Old PCB houses used to do roller tinned before solder mask, which gave you the crinkly soldermask, and of course if you wave soldered it then all the mask over wide traces would wrinkle and peel off, as the underlaying solder melted. Also with a single 4oz layer on the board you will run into issues with warping, especially if you have a large board with a double sided load, you will have issues with components on the other side getting mechanical stress applied, so that you will have things like ceramic capacitors cracking and getting whiskers growing in them, and things like a BGA breaking loose from the balls, and other large SMD devices also suffering from either trace lifting, or joins cracking. Best solution is to take the big current and split it out to a separate board with all the high current paths on it, and then put a board interconnect to another board, which has all the control electronics on it. Bonus is then that any upgrade is easier, only half the board to redo, and also you go into a 3D volume, so the design overall will be smaller, as now your high current side is on a smaller board, which, even if it is a lot more expensive to buy, you get a lot more boards out of a single panel from the PCB house. Daughter board is a standard cheap 4 layer board from them, as a 4 layer board is almost default, if you do not want 4 layers, they simply etch all the copper away on the inner layers in the stack, and assemble them along with the other 4 layer boards. A bonus for the PCB house in the extra copper they get in the etchant recycling, extra profit.
Those were the gool'ol days! You could get 5mm+ thick coatings under your solder mask! I forgot to mention board warping, yes I've seen that happen with even 1oz copper on large boards where there was no matching large plane on the other side.
@@EEVblog I remember that interview with Vincent Himpe, where he was talking about it at one time, and also the AVX capacitor interview, where it was also talked about as a big problem. Plus that mailbag PSU that exploded because of it having cracked capacitors from stress.
I second to this. High current only with Cabel or separate board as it separates the heat from main board. Us thick multicored cable or multitraces on separate boards as mutlitrace divide curent on separate board thus removing some heating problems. 80A Also one option is copper metal bridges with very thick 15mm * 2-6mm metal chunk in a air delivering the current. Multicored cables best ability is current dividation in a outerlayer of each core so multicored vable can deliver more amps than single core.
ANtoher aspect of usage 80A is I never seen DC sircuit needing 80A let see 5v or even 12v in DC systems. He must be speaking AC motor or similar car or high speed industrial machine type that most likely need AC 80 amps rather than DC.
These days we have CPUs, GPUs and FPGAs dissipating 200-400 W, and pulling all that power in at 0.9-1.2 volts! The PCB currents and current densities must be insane, especially since you have to route all of this into multi-thousand pin BGA packages or LGA sockets, which also have hundreds of multi-GHz differential pairs snaking in and out, maybe even a 864 bit wide DRAM bus.
That's why for those designs, they carry usually a 12V line to a multiphase DC-DC converter that lives just on the periphery of the chip. That way you're only carrying 30-40A tot he 12V and the hundreds of amps to the chip are carried only a few millimetres. At that point too, you're also talking chips with a thousand or more pins, so you're looking at half of them carrying only an amp or so. Though, it wouldn't surprise me if the higher power designs start demanding 20V+ power inputs to lower the currents.
Current through a 1" wide trace will have crazy high current densities approaching even a large .093" circular pin. The transition from the THT pin to the 1" copper trace is most likely a larger problem than with the connector pin. A SMT high current FET will at least have a large surface are connection to the PCB trace unlike a THT part. Solder has .1x the electrical conductivity of copper making it mostly impractical to get high currents using solder added on top of the trace. If the solder is 10x as thick as the copper you might get hale the resistance if you can keep the thickness consistent and controlled. There are also thermal conductivity issues and black body emmisivity issues with solder vs. copper and also important if you are pushing the limits. Thank you for the introduction to Saturn PCB. It is a large step from my OLD Bishop Graphics PCB Design Handbook from the pre-CAD early 80's
It's also worth mentioning at 70 - 80 A you're at the practical limit of most TO style packages. Those bond 'wires' do pop if you push them too hard for too long and that seemly super low 0.004 ohm RDSon banner spec for your affordable MOSFET turns out to be 25 W! And that's before you realize that's the *typical*, your max RDSon spec is 0.007 ohms and with your CMOS or TTL level VGS drive means you're trying to dissipate 50W! Also you have to think about protecting it as well, if it's an external connector you need to start worrying about adding ESD protection and other protections from uninformed users with their 48 V center negative PSUs they just love plugging into everything. I've also seen a couple ultra high power density PCBs have thermal pads and heatsinking on the actual PCB traces which is always funny, I think the most recent example I saw it in was for some traces going to a 12VHPWR connector which does 50 A @ 12 V.
I was wondering what type of package could support 80A. I presume it would be some kind of bolted connector, certainly not something that could be on a PCB. Not to mention the heatsink required.
@@axelBr1Phoenix makes green 200A connectors soldered THD. We used ISOTOP transistors and diodes. They are screwed to the board. And copper plates are sandwiched between components and backside PCB. So DC-currents never flow through the board, only the high frequency components do.
It doesn’t have to be a single device, there could be several in parallel that when they are all in use deliver that high current. Thus not being limited by device packaging limits.
@@stevebabiak6997 I checked infineon (Google's top recommendation) and they have some pretty hefty power MOSFETs; at random I found one that can handle 201A in a TO263-3 package (surface mount). I'm shocked so much power (only 10V) can be handled by such a tiny device!
Yeah datasheets quote high 100s A steady state current but fail to mention it'll first desolder itself before it meets that and you'll need an infinite heatsink
Bus bars are really helpful, as they also serve as heatsinks. 1W per 1 square inch, passively cooled. 4W if actively cooled. 0.5mm copper sheets seem to be a good medium ground, and you can place several air-separated layers as fins. Soldering those bars is a bit of a hassle: you have to effectively heat-up the entire bar to ~200C, so use the largest soldering tip at your disposal.
I agree. A big soldering iron. And solder it first before other SMD. You can make an L-shape. The vertical part will almost double the cooling surface and reduce resistance and reduce footprint size.
More than food for thought "my old friend"... I don't comment a lot, but I just wanted to say thank you for all the knowledge and inspiration you brought to my own designs all over the years. After..., I would say a decade...or two...I still get Eureka moments from your videos that make me push a bit further....Thank you for your global public service :)
Two simple answers and solutions: 1) lay one or more pieces of 12 AWG (2-3mm D) bare solid copper wire on the solder / tin plated trace (bend to follow the trace) and solder it to the trace. Most Plasma Cutter and Welder PCBs are made this way. No CAD, no tooling and almost no cost! Though you will need a high powered soldering iron with a large tip to solder the wire to the PCB. 2) get the largest (>1/4") solder wick braid and solder it to the HASL trace. Not quite as good as 12 AWG solid wire, but you can add more layers if needed. 🙂
I had a board that distributed 30 - 40 Amps. I started with a .5" trace, added a slew of surface mount pads .5" x .5". The board arrived with what looked like a .5" solder tinned trace. I added some bare copper house wiring in #12 AWG and soldered it onto the trace using solder the whole way to add to the conductor area. Maybe I should have gone with #10 AWG. Anyway it was only about 8" or so. I had no issues.
It amazes me, that such a wise and experienced, by now old person whom I respect from the early days of the first episodes says: 12 inches - that's what she said. You are just amazing. Genius!
Part of the problem with switching 70-80 amperes with a MOSFET is the mosfet leads. The leads of a TO-220 melt at about 75 amperes and you also have to take into account the trace width available where the copper meets the device. Additionally just because a MOSFET die is rated for 100 amperes or more does not mean a real MOSFET is going to reliability handle that much current. Generally I would not push a PCB mounted MOSFET past 40 amperes and generally shoot for around only 20 amperes if surge currents are likely. Splitting the current between 2 or more devices allows the transistors to run at lower temperatures and gets the current paths down to a reasonable level.
