Choosing right resistor for transistor base in LED matrix

[kanaron]

Hi
I want to create LED matrix 7x5 and I was wondering choosen resistors are correct in this case for transistor's bases and if this approach is valid at all.
If I understand transistors correctly then base resistor determinates current going into base and collector-emiter current is base current times beta. But in case of LED matrix collector current is changing. I'm multiplexing columns so one "row" NPN transistor will always have either one LED off or on so 0mA or 20mA on collector. But "column" PNP transistor can have 0-7 LEDs turned on so 0-140mA on collector. And this leave me puzzled because from my reasoning I have to use different base resistor depending on how much current is on collector. I checked part of this design on breadboard and with 1k resistor on PNP transistor base everything worked. Maybe current consumption showed on my power supply was lower than 140mA with all 7 LEDs on. But I just prefer to ask first if all of it is correct before ordering parts and PCB

If screenshoot is not readable then:
- base resistors for "column" PNP transistor is 1000 ohm
- current limiting resistors are 1 ohm (because voltage drop of LED in datasheet is from 2.6V to 3.4V and supply voltage is 3.3V then I assumed that 1 ohm resistor is probably ok in this case)
- base resistors for "row" NPN transistor is 3600 ohm

[ledtester]

1 Ohm is way too little for the resistors R51 - R55. If the LED voltage happens to be 2.6V and allowing 0.2V for CE drops, you'll have ( 3.3-2.8 )/1R = 500mA going through the LED.

I'd start with 100 Ohms and adjust from there.

I'm multiplexing columns so one "row" NPN transistor will always have either one LED off or on so 0mA or 20mA on collector. But "column" PNP transistor can have 0-7 LEDs turned on so 0-140mA on collector.

With your schematic I don't think you want to do it that way. I would have only one NPN on at a time - ie scan one row at a time. Then you can size the current limiting resistor for just one LED being on. Otherwise you have to pick a resistor value which will work for one to seven LEDs being on -- quite impossible.

[Xena E]

There's no need to explicitly calculate a base resistor for each current scenario.

Pick the maximum current that the driver transistor will have to handle, calculate the current needed to bias the transistor into saturation, collector current divided by worst case gain, (from the data sheet) multiplied by five¹ for a safety margin to ensure the transistor is fully saturated then from that figure calculate your resistor value.

¹Purely rule of thumb, basically the resistor is a current limiter it's just about keeping the transistor switched fully  on.

As for the shared current when multiple leds are on, either multiplex or give each led its own individual resistor from the bus lines.

[PGPG]

And this leave me puzzled because from my reasoning I have to use different base resistor depending on how much current is on collector.

Your reasoning is wrong.
In such schematic you not regulate LED current with transistor base resistors. This resistors should insert into base enough current for transistor to be totally switched on. So if you know that maximum collector current is 140mA and beta is 100 you need to insert in base minimum 1.4mA. This will switch on the transistor but it will be poorly turned on. If in collector there will be 140mA than voltage drop at transistor (collector-emitter) will be about 0.7V, but when only one LED will be on the voltage drop will be 0.2V. And you will get different current in LEDs depending on how many of them are on. You should insert in base more current. The reasonable minimum in your case will be 140mA/20.
The same with npn transistors - minimum is 20mA/20.
Current in the LEDs should have been set by resistors that are in serie with them. Measure voltage drop at LED when 20mA is going through them and assuming that at pnp and npn transistors you have 0.2V calculate the LED serial resistors. 3.3V can be to small voltage if these are LEDs needing about 3V. Typical red LEDs have voltage drop around 1.8V.
To ensure LEDs will light more or less the same you should have a drop at these serial resistors about 1V. In other case you will notice difference in brightness as because of voltage drop differences at diodes you will get different currents and brightness depends mainly on LED current.

And one more thing. I understand that you tested how your LEDs shine and you assumed 4mA average current is what you need.

[kanaron]

Thanks, I will make R51-R55 larger.

I would have only one NPN on at a time - ie scan one row at a time. Then you can size the current limiting resistor for just one LED being on. Otherwise you have to pick a resistor value which will work for one to seven LEDs being on -- quite impossible.

