take off. Also costs like $15 for the custom shaped ones.
I'm not sure I'd double up the press-fit terminals like that, a single one ought to be good.
That's good for heat capacitance and transfer. Also you can use a more flexible motor wiring. High power inrunners often come width 6 leads (also to configure them in y or delta).
I wonder where the high ohmic gate resistors gone
The Redcube terminals are rated at 180A continuous at 25oC. So needed two for 300+A continuous.
Now using some very good thermal pads and anodized heatsink so I think it should outperform the 75/300. https://docs-emea.rs-online.com/webdocs/1181/0900766b811810f2.pdf
Do you really need that expensive material for heat contact? I'm using "thermal conductive silicone" from aliexpress and it does the job perfectly with ~100W of power at 200A phase current continuous.
Also redcube terminals come in 350A size too. They have M10 screw. https://katalog.we-online.de/em/datasheet/7461100.pdf
What kind of gate drivers are you using? Are they isolated?
Are you using heavier copper? I'll also consider not using busbars.
That's a massive piece of it, works out reasonably priced for the amount I need. To get the higher current in smaller package you need to get the heat out better. At 350-400A its 4 times the heat of 200A. So unless you can keep junction temps down things will overheat rapidly. Max continuous current of the fets is 425A or so, and I want to make full use of that.
The smaller terminals spread the current flow over a wider area and mean the connector and wire is not massive. M10 lugs are huge!.
Using the same TI drivers I have been using as I am happy with them and are simple to use.
I am first trying normal copper thickness to see what the limit is, then will go higher if its required. All the current paths on the pcb are extremely short <2mm, the rest of the time it either goes through cables, mosfet case, shunts or output terminals. I think always trying standard plating thickness before going higher is best bet. All my previous designs only used 1oz/0.5oz on a 4 layer board and it is always the mosfets that go, not the board.
Cool, thanks. Just so you know, that heat conductive material also conducts electricity.
Are you using 100V fets, if so aren't they silicon limited to 124A each. Their package can hold up to 375A. Correct me if I'm wrong.
If you use the 6 terminals for delta connection, then you can connect the current sensors to only 3 of the 6.
Just remember to adjust the FOC code (e.g. use a modified Clarke transform).
Advantage: One can use sensors with half current rating. Lower cost and/or lower losses.
Had just an idea for the thermal interface. It should be possible to solder a IMS circuit board upside down on Top of the Mosfets.
Definitely a cool attempt
Yeah the PGS being conductive means you need another solution I guess - unless I'm missing something?
If you ever make a lower current rated version you could only populate half the redcube connectors. It would still be an amazing choice for those needing <150A?
The PGS material comes in an insulated version which has a PET or PEEK film on it depending on which temperature you need. The thicker pads they make are only conductive in Z plane, so if you anodize the heatsink it stops it conducting. The anodizing also increases the conductivity of the aluminium and radiates more as well, plus it looks cool :D.
@e-biker I must have misread the datasheet, yeah the fets are 124A limited so 372A max, still silly current! But it still depends on getting all that heat out.
@RSR They are just connected in parallel on the board and having the shunts stops high current going on pcb, they are pretty cheap.
@sunsam8 The PGS material actually outperforms copper for thermal conductivity, also that sounds really hard to do in production!
All the parts from LCSC will arrive on Tuesday/Wednesday, heatsink arriving Tuesday. Really excited!
The 70um PGS sheet arrived, had to use a microscope to peel the backing paper off the adhesive since its so thin. The first heatsink I got made has a 0.4mm gap to try out a silicone pad first. Then we can see the difference.
What is the breakdown voltage of this material?
It says 1kv, also has a 2kv version.
Parts have arrived today so will build one tonight! Oooo
Hope that the FETs build an even surface, so that the PGS can fill out the differences. Maybe they can get machined after soldering before PGS application.
Big question for me is, will the benefit of the lower inductance be worth the hassle with directFET soldering and the problematic thermal linkage.
I have not simulated it, nor have the experience, but couldn't deal the a200s powerstage (with minor modifications) quite well with low induction Motor windings?
The directfets are not that hard to solder once you get the hang of it. Just put a bit of solder on each of the pads, plop the fet on top, heat it up with hot air and press it down a bit. The side pads sit flush on the pcb, so you will get even mounting surface for thermal interface.