Something that worked well for me once was square cross section bare copper for jewelry making. It was too springy out of the box so I threw it in the fireplace for a couple of minutes and then let it air cool. Clean it up with steel wool, and then bend it to fit over the PCB trace pretty well. Once one end is tacked down with solder flat on the trace and to the component pin you can tweak the fit as you go. I think I used 3mm copper, a little thinner than the trace. It is fine for DIY one-offs. Solid core wire works but imo is harder to work with and get a neat job. I never tried it but I guess if really desperate copper capillary tube might be good for water cooled traces.
@@scoutjonas Do you know if that works with complex high speed electronics, like computers? Will mineral oil work? I have been wanting to send a homemade ROV into very deep sea water. I figure since oil is incompressible, insulating, and doesn't mix with salt water I could maybe fill it with that and use a rubber diaphragm to equalize the pressure. But a concern is oil getting into electrolytic caps, maybe mixing with the electrolyte and throwing them out of whack enough to matter? At 10,000 psi or so, I imagine the oil will get pushed into solid things too,, like resistors and ceramic caps and the PCB itself. Maybe the oil would even become somewhat chemically reactive at that much pressure. And then how quickly will parts equalize to surrounding pressure. Could be a pain in the eye if oil and gas mix in a component and explode when you bring it up too quickly. I don't have a great understanding of how that might work. And then there are all of those factors that become important in high frequency AC stuff. That's mostly black magic to me at this point.
A 10 degC temp rise is really really small. Most PCBS are going to be sat in an ambient environment below 85 degC, so ime, depending on the duty cycle / duration of that high current you can carry far far higher than 20a on a 25mm trace! One interesting experiment is to try to actually BLOW a trace. Find an old pcb, rig up a grunty power supply and try to blow some traces. You'll be surprised how much current and for how long you have to apply that current to cause the trace to delam, lift then fail. This is because the pcb has some thermal inertia. Heat produced does not immediately lead to a proportiojnally direct temperature rise because that heat is sunk away into the pcb. This is why using an IMS (Interal Metal Substrate) pcb really really works well, and could be a very good solution to this problem. The trace generates heat, but that heat is sunk into the metal internal layer of the pcb. Add a suitable method of extracting that heat (passive or active heat sink) and you can run HUGE currents on small(ish) traces. I have a 3 phase IMS inverter board that runs up to 1,000 amps and the entire inverter fits in a 100mm square area!!
Exactly. The 10°C maximum temperature rise is absurd for such an application. Obviously even a big ass trace is not going to handle a lot of current then, since the heat transfer coefficient for passive cooling of that surface is somewhere around 10 W/(m²*K).
Also fun when folks realize how often PCB's are used as heaters on purpose. Most consumer 3DPrinter build plate heaters are just PCB's with a squiggly heater trace...say, 20A@12V for a ~50'C rise for days/years on end.
In my experience in most high current PCB cases you just have to accept +50°C above ambient, thick copper, and the fact that it should probably be a separate board with limited circuitry and size to prevent warping and even then you will need some extra tricks to really push it to the limit. The reason for that is that if you even want such a current on a PCB it alredy means that it will most likely need to be a part of some circuit that will control the current (why even have that on a PCB otherwise). And since the circuit will probably require at least some protection you will have to roughly measure the current and since any shunts of such caliber are either huge and not PCB friendly or stupidly expensive or both you will want to use your trace as a shunt resistor. But with the decision now you actually need to have at least a few tens of millivolts of voltage drop across your trace as with high currents generally comes high noise of all kinds and it automatically puts you into a contradiction when you end up actually requiring to dissipate the power in your trace for the circuit to even work as intended. I have managed to push the compromise to 40A in one of my devices but I had the side of the single side populated board where the trace was covered by a thermal conductive sheet and screwed to a radiator or otherwise the whole thing would just get too hot. Basically it can be done but it may not be quite as easy as one might expect. It is all a huge compromise and depending on the parameters required you may end up with no suitable solutions available at all.
I was involved in the production of a new product that could run on line power or internal gelled lead acid batteries and that had a built in charger for the batteries. The circuit trace for routing charging power to the battery pack was protected by a 40 A fast blowing fuse. The trace itself was of 4 ounce copper and a single sided board. The trace ran along the edge, then turned a corner at the next edge of the board. There was a mounting screw near the corner so the trace tapered down then expanded back where it turned that corner. Barely over a quarter inch wide at the narrowest. Pretty loud pop when it blew up. Fuse was, of course, intact. For a temporary fix we soldered heavy copper braid on top of the trace of all the charging boards.
At Ericsson we always added an analog overcurrent limitation circuit to make sure the fuse did not blow at overcurrents or short-circuits. The ceramic fuses were soldered to the PCB. If they blew you would have downtime and rework cost. Usually the faults cleared themselves. And the overcurrent protection would reset. Fuses were there only to comply with external regulations.
Good explanation! I have designed powerconverters from 5W to 5kW, all of them are mostly on PCBs. When routing 80A for 30mm length between components, these formulas doesn't work. The THD-transistor will wick heat from the trace into the heatsink so double layer is not needed. But SMD transistors will add up to 2W losses into the trace. If you use the power trace as a heatsink you need to be conservative of trace internal temp increase. If 80A is just a transport of current, a homemade bussbar is the solution. I have used 2mm copper plate, screwed or soldered, to the back of PCBs. It was double sided copper PCB with SMD components on top side. Don't short any vias 😂. 525Adc switching converter.
You did not discuss how the current was going to flow in to or out of a pin from the MOSFET device into the trace and out to the final load. A 0.062" pad hole also will strain to accept or deliver that much current. Also it would have been nice to have seen a silver plating result vs solder plated result. .. Thanks for this practical application example video.
Clicking on 'DC' instead of 1 Mhz for the trace frequency which was forced ON by default for the older standard may also approximately double the current.
Back when I worked in science, the electronics engineer of the group panicked about traces being to thin (Horowitz-Hill stated they would be fine...) and solderedcopper wire to the pins connected to those traces.... nailed the Opamp -15V to the digital +5V.... PSU war ensued, killing all opamps with a 5-0-15 supply until the 5V PSU gave up, then the -15V kicked back in to start a genocide on the digital electronics.....
Hahahaha. Great! I had two 12V to 3.3V buck converters. Redundant power on two different boards. Before the two 3.3V were joined together they went through two "True reverse current blocking" IC circuits. When the two power PCBs were connected for the first time, one buck supplied power, the other consumed power, boosting the 3.3V up in reverse until the 12V rail reached 22V and the buck IC burst into flames. Don't trust True reverse current blocking IC's 😅 Later, according to datasheet, it blocks reverse current above 1A!! Why did the buck not have overvoltage protection on the input to save itself from itself.
very handy talk ! and here was me worrying about my 3A power supply layout when most of the main output track is at least 7.62mm wide on both sides of the board.. somehow I think it'll be fine :)
Just for information, 1 oz copper is the measure of the weight of a square foot of the material. If you buy copper sheet which weighs more (used in construction) they use the same measurement methods
My top 3 hates as an electronics repair engineer: Fans blowing shit all over PCB leading to corrosion and tracking. Conformal coating literally not solving this issue and making PCBs almost impossible to work on. electrolyte corrosion.
I've done >150A on a multilayer board. You wont get reasonable trace widths with an IPC calculator because they assume only natural convective heat dissipation. Trick is to bond the board to a heatsink with lots of thermal vias and quality TIM. If you have transistors connected to heatsinks, this will also pull heat from the traces. This makes the current handling problem more of a thermal management issue. Couple 10s watts in trace dissipation seems like a lot but your board could be pushing multi kW at these currents!
IMO the best approach would be ti get some square copper rods/wire, cut them to size, place them on the bare pcb trace just like any other smd part trace and solder them in place.
something I haven't seen mentioned is the duty cycle.. handling 70A for a few milliseconds here and there and averaging 20A is different than handling 70A full time. I don't know the application of this, but I'd look at taking the MOSFET off the board.. and having separate connectors to it. As for adding wire to the trace, remember that old solder wick you clipped off? soldering that to the exposed trace can give you a 20A continuous trace for cheap if you don't need millions
I use Electrodoc pro in my phone to calculate needed trace widths. It doesn’t have too many parameters for the calculation, but for my needs it is perfect. Has also many other useful features.
