That surprises me… I guess it could be down to how the software reset is actually triggering the reset in the device. If it’s just changing the execution vector back to the reset point then that would make sense that it ignores the hardware pins. If it’s creating a reset using a specific process like a CPU register then it’d be a case of maybe it ignores the pins, maybe not. I would wonder about other reset sources like a watchdog reset or reboot via debug interface/USB flashing? That may be a limited enough set of use cases to not matter.
1 Like
Those diodes are both B5819W.
Looks like at 2A of forward current you should see around 600mV of drop.
So around 1.2W of heat to dissipate in the one labeled D2 on the thermal picture (D18).
The one labeled D1 in the thermal picture (D2) is in the switching power supply. This is a different use case for the part. It should be seeing the 570kHz switching in the buck converter output. I’m not sure it should be working as hard as it is.
1 Like
When I checked it, steppers on and pendant in, it was a drop of 0.3V. I looked at other Schottky diodes and they were all more drop, even the ones on the JP1 are double. The 3.3v regulator is doing 4.7v to 3.3V seems like there would not be much heat generated in that small step but I know very little of how they actually work.
I kept a board running all day with the steppers and pendant. I hope it is just because the drivers are sinking so much heat into the copper. They never got any hotter.
On a positive note, somehow The jp2 upstairs had a better signal than the jp1 next to me on the desk. Maybe the wifi is turned up and the esp is sucking in more current than the jp1. That occurred to me after I left the shop so I will have to look into it tomorrow.
Interesting, I was kinda lot at what it was doing there. The other schematics I looked at all had it so I knew it wasn’t wrong. I will dig into that a bit deeper tomorrow as well. Slowly learning some of this stuff.
1 Like
In that application, you can measure the forward drop and you can look at the curve on the data sheet to see how much current is flowing through the part.
I’m not worried about this one.
Switching buck converters are a really fascinating sub discipline. You’re basically switching on and off that coil and caps really fast, charging them up with energy from VMOT to keep the output voltage at the desired point.
Maybe that amount of heat dissipation is normal, the efficiency of this particular converter doesn’t have to super high in this application.
1 Like
How that buck converter chip works is that it takes the node labelled PH and alternates between connecting it to Vin (Vmot) and open circuiting it. When connected to Vin, you’ve got say +24V on one side of the inductor and 5V on the other, so 19V across the inductor. That voltage causes the current in the inductor to ramp up, storing energy in the inductor. After a period of time in that state it will disconnect Vmot leaving that pin high impedance. The current needs to continue flowing in the inductor so instead it flows from ground through D2, sometimes called ‘freewheeling’, but basically discharging the energy stored in the inductor out into the load.
The ratio of on/off (the duty cycle) is controlled by the Vsense pin, but if everything is stable then the duty cycle is pretty much just determined by the input and output voltages. D = Vout / Vin.
So if you’ve got 10V in and 5V out then the converter will spend an equal amount of time charging the inductor and discharging it. This is because the inductor has 5V across it (10V in - 5V out) while charging and then 5V across it (just whatever the load is) while discharging. If that ever stops being balanced then the output voltage will start to change which will be measured by the Vsense pin and used to adjust the duty cycle to regain balance.
In the example of 24V in and 5V out, the FET in the buck converter IC will be on ~20% of the time and the diode will be conducting 80% of the time. 19V while charging the inductor, 5V while discharging.
In steady state, the inductor current needs to be the same as the load current, otherwise it wouldn’t be steady state. An equal amount of current needs to flow into the +5V node from the inductor as flows out through all other paths. So if you’ve got 2A of load on the 5V then there will always be 2A through that inductor. The diode will see that 2A while it’s on, so it’ll be at 2A for 80% of the time, 0A for 20% of the time. The FET in the boost converter IC will be the opposite, 2A for 20% of the time, 0A for 80% of the time. The average current through the FET is the same as the input current, so in this case it’d be 0.4A coming in from the 24V node.
That’s all the idealised scenario, of course, ignoring anything like resistance in the FET or inductor, voltage drop in the diode, etc. It also shows that it’s important to consider the instantaneous current through each device, not the average. In a scenario where you have a 5V 2A load being supplied by a 10V input, the diode will be seeing an average of 1A but that’s actually 2A for 50% of the time. At 1A that diode has a voltage drop of 0.5V so you could say it has an average loss of 500mW. In reality, at 2A, the diode has a voltage drop of 0.65V, so it will have a power loss of 1.3W for 50% of the time, so 650mW average. There will also be higher peak die temperatures with a ‘pulsating’ loss like that, but that’s getting way beyond what’s worth considering at this level.
That diode is rated for 500mW at 25C and then derates linearly beyond that, so if you were calculating this based on the average current you’d say it’s bang on rating, provided the ambient stays limited. If you calculate it using the integral of the instantaneous loss then you’d conclude that it’s running 30% over rated.
There usually aren’t specs for case to ambient so it’s difficult to estimate how much loss is coming from that package, but at 60 degrees C I definitely wouldn’t worry about it.
1 Like
OMG, I’m learnin! That is exactly how I tried to compare other diodes. What a validating feeling. I felt like a cheat looking at the charts felt like I was doing it wrong. You just made my night.
I am really hoping it is the hot PCB combined for the heat delta. Seems most components are never rated for a “3 degrees per watt”, but more of a “3 degrees over ambient” so hopefully 8 degrees above the substrate temp is normal.
Also I should add the heatsinks and a thermistor to keep an eye on temps, maybe even see how my fan ducting works on the new board box.
1 Like
Ohhhhh, I am following along!
okay.