The comparatively HUGE hsof-8 package is a pain in the ass to work with. You can't cool the case, so you need a big busbar on the back to get the heat away, which takes up loads of space and costs a lot.
You could arrange them in a way that reduces the inductance of DC link, but its still much more efficient from a size and cost standpoint to use directfets.
Heatsink has shipped! Will arrive tomorrow.
Managed to get one together, the caps were a bit too big with the heatshrink on them. Apart from that everything else seems to fit good and quite easy to assemble.
I can't wait to build one of these.
I've found soldering directfets to be just like any other part with pads on the bottom. The magic of surface tension does all the work and keeps the mosfets at similar height. I have found it helpful to tin the pads instead of using paste, to prevent uncoordinated smearing of the paste while placing the fet.
Heatsink arrived. Looks gorgeous!
Take all my money
I got a sheet of the PGS material and was confused why it was conducting. Turns out I bought the wrong one! Going to get this one EYGA091207KV, which has polyamide tape on it. Can get about 4 pieces out of each so its not too bad price wise. The adhesive works well and cutting the pad is very easy to get it the correct shape.
This should help you define cutout the way you want it. You can also draw keepout in any other software then import it as DXF or draw your board in 3D, import the body and then define board shape from it. Boards will look nicer and you won't have to cut corners to fit it in your heatsink.
I had to cut the corners as I designed the heatsink after I ordered the board and forgot to round the corners on the pcb when I ordered them. Was in a hurry to get them ordered before Chinese new year.
As for the outline on the top board, I am just going to remove the sticking out sections between the redcube terminals. Then have the 3d printed case use those holes to mount.
Think I might slow down the switching a bit lol.
Tested today with a motor and its working really well! Detection passed first time and it can spin up the motor with perfect tracking at 3A. Zero rpm crossings are butter smooth as well. Now to test some high current stuff!
Nice job on getting the price down! Very compact, high power controller. Hope you testing goes well
Decided to just use silicone pad. The PGS is great but a bit of a pain to take off again and the insulated version loses quite a lot of thermal conductivity.
You don't have any mosfet gate resistors on it at all?
There is one common high power gate resistor right next to the fets on the power board. I saw the 75/300 has either none or a tiny one right next to the gate driver. We will see during testing if it is going to be a problem.
Aside from lowering gate oscillations, don't you also need dedicated gate resistors to better match turn on and turn off times. Otherwise the fets will be at the mercy of part to part variation.
I think because they turn on so quickly it doesn't matter as much. Power dissipated in a say 10ns difference in turn on time really doesn't do anything. It will be insignificant compared to the power you would lose by say doubling the switching time.
This explains is perfectly. Section 6.
So basically it doesn't matter unless you have bad ringing. Their idea of fast switching is probably like >500kHz.
So tested today on my Revolt 160sh. Set it to 200A motor, 300A abs max. Detects the motor fine and spins it up almost instantly with 150A current mode.
Ran 150A foc_openloop and there are some interesting noises lol. Which might be gate ringing, or the fact I am only using 30v 20A psu at the moment. Will try increasing the gate resistance and see if it makes it sound different, if not then will try some lipos.
Plan for the weekend is to test on my big generator pair and see what its continuous max current is without extra heatsinks or fans.
But so far seems quite happy sitting at 150A for at least 30-40s. With the heatsink getting to maybe 35oC in that time. Not bad for the first prototype and significantly better than V1.3.
Roughness with foc open loop through terminal is probably down to the lack of ramping and open loop jankiness. It always sounded awful on all the vesc hardware and motors I tested it on.
Good to hear it is staying nice and cool. Hopefully that means the gates are doing ok too. I guess the internal gate resistance on each fet is enough?
Yeah it was making weird electrical noises, whines, buzzes, which only got worse as I increased the current. Its probably more to do with how its doing 150+A at very low modulation, so it just sounds strange. There is also probably a lot of ripple on DC bus while its happening too. I need to get the temperature sensors working properly then I will be able to run a longer test. I will record a video of what its doing tonight.
The driver is 0.9Ohm, gate resistor is 1.6Ohm and gate is 1.5Ohm.
Switching is probably so fast it couples to everything. I had somewhat similar problems on my first version of BESC. You'll probably have to slow it down if you want to have other electronic devices near your driver and not have EMI (or to pass emc ). Make a loop at the end of your probe and go around your driver with it to see what couples.