Thanks. Highly informative. I am looking at doing up to 40A on a 2 layer board. 1ounce. I was thinking of making approx 7mm traces, leaving the solder mask off, then soldering 8mm2 wire along the top. Essentially it's the bus bar idea. should work for small runs.
Don't add solder mask on the trace and throw in two, three or even 4 heavier gauge wires like 19 or so with a lot of solder after the PCB has been manufactured. Obviously it will cause warping but so will be the case with any other method for that amount of current on a single layer board.
Never saw you try the different PCB materials. (Substrate Options) Curious how much difference different board materials would make regarding heat dissipation.
I've handled this before by leaving mask off of the tracks and then hand-soldering along them to make nice thick tracks. Obviously doesn't scale to high volume production, but for a few specialised boards it works well.
Look up "heavy copper PCB". There are some manufacturers that can create traces that are 20oz or thicker. They can do multiple trace thicknesses on the same layer. It's certainly not cheap though.
Just add more copper from outside(copper wire) with layer of tin(solder) to make good thermal conductivity and heavy current carrying capability and you can do it with just one layer pcb.
1. Update your Saturn PCB calculator. There is already a version 8.31 or higher available. 2. Plating thickness matters if you have at least a two-sided rigid pcb with vias. It is not a thickness of ENIG, HASL or other finishes. It is an additional copper plated onto the pcb during manufacturing process. A minimum guaranteed amount of copper plated onto the PCB for vias plating depends on the IPC class it is manufactured. Many fabs plate about 20um of copper (IPC class 2 as far as I remember). Pay attention to "total copper thickness" below as it changes when you use plating thickness modifier. Total thickness greatly changes differential impedance, for example.
That is pretty informative. I knew about some stuff. But I never considered that having just some copper (plane preferably) on another side, helps. It is obvious tho, and interesting approach. I was designing a smart relay board, with some power metering, and wanted to support 16/25A. Doing 10A was easy. But going into 16-25A, was really tricky, due to compactness of a design, and also logic / digital circuits (so could not use super thick layers). A lot of trickery was also around current sensing resistor (I wanted to have resistor shunt, not magnetic sensors or anything like that), voltage isolation distance, and connectors. In many cases, I just went with copper fills, instead normal tracks, to give it as much thermal dissipation as possible, and removed most of the thermal reliefs (which are there really only for assembly).
I needed to build a laser power supply that could supply up to 50A each into two channels (~2v). Board was all surface mount and I couldn't really go beyond 2oz copper (hard enough with 2oz and .5mm pins on the controller IC). I via stitched pretty large planes on the front and back of the board and bonded the back of the board via a non-conductive thermal pad to a heat sink that uses forced air. Works pretty well, but the board is hardly ever driven to full capacity and there are on-board temperature sensors that shut the thing down if it starts heating up too much. The distances from MOSFETs to power connectors were very small which helped. It was a hobby project, so if dies its only me who's impacted (but big impact as it's connected to a pair of very pricey fiber coupled lasers diodes).
Super useful and interesting, especially that cool Saturn PCB Toolkit. Do have a question, though. I'm guessing that all the calculations are done for continuous current handling? I'm sure it's always best to be more cautious with high current designs, but is there a case for going with slightly underrated traces if the currents won't always be at peak draw?
I have done a big professional experiment for this, but for EV-motor-cables. I ran max continuous current until stable temp. After cooling down I doubled the current and documented the time it took to reach the same temperature. After redoing for different wages and current ratios and interpolations, I got a table of pulse capabilities for this cable type. I learned that the radox rubber insulation can absorb and store a lot of heat energy. More than copper. Also that 30 minutes can be regarded as constant load for cables.
Oh dang, that's a lot of work, but I can see how that information could end up saving a lot of money, and possibly weight if you can ensure the application is within spec of your findings. @@scoutjonas
Interesting read by Doug Brooks and Johannes Adam called "PCB Design Guide to Via and Trace Currents and Temperatures" covers a lot of this in detail with experimental data.
The thickest copper I have ever seen on a PCB was 20oz, it made the PCB heavy and very expensive. Copper bus bars would make more sense in an application that needs that kind of current.
Cooling a board with a fan will double current capabillity in top layer. You can also use gappads to wick heat from power traces into an aluminium case or cooling plate or heatsink.
If you consider some PTH connectors for power arrival and departure, and if you have multilayer board, how do you deal the fact that, with time, some internal plated holes (vias, connector pins) could internally break and then current will only go through one layer instead of many, dangereously increasing temperature? (Presence of plastic connectors, fire hazards..)
The last Class D amp 4 x 120w RMS, power rails, were foreseen with serious traces copper, with on top a full bar copper on it. So, hands off there. And they were extremely short. Good I just had to replace 1 end stage and change the smaller caps on the SMPS. Dangerous toys :)
Isn't the copper plating different from the finish? For example, when you have a 3oz/2oz board, wouldn't they start with 2oz base copper and then plate the through holes such that the external layer thickness is 3oz?
It depends. Some chinese manufacturers try and save money by using a 2oz board and plating 25um on it which they then call "3oz finished copper". In reality this is only about 90um. Just make sure you want 3oz base copper. You will end up with about 125um total thickness after plating that way.
Usually if i am dealing with a single or few high current signals, i just use a high power 0-ohm shunt to pass the current through the board in a few jumps.
At the end of the day, you still have the get the current in/out of the board and for most high power FETs, Relays, Switches etc..., this winds up being a heavy duty pad of some sort. Ohms per square applies here. QFNs and TO packages use large pads while QFPs etc...or other packages will use multiple IOs. The PCB on fire is highly unlikely...I have seen many burnt PCBs and never one that caught fire...The FR4 materials etc..are designed not to combust.
What if you ran a bare copper trace, no solder mask, then soldered plain old square/rectangular copper stock to it? In effect, a surface mount bus bar?
I used to work for a company that had a completely finished board that was failing. they had run a real hot trace through the middle of a multilayer board and they couldn't figure out why the boards were failing. i may have missed it but you did not point out that when you have a lots heat in a trace it comes back out through the parts and the parts fail. the trace itself won't but the parts will. I went back and looked up what you did mention in the old documents and they assumed one layer one side of a board only. Once you do a multilayer board the thermal conductivity changes especially if you've got hot traces adjoining each other in several layers where the heat can go nowhere even in the PCB plastic. One of the solutions is just to go out and run a wire from one point to the other and take the heat in the wire. however if you're in a multi trace board this is not always doable.
@Dave How about surface between THT component leg and PCB layer? You can make trace wider but apart from thicker trace, component leg surface in contact with trace will not change. Sure tin will make some additional surface above board.
