DRV8825 current and thermal issues

Hi Dario,

First of all let me thank you for providing so much vital information to help you with this issue. I am positive we will get to the bottom of it pretty quick.

With 0.625V in VREF and 0.2 Ohm resistor, the ITRIP should definitely be set to 625 mA. So my first question is how are you measuring the current? Are you using a current probe, a multimeter, or the power supply meter? Only the current probe tends to give good results. The current regulation waveform is hard to measure with any other equipment. Also, are you measuring the current on a per H Bridge basis or the whole device? What is the application voltage? 

Now, I must admit a 4 x 6 cm is a fairly small board. You should be able to easily drive 625 mA. The 1.4A RMS may fall into a hairy edge, though. We have run an experiment on a similarly small board (6.35 x 6.35 cm), but only two layers and we were able to run at 1.75A RMS per H Bridge without reaching thermal shutdown for a pretty long time (close to an hour). Eventually we reached TSD, however, and we do measure pretty high temperatures (close to 140C) during the running time. Since your board is slightly smaller, although 4 layers, it is possible the size does have something to do with the described behavior. The thermal impedance is just too large and it does not allow the heat building up to be released into the enviroment as soon as we would like.

Thermal impedance is everything. We recently ran another experiment in which we use a special heat sink and the device was able to run at maximum current with the temperature never exceeding 65C. This was on the same 2 layer board.

On another note, although you used hot air to solder the device, I would ask you to reflow the device again and desolder it out of the board to see if the power pad in fact got any solder into it. What I see often is that if the power pad vias are too big, solder tends to wick out and the power pad remains unsoldered. It is a tricky endeavor, but there are ways around it.

Finally, if you want to send me the gerber files to jquinones@ti.com I will gladly take a look and see if there is anything else obvious we can look at improving

Hope the info helps. Best regards,

JIQ 

Hi Jose, thanks very much for your answer,

First of all I found out that both the two shunts were wrong, stupidly I've not measured them (were 20 mohm instead of 200 mohm). Now I've replaced them with two 330mohm 1% shunt resistors and all seems to be better than before.

Unfortunately I've no current probes, I tried to measure the current indirectly using my osc voltage probe. I took the average peak voltage on the shunt resistor (with trigger in sample mode - I had a noisy sin voltage, as I expected when i use 32-step microstepping), and I divided it for the shunt resistor value.

So with the 330mohm shunt, the Vref = 3.3V and supply at 24.5V I found :

- Vpeak on the shunt = 600-605mV

- Rshunt = 328mohm (measured with volt-amperometric method)

so I had (in the worst case) 600mV peak / 328mohm = 1.829A peak. At the same time the power supply meter told me 0.54A (obvioulsy not a correct measure, I thought), so, since the current is a sin, I tried to measure a single phase current putting  the multimeter, in AC-current mode, in series to the phase, reading 1.26A rms that should be 1.78A peak, isn't it? It's a correct measurement way or not?

In this situation I never reach the TSD, also using the motor for 1 hour consecutively but I'm not sure about my current measurement. Can you give me some tricks? If the current measurement is right I have almost the maximum current the DRV8825 could give, isn't it?

About the soldering of the chip: I also find out that with many PCB services the thermal vias in the power pad are too big (I designed those with the minimum allowable dimensions) and every time the solder paste flows through the vias, I can see it from the other side of the PCB. This is not good as I read in many papers. But, in order to be sure about the soldering of the thermal pad, I let all the vias fill with solder paste and then I soldered again the chip. Now I hope the thermal pad should be well soldered.

Finally, I'm very glad to have the possibility of showing my work to an expert, so I'm sending to you my gerber.

Thanks a lot and best regards.

Dario

Hi Dario,

Thanks for sending the Gerber files. You were not kidding. This board is very small! You have definitely followed the SLMA002 guidelines the best that you could with the available space. Do note this space is still quite constrained and the available copper and board structure are just not enough to give you a low enough thermal impedance, which is why I believe the device will get hot under high current circumstances.

If there is some metal in the vecinity that you can tap into, that could help. I realize this is not even easy to contemplate, but it may be the only choice to decrease the thermal impedance.

The numbers that you are measuring almost comply, except that according to the ITRIP = VREF/(5*RSENSE) equation, something is not making sense. If the VREF is really 3.3V and the RSENSE is 0.33 ohms, then we should see 3.3/1.62=2A peak which means you should be seeing 660 mV at the Sense resistor. Could it be that your peaks are 660 mV instead of 600 mV? I realize measuring current by looking at the SENSE resistor is not the easiest due to all the switching noise. I think we are close, though. Or at least closer than this morning.

At the end, however, 2A Peak or 1.8A peak will not make that much difference. If that is the current you want to drive, this board's size may haunt you. It may work at 25C, but as you go up in ambient temperature you will approach TSD levels. Not to mention it will run very hot. If you are trying to source 600 mA, I believe you should be fine.

BTW, I noticed the vias on the power pad are 13 mils. If you want to avoid solder wicking through, you can make the via hole size 12 or less. It may raise the cost a little bit but will definitely help with the solderability of the power pad. The other option is to use via tenting at the top, in which case you rule out solder wicking.

Hope the info helps. Best regards,

Jose Quinones 

Hi Jose,

I know the board is really small, unfortunately! I have this constrain, the board dimension! I know that it is really really small and with not so much space for copper land but I have to drive 1.8A, not less!!! Do you think that placing the components in a different way could be better to achieve a lower thermal impedance? (for example using larger dimensions for the copper land beside and behind the DRV8825)

You are right about the temperature, in fact the DRV8825 during high current conduction get quite hot. I was afraid about the thermal impedance of this PCB, but I had to try! Could be a solution using a dissipator and a layer of mica? Could in this way the DRV8825 be not so hot, or is not enough?

You are right also about the measurement: using a new PCB (just mounted) a precise measurement of the VREF is 3.25V, not exactly 3.3V (not a so relevant variation but...). Probably the measurement done with the osc could be not so precise, there is a discrete amount of noise. A new and more accurate measure with this new board shows 620-625mV peak, nothing more. So the ITRIP should be = 3.25V/(5*329mohm) = 1.975A whilst the measured value is = 625mV/329mohm = 1.9A.there is yet a difference of quite 80mA. 

The dimension of the via is the smallest that can be done by the PCB producer I usually use, 11mils should be a correct dimension, isn't it? The second option, the via tenting it's a great solution, thanks so much.

Thanks again and best regards,

Dario