I'm considering using the TLC5925 constant current LED drivers from TI in an LED bill board app:
formatting link
I notice these have max current of 45mA programmable via a resistor. I assume this is quoted when all outputs (16 of them) are sinking current, but my app is a multiplexed system driving dot matrix LEDs. Is it possible to program more current than 45mA as long as there is a sufficiently small duty cycle (I'm using 1:8), and if so how much more? The specs only state matched outputs of < 6% between chips when < 45mA is used, so I'm guessing it would probably be safe but the matching might not be as good. Anyone had experience of these chips?
My view is that I wouldn't think it is safe to assume this, but it could be the case. Best to check the other specs, first.
It says 45mA is a maximum specification and that in such case Vo >= 1V. This suggests a dissipation, per pin, of 45mA*1V or >= 45mW. Times 16, this gets close to 3/4 watt. They give theta(JA) as on the order of 100, for single layer boards, so about a 75C rise over ambient. Since the junction needs to be kept under 150C, this leaves you some reasonable margin for "ambient." Also, the absolute maximum spec says
750mA for the ground pin. Which is roughly 16*45mA, so that seems to match up. So yes, it looks as though you can _actually_ get 45mA per pin all at the same time. Not accounting for frequency of operation and the additional power that higher speed operation may yet add.
If you ask for still more current from a pin, you very well may exceed the ground line _absolute_ maximum. Besides, the Vo voltage will rise still further. And the spec doesn't say exactly how. You will be testing parts to find out just what the damage is, there. And it will mean still more dissipation in the package, of course. And that will start pushing your towards that 150C max junction temp, as well, with lower or even unacceptable margins remaining.
No experience with them.
Another thing to keep in mind is that the designers may have truly designed the trace widths on the lines leading out to each pin to truly provide not so much margin over 45mA. The Absolute Maximum Ratings _also_ say 45mA per pin, not just in the place where those recommended values appear. So it may very well be they have _designed_ it from the get-go to not handle much more than 45mA on the traces leading out. Demanding more may lead to electromigration and then worsened by Joule heating and may even eventually turn the trace into a fuse. This applies to the ground, too.
Get some and test. You may know better than anyone how many of the 16 outputs will be used, worst case, at one time. It's possible that you can "work the issue" a bit. Just keep in mind power dissipation and junction temperature, maximum ground current, Vo at the pin, and all the rest. You are considering moving out of the specification region.
Actually I cocked up, these will be on the columns and the rows are multiplexed, so all 16 could be sinking 45mA (giving an average current through each LED of 45/8 mA).
Yes, I need to look at junction temps and choose the right package to suit (SOIC is better than SSOP or TSSOP). I've done some maths now:
In reality the LEDs will need to be powered by something (say Vled), so the actual Vo is going to be Vled - Vf where Vf is the forward voltage drop of the LED. To cope with the range of forward voltages of the LEDs, say 1.7V to
2.5V, and add the 1V voltage output at 45mA, Vled has to be at least 3.5V, and worst case Vo sinking 45mA is going to be 3.5 - 1.7, i.e. 1.8V. So in theory it could dissipate 0.045 * 1.8 * 16, i.e 1.3W or so which basically means using the SOIC package only.
The theta(JA) for this package on a single sided PCB is 80.5 degC/W, which is 104 deg C, and that allows 46 deg C of ambient to keep junction below 150 degC. The reality is this will be a double sided PCB with flooded pour so probably more like 60degC/W so we should be OK with that. At a push I could have the SOICs stuck down with thermal adhesive onto a copper pour below which would make it even better.
So SOIC it is then, and I'll use 45mA max per pin. I take all your points about the reasons why it might not handle more than that, I could end up gradually destroying the device so I'll stick with the max spec.
The programming resistor is for the max current per output. Look at page 13 "Adjusting Output Current".
Since these devices do not have built in PWM you will have to do it yourself(externally).
No, you cannot program more than the max. You must find a driver that is capable of your peak current per led per row since this is what will be supplied continuously by the driver.
Suppose you have n led's multipled per driver output. This means that you have a duty cycle of 1/n. To get the same average lumens(approximately) you'll need to supply a current per led of n times the average led current. But the driver then has to apply that same current for each of the n led's which is also the same as the continuous current the driver has to supply.
For example,
Suppose you have 10 led's multiplexed on one output of the driver. Suppose you want to supply an average current per led of 10mA. This means that the driver will have to supply a max current of 0.1A which is also the average duty current.
That is, for 10 multiplex LED's we have to supply 100mA per LED to get an average of 10mA. This means that the driver has to supply 100mA continuously(since the driver is always driving an LED).
If the average LED current you want per LED(related to brightness) is C and n is the number of multiplex LED's. Then
Duty Cycle per led = 1/n Duty Current = C*n Average Current = C Driver Current = C*n PWM clock rate(for variable current) = 2*M*n*PWM_Bits*RR
where M = # of driver channels, PWM_Bits is the PWM "resolution" you want such as 8 for 256, etc.. RR is the refresh rate which must be greater than about 100 for blink free. It should actually be much greater and depends on the duty cycle. You want to try to maximize this quantity.
For 10-bit PWM resolution with M = 16 channel, n = 10, and RR = 100
PWM rate = 320kHz which is the minimum clock rate.
To see this, 320kHz is the clock rate, 160kHz is the data rate, 10kHz is the data rate per channel, 1kHz is the data rate per LED, and 100Hz is the refresh rate per LED.
Note that you do not have complete control over all the variables and must work out an optimal configuration. You cannot simply multiplex and arbitrary number of LED's together. From what I have seen, most LED's with a 1/10 duty are much less efficient. There is a peak current, at least for the LED's I have seen, with a duty of about 1/2 to 1/3.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.