80 Amps continuous for low frequency / DC switching is not an easy task in electronics packaging / electronic subassemblies. Also, the circuit voltage was not specified which makes a huge difference in device selection and thus impacts device packaging and the overall packaging approach. It is not just the routing of current but also the power dissipation of the switch which will determine a packaging approach. For example, if the MOSFET had 15mOhm on resistance, at 80 Amps the device power dissipation is 96 Watts. For 96 watts you would want to spread out the heat over several parallel devices, especially at higher voltages. Fortunately, for sharing we have the newer Sic MOSFETs. When you get someone with Dave's background you will also see an experienced eye for "does it look like it would work" first approximation. In this respect you might see perhaps 4 parallel Sic MOSFETs on a heat sink having TO-247 packages. With the current to and from the switching circuit using 4AWG wire with crimp wire lugs to PCB soldered lug connectors. Which means the PCB high current traces lengths are minimized (perhaps at most 750 mils from the PCB lug) with input routed on the top layer and output routed on the bottom layer while maintaining minimal loop area. The PCB cooling to a great deal local to the switches is in large part due to the huge wire bolted to the PCB connections. The power devices are mounted to a heatsink. Essentially, this allows you to use a common power package (TO-247) without using an expensive power module which has screw terminals, and supports a PCB approach using 2 oz copper which would support a microcontroller QFP package. One rule of thumb is when using a PCB at high power is to only route on the outer layers, even though the board may have 4 or 6 layers. Localized hotspots within a PCB can lead to the PCB overheating (catching fire). If the board has internal layers normally they are not used an are omitted from the high current area of the board. High current vias through plane layers is not recommended and if a short occurs can create a fire hazard. So, no internal planes in the high current region. For a poor mans bus bar and with a soldered wire connection, good to about 30 amps at 16 Volts, I have specified a stripped stranded copper 10 AWG wire (tin plated) which is soldered full length on top of a long tinned trace to increase the current of the circuit. The wire acts like a bus bar. But, I only did this during the prototyping phase. In production you would have a bus bar and PCB lug connection. Some times the lug is part of the bus bar at its terminating end. However, custom bus bars are expensive. For very compact high power amplifiers (like three phase CNC spindle motor drives that can reach up to 60HP) where you can justify the use of custom bus bars, the design becomes a very interesting 3D exercise, this is because bus bars can route above a PCB when necessary and form conceptually another routing layer in space above or below a PCB. Bus bars can be clamped together using insulators, then secured to screw connections on power modules or PCB lug screw connections. Custom bus bars are usually made out of copper and are tinned plated. But, you can also have them silver plated, or other plating combinations. If you design your own bus bars that are large (5 to 8mm thick) you can make a tab off of a ground bus bar with a hole to connect a banana plug to or clamp on with an alligator clamp for your DMM.
2 AWG copper flex FTW. Should be able to comfortably handle 100A. I'm going to assume that in an application like this space is no object. You'll need a lot of room to get those cables around.
It was literally my first thought - why not a buss bar or just thick ass wire or sheet cutout. So spoiled we are nowadays, my tube amp doesn't even have a board in neither the low nor the high power circuits. It is all built elegantly right into the chassis with wires or component to component joints, and has survived many decades of operating on top of a rumbling speakers cabinet, that's way way above the standards to which modern equipment is built. Not to mention all parts that wear off are conveniently socketed for trivial replacement.
Using flar busbars or copper plates sandwiched to the PCB is better than using a insulated wire. The insulation is usually max 70dgC and the isulation will reduce cooling of the wire. A bussbar can be 100dgC and doule cooling. Its also better for vibrations.
If I were the author of the question, I would first look at how this is done in welding inverters. However, it would be much better to ask the question of the correct choice of connectors.
Depending on a number of conditions, soldermask can even be beneficial, as emissivity of bare copper (or any other shiny metal) is very low. If dissipation has a sizeable radiative component, e.g. still air, high temperature, the minute thermal resistance added by the soldermask is negligible with respect to the improved emissivity.
To the guy who asked the question to dave. 80 amps is no big deal. Everyone is freaking out about it. I would suggest you start looking into ESC's (Electronic Speed Control ) for RC. Those things hand handle hundreds of amps.
Funny I was looking at hall effect sensors such as ACS772 that can do 200 amps and wondering how do you even make a PCB to handle that, and this video popped up. I guess for those you would use a bus bar with lug connections. Even soldering wire down becomes a little gnarly as you're into like 2/0 wire at that point which is super thick. I'd almost be tempted to skip the PCB altogether and try to solder the wires directly on the chip but that might be challenging to do as well.
Thin copper plate busbar without insulation and with some forced air can handle 10x the current in a copper cable of similar size. Improved cooling doesn't reduce voltage drop, but for short distances it is ok to push high current density.
If you look at some welding power supplies, you can find some really heavy currents, like 150 or 250 amperes, or even more. Almost certainly you find 4 or 6 oz copper, on two sides and a large number of vias between those two layers. Passing the current on two sides and a large number of vias benefits you with balancing thermal warp and also locking the two copper layers to the FR4, which has probably even higher thermal expansion rate than copper. But adding other metals like tin don't benefit you much, because they have much lower conductivity than copper. I assume that you don't even think of using silver (which is about the only material better than copper).
Some of the easiest and cheapest ways is to use "current shunts" that are just the bare copper or steel wire. High current low resistance you can jump components and lanes with it and they are cheap as hell.
Is there a place where we can order busbars like we would a PCB? Just upload some design files and get a quote? Or is it not feasible for DIY/prototypes/small runs and stuff like that? 🤔
80 A - sounds like bodge wire o'clock. edit - ah - commented prematurely, I see you make that point towards the end . . . Wires might also make sense if the mosfets themselves have to dissipate a lot of heat as you can mount them wherever you like on the enclosure and use that for heat dissipation. Nothing beats the classic look of some externally mounted TO-3s.
It seems sketchy as hell that ( 16:11 ) moving the conductor layer to be an internal layer had zero effect, but having a plane present ( 17:10 ) had such a huge effect. Even if this calculator is just a rule-of-thumb thing, that fact that is has no problem with wrapping your heatsink with fibreglass makes it totally untrustworthy? Also, skin effect has a huge impact on current carrying capacity of a wire. I sincerely hope the only reason it made no difference to this particular calculation was because it was truly negligible on a 1 inch x 35 um trace at 1MHz, as opposed to another thing the calculator is incorrectly ignoring.
i have a couple SK 10123/2350 beasts rated at 50A....T0-3 cases with 1/16" thick B and E pins......googled that ID and got nothing, tried Mouser and it seems that it is now an Onsemi part, MJ11033G....
Old PCB houses used to do roller tinned before solder mask, which gave you the crinkly soldermask, and of course if you wave soldered it then all the mask over wide traces would wrinkle and peel off, as the underlaying solder melted.
Also with a single 4oz layer on the board you will run into issues with warping, especially if you have a large board with a double sided load, you will have issues with components on the other side getting mechanical stress applied, so that you will have things like ceramic capacitors cracking and getting whiskers growing in them, and things like a BGA breaking loose from the balls, and other large SMD devices also suffering from either trace lifting, or joins cracking. Best solution is to take the big current and split it out to a separate board with all the high current paths on it, and then put a board interconnect to another board, which has all the control electronics on it. Bonus is then that any upgrade is easier, only half the board to redo, and also you go into a 3D volume, so the design overall will be smaller, as now your high current side is on a smaller board, which, even if it is a lot more expensive to buy, you get a lot more boards out of a single panel from the PCB house. Daughter board is a standard cheap 4 layer board from them, as a 4 layer board is almost default, if you do not want 4 layers, they simply etch all the copper away on the inner layers in the stack, and assemble them along with the other 4 layer boards. A bonus for the PCB house in the extra copper they get in the etchant recycling, extra profit.
Those were the gool'ol days! You could get 5mm+ thick coatings under your solder mask! I forgot to mention board warping, yes I've seen that happen with even 1oz copper on large boards where there was no matching large plane on the other side.
@@EEVblog I remember that interview with Vincent Himpe, where he was talking about it at one time, and also the AVX capacitor interview, where it was also talked about as a big problem. Plus that mailbag PSU that exploded because of it having cracked capacitors from stress.
Seen a lot of that crap in my repair practice, haha! I really like the daugterboard idea.
I second to this. High current only with Cabel or separate board as it separates the heat from main board. Us thick multicored cable or multitraces on separate boards as mutlitrace divide curent on separate board thus removing some heating problems. 80A Also one option is copper metal bridges with very thick 15mm * 2-6mm metal chunk in a air delivering the current. Multicored cables best ability is current dividation in a outerlayer of each core so multicored vable can deliver more amps than single core.