Thank you guys so much for teaching me. I keep saying it but I do enjoy the process of learning on a real project, with good teachers explaining things I get wrong. I am even getting a bit more competent on my scope, had 3 probes today and almost used a 4th…and I knew what all the values meant.
7 Likes
There’s a great satisfaction in learning, and a great satisfaction in helping others learn.
4 Likes
I swear if you guys were there this morning there would have been so many high-fives, and me repeating, “oh…I get it”
6 Likes
Well, that’s possibly one of the most bizarre engineering related sentiments I’ve ever heard expressed out loud… Charts are where the real information lives! Then you go beyond that and start using multiple charts to infer things about the overall shape of the data, like how the line on a graph changes with temperature, what its error bars are, that kinda thing.
Usually it’s a maximum temperature then a junction-to-ambient or junction-to-case thermal resistance. I would look at the power vs temperature ratings with suspicion, usually. They’re a good ‘first pass’ sanity check but not something I’d design to.
In this case, it’s 200C/W and a maximum junction temp of 125 degrees C. How close you get to that maximum is a judgement call. The part won’t outright fail at 126 degrees but the lifespan calculations are all done at 125 degrees. If you want long life and high reliability, you might go 80-100C, if you’re trying to aggressively minimize price then you might go to 110-120C. You’d then have a worst case operating spec, say 50 degrees C inside a case. Based on that you might say 50 degC of operating margin… That gives you 250mW of loss allowance for that component before you need to change something.
For a diode, I’d then follow the process above to figure out the actual real world operating loss. Instantaneous current and duty cycle from the input voltage, output voltage and output current specs. Voltage drop from the charts, using the line for the higher junction temperature because that’s when it matters. I’d also usually try to leave some headroom because often the charts are ‘typical’ not ‘worst case’. For that diode it says Vf is 0.6V max at 1A and 25degC, 0.9V max at 3A and 25 degC. The graph says closer to 0.5V at 1A and 0.8V at 3A so I’d assume +100mV as worst case. If my actual usage condition was 2A for 50%, I’d go to the graph and see that it’s typically 0.55V, I’d add 100mV for the ‘max’ scenario and say 0.65V @ 2A so 1.3W and go from there, etc.
There’s also reverse recovery current/losses which comes into it which is where things can get a little more crazy, but I wouldn’t worry too much. Basically a diode doesn’t ‘snap shut’ as it turns off, it actually lets a bit of current flow backwards through it. How much current and for how long can depend on a lot of things, like how much forward current there was, how quickly it went from conducting to not, what the reverse applied voltage on it is, all sorts of stuff.
There’s often a balance between component size and ratings. You can often get things in small packages with high ratings but they’re quite ‘highly strung’. They won’t have much thermal mass and in some cases can more easily end up with a loop of overheating leading to higher losses leading to more overheating etc. I tend to be quite suspicious of SOD-123 and SOT-23 packages as they have very little die area and tend to vaporize if something goes wrong!
3 Likes
That all made sense. Thank you.
I did do a larger package size on the inductor…but smaller on the diodes. I might have to go back up a size, if it bugs me after cooling down the pcb.
Actually, the next trick from dealing with charts is to use something like WebPlotDigitizer to extract the data from a plot and use it in a more precise manner.
From there, sometimes we end up trying to combine data from multiple plots to create a multidimensional surface plot or similar, all depending on what we’re trying to figure out.
The step beyond that is that there are some tools that allow you to create models from that data or use models from the manufacturer where those data points are the output of other circuit conditions. Plecs is an example of that. For a MOSFET you can add curves for voltage vs current at a single temperature, which is a bare minimum. You can then add either another curve for voltage vs temperature at a single current or multiple curves at different temperatures and it will extrapolate for you. You can also add in the thermal model of the die you’re using and it will account for that, too. You might turn the switch on at 100 degC and it will figure out what the loss is at that point, then track how much hotter the die gets due to that loss, changing the loss instantaneously as the die temperature changes. You might have an average of 100degC but over a switching event it might go from 95 degC to 105degC and those changes might make enough of a different in loss that it’s worth taking into account.
Have a look through that datasheet. That’s a device I used in one of the previous power converters I worked on. 900V DC bus, ~100Arms output current, switching events occurring at close to 200A instantaneously sometimes. Even then, I’d say those graphs aren’t 100% comprehensive. We relied a lot on the model for the above software that Infineon provides (sometimes helping them to preview and provide feedback on new coming devices) which would provide another layer of depth to the information in that datasheet again.
1 Like
Colder is almost always better, but 60degC in a decently warm ambient without forced cooling isn’t anything I’d personally be worried about.
1 Like
That’s cool!
3 Likes
Jackpot2 Box, Jackpot2 Box by V1 Engineering | Download free STL model | Printables.com
Thinking about hiding the fan, but this is the box I have for now.
1 Like
I am putting 10 up for sale, with a massive discount for brave testers in the US that have a little time to give it a shot. Help make sure I am not missing something, The Jackpot2 CNC Controller – V1 Engineering Inc
When we get a few reports back of no new issues I will put up more.
6 Likes
I updated my JP1 to 3.9.8 to check to see if my macros work with my cyd pendant. They still work on 3.9.8 and webuiv3.
I just ordered a JP2 so I can test my cyd pendant and the macros on it for you.
3 Likes
Are you only using the single preference file macros of webuiv3, or did you keep the webui2 style separate file macros?
The separate files work, but webuiv3 does not need them, you can enter all your macros in one file.
If we have to, I have no issue doing it the extra file way, but it seems like a bug.
I have individual macro files on my SD card.
1 Like
Okay that does work. WebUI v3 lists them in the preference file. The pendant sees them it will just not run them. Super odd.
1 Like