ANtoher aspect of usage 80A is I never seen DC sircuit needing 80A let see 5v or even 12v in DC systems. He must be speaking AC motor or similar car or high speed industrial machine type that most likely need AC 80 amps rather than DC.
These days we have CPUs, GPUs and FPGAs dissipating 200-400 W, and pulling all that power in at 0.9-1.2 volts! The PCB currents and current densities must be insane, especially since you have to route all of this into multi-thousand pin BGA packages or LGA sockets, which also have hundreds of multi-GHz differential pairs snaking in and out, maybe even a 864 bit wide DRAM bus.
And voltage spikes at switching speeds!
Yeah but the distances are very short!
That's part of why multi- phase converters are so useful: partly to spread out the current on the board.
ya man, signal integrity is a huge concern these days for this very reason.
That's why for those designs, they carry usually a 12V line to a multiphase DC-DC converter that lives just on the periphery of the chip. That way you're only carrying 30-40A tot he 12V and the hundreds of amps to the chip are carried only a few millimetres. At that point too, you're also talking chips with a thousand or more pins, so you're looking at half of them carrying only an amp or so. Though, it wouldn't surprise me if the higher power designs start demanding 20V+ power inputs to lower the currents.
Current through a 1" wide trace will have crazy high current densities approaching even a large .093" circular pin. The transition from the THT pin to the 1" copper trace is most likely a larger problem than with the connector pin. A SMT high current FET will at least have a large surface are connection to the PCB trace unlike a THT part.
Solder has .1x the electrical conductivity of copper making it mostly impractical to get high currents using solder added on top of the trace. If the solder is 10x as thick as the copper you might get hale the resistance if you can keep the thickness consistent and controlled. There are also thermal conductivity issues and black body emmisivity issues with solder vs. copper and also important if you are pushing the limits.
Thank you for the introduction to Saturn PCB. It is a large step from my OLD Bishop Graphics PCB Design Handbook from the pre-CAD early 80's
I really like this kind of content because it mixes theory and experience in a way that is directly applicable to solving engineering problems.
It's also worth mentioning at 70 - 80 A you're at the practical limit of most TO style packages. Those bond 'wires' do pop if you push them too hard for too long and that seemly super low 0.004 ohm RDSon banner spec for your affordable MOSFET turns out to be 25 W! And that's before you realize that's the *typical*, your max RDSon spec is 0.007 ohms and with your CMOS or TTL level VGS drive means you're trying to dissipate 50W!
Also you have to think about protecting it as well, if it's an external connector you need to start worrying about adding ESD protection and other protections from uninformed users with their 48 V center negative PSUs they just love plugging into everything.
I've also seen a couple ultra high power density PCBs have thermal pads and heatsinking on the actual PCB traces which is always funny, I think the most recent example I saw it in was for some traces going to a 12VHPWR connector which does 50 A @ 12 V.
I was wondering what type of package could support 80A. I presume it would be some kind of bolted connector, certainly not something that could be on a PCB. Not to mention the heatsink required.
@@axelBr1Phoenix makes green 200A connectors soldered THD.
We used ISOTOP transistors and diodes. They are screwed to the board. And copper plates are sandwiched between components and backside PCB. So DC-currents never flow through the board, only the high frequency components do.
It doesn’t have to be a single device, there could be several in parallel that when they are all in use deliver that high current. Thus not being limited by device packaging limits.
@@stevebabiak6997 I checked infineon (Google's top recommendation) and they have some pretty hefty power MOSFETs; at random I found one that can handle 201A in a TO263-3 package (surface mount). I'm shocked so much power (only 10V) can be handled by such a tiny device!
Yeah datasheets quote high 100s A steady state current but fail to mention it'll first desolder itself before it meets that and you'll need an infinite heatsink
Bus bars are really helpful, as they also serve as heatsinks. 1W per 1 square inch, passively cooled. 4W if actively cooled. 0.5mm copper sheets seem to be a good medium ground, and you can place several air-separated layers as fins. Soldering those bars is a bit of a hassle: you have to effectively heat-up the entire bar to ~200C, so use the largest soldering tip at your disposal.
I agree. A big soldering iron. And solder it first before other SMD.
You can make an L-shape. The vertical part will almost double the cooling surface and reduce resistance and reduce footprint size.
More than food for thought "my old friend"... I don't comment a lot, but I just wanted to say thank you for all the knowledge and inspiration you brought to my own designs all over the years. After..., I would say a decade...or two...I still get Eureka moments from your videos that make me push a bit further....Thank you for your global public service :)
Robert Feranec did some absolutely excellent videos on this subject btw!
Lol i felt the flex when you showed that big bus bar :) , such great info as usual! ✨✨✨
dont forget that the high current also needs a return path
Two simple answers and solutions: 1) lay one or more pieces of 12 AWG (2-3mm D) bare solid copper wire on the solder / tin plated trace (bend to follow the trace) and solder it to the trace. Most Plasma Cutter and Welder PCBs are made this way. No CAD, no tooling and almost no cost! Though you will need a high powered soldering iron with a large tip to solder the wire to the PCB.
2) get the largest (>1/4") solder wick braid and solder it to the HASL trace. Not quite as good as 12 AWG solid wire, but you can add more layers if needed. 🙂
Good on ya, for this video. I enjoyed it. Also love the white board tutorials.
I had a board that distributed 30 - 40 Amps. I started with a .5" trace, added a slew of surface mount pads .5" x .5". The board arrived with what looked like a .5" solder tinned trace. I added some bare copper house wiring in #12 AWG and soldered it onto the trace using solder the whole way to add to the conductor area. Maybe I should have gone with #10 AWG. Anyway it was only about 8" or so. I had no issues.
It amazes me, that such a wise and experienced, by now old person whom I respect from the early days of the first episodes says: 12 inches - that's what she said.
You are just amazing. Genius!
Ken Wood (Saturn PCB), who wrote the calculator, does excellent work very quickly! I have used him for several PCB layouts.
Thank you Ken for a really great useful tool!
Part of the problem with switching 70-80 amperes with a MOSFET is the mosfet leads. The leads of a TO-220 melt at about 75 amperes and you also have to take into account the trace width available where the copper meets the device. Additionally just because a MOSFET die is rated for 100 amperes or more does not mean a real MOSFET is going to reliability handle that much current. Generally I would not push a PCB mounted MOSFET past 40 amperes and generally shoot for around only 20 amperes if surge currents are likely. Splitting the current between 2 or more devices allows the transistors to run at lower temperatures and gets the current paths down to a reasonable level.
Hi Dave, your channel is my number one choice when it comes to electronics, keep it up and greetings from Germany
Good info. I learned more about PCB copper thicknesses by watching this video, thank you. I often run wires on the veroboard for my amateur projects.
Fascinating content as always, Dave. I came here from Twitter after seeing those beautiful copper bus bars.
Something that worked well for me once was square cross section bare copper for jewelry making. It was too springy out of the box so I threw it in the fireplace for a couple of minutes and then let it air cool. Clean it up with steel wool, and then bend it to fit over the PCB trace pretty well. Once one end is tacked down with solder flat on the trace and to the component pin you can tweak the fit as you go. I think I used 3mm copper, a little thinner than the trace. It is fine for DIY one-offs.
Solid core wire works but imo is harder to work with and get a neat job.
I never tried it but I guess if really desperate copper capillary tube might be good for water cooled traces.
Or stuck the PCB in oil.
@@scoutjonas Do you know if that works with complex high speed electronics, like computers? Will mineral oil work? I have been wanting to send a homemade ROV into very deep sea water. I figure since oil is incompressible, insulating, and doesn't mix with salt water I could maybe fill it with that and use a rubber diaphragm to equalize the pressure. But a concern is oil getting into electrolytic caps, maybe mixing with the electrolyte and throwing them out of whack enough to matter?
At 10,000 psi or so, I imagine the oil will get pushed into solid things too,, like resistors and ceramic caps and the PCB itself. Maybe the oil would even become somewhat chemically reactive at that much pressure.
And then how quickly will parts equalize to surrounding pressure. Could be a pain in the eye if oil and gas mix in a component and explode when you bring it up too quickly. I don't have a great understanding of how that might work.
And then there are all of those factors that become important in high frequency AC stuff. That's mostly black magic to me at this point.
A 10 degC temp rise is really really small. Most PCBS are going to be sat in an ambient environment below 85 degC, so ime, depending on the duty cycle / duration of that high current you can carry far far higher than 20a on a 25mm trace!
One interesting experiment is to try to actually BLOW a trace. Find an old pcb, rig up a grunty power supply and try to blow some traces. You'll be surprised how much current and for how long you have to apply that current to cause the trace to delam, lift then fail. This is because the pcb has some thermal inertia. Heat produced does not immediately lead to a proportiojnally direct temperature rise because that heat is sunk away into the pcb.
This is why using an IMS (Interal Metal Substrate) pcb really really works well, and could be a very good solution to this problem. The trace generates heat, but that heat is sunk into the metal internal layer of the pcb. Add a suitable method of extracting that heat (passive or active heat sink) and you can run HUGE currents on small(ish) traces. I have a 3 phase IMS inverter board that runs up to 1,000 amps and the entire inverter fits in a 100mm square area!!
Exactly. The 10°C maximum temperature rise is absurd for such an application. Obviously even a big ass trace is not going to handle a lot of current then, since the heat transfer coefficient for passive cooling of that surface is somewhere around 10 W/(m²*K).
Also fun when folks realize how often PCB's are used as heaters on purpose. Most consumer 3DPrinter build plate heaters are just PCB's with a squiggly heater trace...say, 20A@12V for a ~50'C rise for days/years on end.
In my experience in most high current PCB cases you just have to accept +50°C above ambient, thick copper, and the fact that it should probably be a separate board with limited circuitry and size to prevent warping and even then you will need some extra tricks to really push it to the limit.
The reason for that is that if you even want such a current on a PCB it alredy means that it will most likely need to be a part of some circuit that will control the current (why even have that on a PCB otherwise). And since the circuit will probably require at least some protection you will have to roughly measure the current and since any shunts of such caliber are either huge and not PCB friendly or stupidly expensive or both you will want to use your trace as a shunt resistor. But with the decision now you actually need to have at least a few tens of millivolts of voltage drop across your trace as with high currents generally comes high noise of all kinds and it automatically puts you into a contradiction when you end up actually requiring to dissipate the power in your trace for the circuit to even work as intended. I have managed to push the compromise to 40A in one of my devices but I had the side of the single side populated board where the trace was covered by a thermal conductive sheet and screwed to a radiator or otherwise the whole thing would just get too hot.
Basically it can be done but it may not be quite as easy as one might expect. It is all a huge compromise and depending on the parameters required you may end up with no suitable solutions available at all.
I was involved in the production of a new product that could run on line power or internal gelled lead acid batteries and that had a built in charger for the batteries. The circuit trace for routing charging power to the battery pack was protected by a 40 A fast blowing fuse. The trace itself was of 4 ounce copper and a single sided board. The trace ran along the edge, then turned a corner at the next edge of the board. There was a mounting screw near the corner so the trace tapered down then expanded back where it turned that corner. Barely over a quarter inch wide at the narrowest. Pretty loud pop when it blew up. Fuse was, of course, intact. For a temporary fix we soldered heavy copper braid on top of the trace of all the charging boards.
Inadvertently created a PCB fuse?
At Ericsson we always added an analog overcurrent limitation circuit to make sure the fuse did not blow at overcurrents or short-circuits. The ceramic fuses were soldered to the PCB. If they blew you would have downtime and rework cost. Usually the faults cleared themselves. And the overcurrent protection would reset.
Fuses were there only to comply with external regulations.
Good explanation!
I have designed powerconverters from 5W to 5kW, all of them are mostly on PCBs. When routing 80A for 30mm length between components, these formulas doesn't work. The THD-transistor will wick heat from the trace into the heatsink so double layer is not needed. But SMD transistors will add up to 2W losses into the trace. If you use the power trace as a heatsink you need to be conservative of trace internal temp increase.
If 80A is just a transport of current, a homemade bussbar is the solution. I have used 2mm copper plate, screwed or soldered, to the back of PCBs. It was double sided copper PCB with SMD components on top side. Don't short any vias 😂.
525Adc switching converter.
Good as video!! Love the nerdy stuff like this
You did not discuss how the current was going to flow in to or out of a pin from the MOSFET device into the trace and out to the final load. A 0.062" pad hole also will strain to accept or deliver that much current. Also it would have been nice to have seen a silver plating result vs solder plated result. .. Thanks for this practical application example video.
The video is already 30min long.
@@EEVblog So!
@@EEVblogpart 2? 😊
Clicking on 'DC' instead of 1 Mhz for the trace frequency which was forced ON by default for the older standard may also approximately double the current.
Nah, clicking dc doesn't change the current in the calculator.
No it doesn't change anything.
A couple years ago I was wondering what the current carrying capability was couldn't find anything this was very handy.
Back when I worked in science, the electronics engineer of the group panicked about traces being to thin (Horowitz-Hill stated they would be fine...) and solderedcopper wire to the pins connected to those traces.... nailed the Opamp -15V to the digital +5V.... PSU war ensued, killing all opamps with a 5-0-15 supply until the 5V PSU gave up, then the -15V kicked back in to start a genocide on the digital electronics.....
Hahahaha. Great!
I had two 12V to 3.3V buck converters. Redundant power on two different boards. Before the two 3.3V were joined together they went through two "True reverse current blocking" IC circuits. When the two power PCBs were connected for the first time, one buck supplied power, the other consumed power, boosting the 3.3V up in reverse until the 12V rail reached 22V and the buck IC burst into flames. Don't trust True reverse current blocking IC's 😅 Later, according to datasheet, it blocks reverse current above 1A!!
Why did the buck not have overvoltage protection on the input to save itself from itself.
Yeah, I'd like to see more of this style of video.
very handy talk ! and here was me worrying about my 3A power supply layout when most of the main output track is at least 7.62mm wide on both sides of the board.. somehow I think it'll be fine :)
Just for information, 1 oz copper is the measure of the weight of a square foot of the material. If you buy copper sheet which weighs more (used in construction) they use the same measurement methods
I've seen boards where they soldered solder braid onto the traces.
Solder a thick piece of copper wire along the trace.
Another factor that influences things is airflow over the pcb, how much cooling can be provided by fans, if possible?
My top 3 hates as an electronics repair engineer: Fans blowing shit all over PCB leading to corrosion and tracking. Conformal coating literally not solving this issue and making PCBs almost impossible to work on. electrolyte corrosion.
I've done >150A on a multilayer board. You wont get reasonable trace widths with an IPC calculator because they assume only natural convective heat dissipation. Trick is to bond the board to a heatsink with lots of thermal vias and quality TIM. If you have transistors connected to heatsinks, this will also pull heat from the traces. This makes the current handling problem more of a thermal management issue.
Couple 10s watts in trace dissipation seems like a lot but your board could be pushing multi kW at these currents!
Sometimes you don't even need a PCB and it be better to use chassis mount packages.
IMO the best approach would be ti get some square copper rods/wire, cut them to size, place them on the bare pcb trace just like any other smd part trace and solder them in place.
something I haven't seen mentioned is the duty cycle.. handling 70A for a few milliseconds here and there and averaging 20A is different than handling 70A full time.
I don't know the application of this, but I'd look at taking the MOSFET off the board.. and having separate connectors to it.
As for adding wire to the trace, remember that old solder wick you clipped off? soldering that to the exposed trace can give you a 20A continuous trace for cheap if you don't need millions
I use Electrodoc pro in my phone to calculate needed trace widths. It doesn’t have too many parameters for the calculation, but for my needs it is perfect. Has also many other useful features.
Solder thick bare copper wire or braid to the track, done that many times for motor controllers.
Thanks. Highly informative. I am looking at doing up to 40A on a 2 layer board. 1ounce. I was thinking of making approx 7mm traces, leaving the solder mask off, then soldering 8mm2 wire along the top. Essentially it's the bus bar idea. should work for small runs.
Yes, not an uncommon technique. In that case though you'd usually just use the bidge wire on it's own. Less messy.
@@EEVblogThanks. I might just do that.
L-shaped copper busbar. More cooling. Less footprint.
Don't add solder mask on the trace and throw in two, three or even 4 heavier gauge wires like 19 or so with a lot of solder after the PCB has been manufactured. Obviously it will cause warping but so will be the case with any other method for that amount of current on a single layer board.
23:18
i wouldn't call it a bodge when they silk-screened the wire path.
bodge would imply last minute quick fix at the end of the production line.
Thats a very important topic. Im never sure about the temp. I want to allow a trace to get.
100dgC. At 130dgC PCB start carbonize after some years.
Put if your PCB is 100dgC you need to derate components and capacitors.
awesome video! Keep doing more of these!
Never saw you try the different PCB materials. (Substrate Options) Curious how much difference different board materials would make regarding heat dissipation.
I've handled this before by leaving mask off of the tracks and then hand-soldering along them to make nice thick tracks. Obviously doesn't scale to high volume production, but for a few specialised boards it works well.
Look up "heavy copper PCB". There are some manufacturers that can create traces that are 20oz or thicker. They can do multiple trace thicknesses on the same layer. It's certainly not cheap though.
That's a heck of a current. I would certainly choose a non PCB mountable MOSFET package and connect that beast through its screw terminals.
Thanks for the software tip. It looks excellent.
Just add more copper from outside(copper wire) with layer of tin(solder) to make good thermal conductivity and heavy current carrying capability and you can do it with just one layer pcb.
1. Update your Saturn PCB calculator. There is already a version 8.31 or higher available. 2. Plating thickness matters if you have at least a two-sided rigid pcb with vias. It is not a thickness of ENIG, HASL or other finishes. It is an additional copper plated onto the pcb during manufacturing process. A minimum guaranteed amount of copper plated onto the PCB for vias plating depends on the IPC class it is manufactured. Many fabs plate about 20um of copper (IPC class 2 as far as I remember). Pay attention to "total copper thickness" below as it changes when you use plating thickness modifier. Total thickness greatly changes differential impedance, for example.
That is pretty informative. I knew about some stuff. But I never considered that having just some copper (plane preferably) on another side, helps. It is obvious tho, and interesting approach. I was designing a smart relay board, with some power metering, and wanted to support 16/25A. Doing 10A was easy. But going into 16-25A, was really tricky, due to compactness of a design, and also logic / digital circuits (so could not use super thick layers). A lot of trickery was also around current sensing resistor (I wanted to have resistor shunt, not magnetic sensors or anything like that), voltage isolation distance, and connectors. In many cases, I just went with copper fills, instead normal tracks, to give it as much thermal dissipation as possible, and removed most of the thermal reliefs (which are there really only for assembly).
I needed to build a laser power supply that could supply up to 50A each into two channels (~2v). Board was all surface mount and I couldn't really go beyond 2oz copper (hard enough with 2oz and .5mm pins on the controller IC). I via stitched pretty large planes on the front and back of the board and bonded the back of the board via a non-conductive thermal pad to a heat sink that uses forced air. Works pretty well, but the board is hardly ever driven to full capacity and there are on-board temperature sensors that shut the thing down if it starts heating up too much. The distances from MOSFETs to power connectors were very small which helped. It was a hobby project, so if dies its only me who's impacted (but big impact as it's connected to a pair of very pricey fiber coupled lasers diodes).
This is a good video dave! Thanks
Super useful and interesting, especially that cool Saturn PCB Toolkit. Do have a question, though. I'm guessing that all the calculations are done for continuous current handling? I'm sure it's always best to be more cautious with high current designs, but is there a case for going with slightly underrated traces if the currents won't always be at peak draw?
I have done a big professional experiment for this, but for EV-motor-cables.
I ran max continuous current until stable temp. After cooling down I doubled the current and documented the time it took to reach the same temperature.
After redoing for different wages and current ratios and interpolations, I got a table of pulse capabilities for this cable type. I learned that the radox rubber insulation can absorb and store a lot of heat energy. More than copper.
Also that 30 minutes can be regarded as constant load for cables.
Oh dang, that's a lot of work, but I can see how that information could end up saving a lot of money, and possibly weight if you can ensure the application is within spec of your findings. @@scoutjonas
Interesting read by Doug Brooks and Johannes Adam called "PCB Design Guide to Via and Trace Currents and Temperatures" covers a lot of this in detail with experimental data.
Professional thumbnail
The thickest copper I have ever seen on a PCB was 20oz, it made the PCB heavy and very expensive.
Copper bus bars would make more sense in an application that needs that kind of current.
The "Thickest Copper" is sometimes found unfortunately on the beat in the UK.😠
Bit surprised you also didn't talk about the voltage drop that would occur when allowing for the traces to get hotter.
The video is already long enough.
Cooling a board with a fan will double current capabillity in top layer. You can also use gappads to wick heat from power traces into an aluminium case or cooling plate or heatsink.
Would it work to do a combination of thick copper trace and a wire to handle the current?
Yes. But if you are adding the wire, then why not free up the PCB real estate and have no trace?
If you consider some PTH connectors for power arrival and departure, and if you have multilayer board, how do you deal the fact that, with time, some internal plated holes (vias, connector pins) could internally break and then current will only go through one layer instead of many, dangereously increasing temperature? (Presence of plastic connectors, fire hazards..)
You pronounced my name right 😂 0:26
The last Class D amp 4 x 120w RMS, power rails, were foreseen with serious traces copper, with on top a full bar copper on it. So, hands off there. And they were extremely short. Good I just had to replace 1 end stage and change the smaller caps on the SMPS. Dangerous toys :)
Thanks, a useful, entertaining and informative piece of 'waffle'.
On version 8.32 of the sw, they do take into account the thermal effects of the conductor layer even on IPC-2152 with modifiers mode
A 3d printed jig to reliably bend a solid copper wire to the right shape, then hot-air solder it on top of the trace tends to work well.
Isn't the copper plating different from the finish? For example, when you have a 3oz/2oz board, wouldn't they start with 2oz base copper and then plate the through holes such that the external layer thickness is 3oz?
It depends. Some chinese manufacturers try and save money by using a 2oz board and plating 25um on it which they then call "3oz finished copper". In reality this is only about 90um. Just make sure you want 3oz base copper. You will end up with about 125um total thickness after plating that way.
Usually if i am dealing with a single or few high current signals, i just use a high power 0-ohm shunt to pass the current through the board in a few jumps.
At the end of the day, you still have the get the current in/out of the board and for most high power FETs, Relays, Switches etc..., this winds up being a heavy duty pad of some sort. Ohms per square applies here. QFNs and TO packages use large pads while QFPs etc...or other packages will use multiple IOs. The PCB on fire is highly unlikely...I have seen many burnt PCBs and never one that caught fire...The FR4 materials etc..are designed not to combust.
What if you ran a bare copper trace, no solder mask, then soldered plain old square/rectangular copper stock to it? In effect, a surface mount bus bar?
That's what Tesla does. PNP SMT busbar links.
I used to work for a company that had a completely finished board that was failing. they had run a real hot trace through the middle of a multilayer board and they couldn't figure out why the boards were failing. i may have missed it but you did not point out that when you have a lots heat in a trace it comes back out through the parts and the parts fail. the trace itself won't but the parts will. I went back and looked up what you did mention in the old documents and they assumed one layer one side of a board only. Once you do a multilayer board the thermal conductivity changes especially if you've got hot traces adjoining each other in several layers where the heat can go nowhere even in the PCB plastic. One of the solutions is just to go out and run a wire from one point to the other and take the heat in the wire. however if you're in a multi trace board this is not always doable.
@Dave How about surface between THT component leg and PCB layer? You can make trace wider but apart from thicker trace, component leg surface in contact with trace will not change. Sure tin will make some additional surface above board.
I've just discovered the existence of embedded resistors and capacitors. Could you do a video on this?
80 Amps continuous for low frequency / DC switching is not an easy task in electronics packaging / electronic subassemblies. Also, the circuit voltage was not specified which makes a huge difference in device selection and thus impacts device packaging and the overall packaging approach. It is not just the routing of current but also the power dissipation of the switch which will determine a packaging approach. For example, if the MOSFET had 15mOhm on resistance, at 80 Amps the device power dissipation is 96 Watts. For 96 watts you would want to spread out the heat over several parallel devices, especially at higher voltages. Fortunately, for sharing we have the newer Sic MOSFETs.
When you get someone with Dave's background you will also see an experienced eye for "does it look like it would work" first approximation. In this respect you might see perhaps 4 parallel Sic MOSFETs on a heat sink having TO-247 packages. With the current to and from the switching circuit using 4AWG wire with crimp wire lugs to PCB soldered lug connectors. Which means the PCB high current traces lengths are minimized (perhaps at most 750 mils from the PCB lug) with input routed on the top layer and output routed on the bottom layer while maintaining minimal loop area. The PCB cooling to a great deal local to the switches is in large part due to the huge wire bolted to the PCB connections. The power devices are mounted to a heatsink. Essentially, this allows you to use a common power package (TO-247) without using an expensive power module which has screw terminals, and supports a PCB approach using 2 oz copper which would support a microcontroller QFP package. One rule of thumb is when using a PCB at high power is to only route on the outer layers, even though the board may have 4 or 6 layers. Localized hotspots within a PCB can lead to the PCB overheating (catching fire). If the board has internal layers normally they are not used an are omitted from the high current area of the board. High current vias through plane layers is not recommended and if a short occurs can create a fire hazard. So, no internal planes in the high current region.
For a poor mans bus bar and with a soldered wire connection, good to about 30 amps at 16 Volts, I have specified a stripped stranded copper 10 AWG wire (tin plated) which is soldered full length on top of a long tinned trace to increase the current of the circuit. The wire acts like a bus bar. But, I only did this during the prototyping phase. In production you would have a bus bar and PCB lug connection. Some times the lug is part of the bus bar at its terminating end. However, custom bus bars are expensive.
For very compact high power amplifiers (like three phase CNC spindle motor drives that can reach up to 60HP) where you can justify the use of custom bus bars, the design becomes a very interesting 3D exercise, this is because bus bars can route above a PCB when necessary and form conceptually another routing layer in space above or below a PCB. Bus bars can be clamped together using insulators, then secured to screw connections on power modules or PCB lug screw connections. Custom bus bars are usually made out of copper and are tinned plated. But, you can also have them silver plated, or other plating combinations. If you design your own bus bars that are large (5 to 8mm thick) you can make a tab off of a ground bus bar with a hole to connect a banana plug to or clamp on with an alligator clamp for your DMM.
2 AWG copper flex FTW. Should be able to comfortably handle 100A.
I'm going to assume that in an application like this space is no object. You'll need a lot of room to get those cables around.
Where can one set the pcb thickness to 2oz, 3oz or 4oz. Can it be done in the software for the pcb designing?
It was literally my first thought - why not a buss bar or just thick ass wire or sheet cutout. So spoiled we are nowadays, my tube amp doesn't even have a board in neither the low nor the high power circuits. It is all built elegantly right into the chassis with wires or component to component joints, and has survived many decades of operating on top of a rumbling speakers cabinet, that's way way above the standards to which modern equipment is built. Not to mention all parts that wear off are conveniently socketed for trivial replacement.
I would recommend leaving the trace unmasked and then soldering to it some thick copper wire bent to shape.
Using flar busbars or copper plates sandwiched to the PCB is better than using a insulated wire. The insulation is usually max 70dgC and the isulation will reduce cooling of the wire. A bussbar can be 100dgC and doule cooling. Its also better for vibrations.
i was told to use bus bars over thick traces as it was cheaper and safer.
If I were the author of the question, I would first look at how this is done in welding inverters. However, it would be much better to ask the question of the correct choice of connectors.
How about soldering a copper wire over the trace?
Depending on a number of conditions, soldermask can even be beneficial, as emissivity of bare copper (or any other shiny metal) is very low.
If dissipation has a sizeable radiative component, e.g. still air, high temperature, the minute thermal resistance added by the soldermask is negligible with respect to the improved emissivity.
To the guy who asked the question to dave. 80 amps is no big deal. Everyone is freaking out about it. I would suggest you start looking into ESC's (Electronic Speed Control ) for RC. Those things hand handle hundreds of amps.
Funny I was looking at hall effect sensors such as ACS772 that can do 200 amps and wondering how do you even make a PCB to handle that, and this video popped up. I guess for those you would use a bus bar with lug connections. Even soldering wire down becomes a little gnarly as you're into like 2/0 wire at that point which is super thick. I'd almost be tempted to skip the PCB altogether and try to solder the wires directly on the chip but that might be challenging to do as well.
Thin copper plate busbar without insulation and with some forced air can handle 10x the current in a copper cable of similar size. Improved cooling doesn't reduce voltage drop, but for short distances it is ok to push high current density.
If you look at some welding power supplies, you can find some really heavy currents, like 150 or 250 amperes, or even more. Almost certainly you find 4 or 6 oz copper, on two sides and a large number of vias between those two layers. Passing the current on two sides and a large number of vias benefits you with balancing thermal warp and also locking the two copper layers to the FR4, which has probably even higher thermal expansion rate than copper. But adding other metals like tin don't benefit you much, because they have much lower conductivity than copper. I assume that you don't even think of using silver (which is about the only material better than copper).
Some of the easiest and cheapest ways is to use "current shunts" that are just the bare copper or steel wire. High current low resistance you can jump components and lanes with it and they are cheap as hell.
Is there a place where we can order busbars like we would a PCB? Just upload some design files and get a quote? Or is it not feasible for DIY/prototypes/small runs and stuff like that? 🤔
In the end, it's just a metal stick.
You can make bus bars out of copper tubing
80 A - sounds like bodge wire o'clock.
edit - ah - commented prematurely, I see you make that point towards the end . . .
Wires might also make sense if the mosfets themselves have to dissipate a lot of heat as you can mount them wherever you like on the enclosure and use that for heat dissipation.
Nothing beats the classic look of some externally mounted TO-3s.
It seems sketchy as hell that ( 16:11 ) moving the conductor layer to be an internal layer had zero effect, but having a plane present ( 17:10 ) had such a huge effect. Even if this calculator is just a rule-of-thumb thing, that fact that is has no problem with wrapping your heatsink with fibreglass makes it totally untrustworthy? Also, skin effect has a huge impact on current carrying capacity of a wire. I sincerely hope the only reason it made no difference to this particular calculation was because it was truly negligible on a 1 inch x 35 um trace at 1MHz, as opposed to another thing the calculator is incorrectly ignoring.
MyVanitar channel also talked about this in a video where Altium itlsef can calculate these parameters
Can u tell slope compensation in current mode power supply
how about edge plating as well?
Thanks for the Saturn PCB tool. It's pretty cool looking, even though the thing exceeds the height of my laptop screen, lol.
i have a couple SK 10123/2350 beasts rated at 50A....T0-3 cases with 1/16" thick B and E pins......googled that ID and got nothing, tried Mouser and it seems that it is now an Onsemi part, MJ11033G....
I remember seeing a board with wires soldered extra on top of the traces, effective but it's not perfect to the eyes.
Anyone knows the videos that dave made on stackup