So... multiplexing by each LED one by one? Then each row will take 7 times longer and entire matrix will takie 35 times longer to multiplex. It is simpler and I would only need 74HC164 shift registers for columns and rows and I could get rid of HEF4094 shift and store registers. I will test this approach tommorow to check if that 35 times longer for a matrix is visible or not. But based on my previous tests I think 1/35 duty cycle will result in much dimmer or almost invisible dots

As for the shared current when multiple leds are on, either multiplex or give each led its own individual resistor from the bus lines.

Thanks, I will calculate it as you said. And I think I will go with individual resistor per led. I'm already scared of soldering session (or weeks of soldering) as I have to place 24 of these matrixes on 10x5 cm space... 840 leds and now 840 resistors next to each led

3.3V can be to small voltage if these are LEDs needing about 3V. Typical red LEDs have voltage drop around 1.8V.

Well, these leds are green and in datasheet it says that voltage drop is 2.6V - 3.4V... I'm not sure why there is 0.8V disperssion. Is it voltage range on which leds will be on or are there manufacturing errors and one led will have 2.6V drop and another could have 3.4V... I hope not the latter...

To ensure LEDs will light more or less the same you should have a drop at these serial resistors about 1V. In other case you will notice difference in brightness as because of voltage drop differences at diodes you will get different currents and brightness depends mainly on LED current.

If I should have 1V drop on resistor then for sure 3.3V supply can be too small. However brightness difference is not a big problem as for most of the time majority of leds will be off. Some of them will show few digits and other will not show anything most of the time. Of course I hope that we are not talking about massive brightness differences

And one more thing. I understand that you tested how your LEDs shine and you assumed 4mA average current is what you need.

Well I didn't test these particular leds as I still have to make a decision about package. I hoped for 0603 to make it easier to solder but if I have to place resistor next to each led then I think I will have to go with 0402. I tested this with factory made green led matrix 7x5. I hope there wont be much of a difference

[mariush]

Well, if you want to make it yourself, I would say make it as bright as you want. 

You will light up either one row out of 7 at a time, or you'll light up one column at a time out of 5...

So any led will light up 1/7 or 1/5 of time, which means  you can run the leds at currents higher than default current, because they'll have time to cool down.

For example, lots of leds rated for 20mA will also be rated for 100mA at 1/10 duty cycle (ex run for 10ms, off for 90 ms or something like that) so you could configure the default current at 40-50mA and as long as you don't screw up your code and keep looping through the leds, you'd be fine.


A slightly better solution would be to use led driver chips.

The easiest would be to use chips designed to run seven segment digits... lots of chips are designed to control between 4 and 16 "digits" , with anything between 7 and 16 "segments". You just tell the chip which segments to light up, and the chip will take care of looping through the digits updating the segments.

So in your 7 x 5 matrix , you could treat each group of 7 leds as the segments of a digit, and therefore you have 5 digits with 7 "segments" each.

TM (Shenzen Titan Micro) makes a bunch of seven segment led drivers that are very easy to use.

For example

TM1640B  can do 16 digits, each with 8 segments : https://www.lcsc.com/product-detail/TM-Shenzhen-Titan-Micro-Elec-TM1640B_C2962200.html

You could literally have 3  8x5 matrixes controlled by a single chip, or you could link together 3 "grid" pins at a time so that each of the 5 rows/columns will light up 3/16 of the time.

TM1638 can do 8 digits, each with 10 segments : https://www.lcsc.com/datasheet/lcsc_datasheet_2401261512_TM-Shenzhen-Titan-Micro-Elec-TM1638-TA2009B_C20611979.pdf

So you'd have 16  5 led columns, basically 2  7x5 matrixes.

TM1629B can do 8 digits, each with 14 segments : https://www.lcsc.com/datasheet/lcsc_datasheet_1809191914_TM-Shenzhen-Titan-Micro-Elec-TM1629B_C46800.pdf

So you could use 14 segments to control 2  rows of 7 leds  .... basically you have 16  7 led lines, so you could control 3  7x5 matrixes with a single chip


All these are super cheap, like 10-20 cents ... cheaper than buying the pnp transistors and resistors and use much less space on a circuit board. They're also not that hard to solder on a prototyping board, you could just bend the leads or use a small soic / tssop to DIP adapter board..

There's also DIP versions but more limited ex TM1637 (6 digits, 8 segments per digit) : https://www.lcsc.com/datasheet/lcsc_datasheet_2401261512_TM-Shenzhen-Titan-Micro-Elec-TM1638-TA2009B_C20611979.pdf  or TM1630  (5 digits x 7 segments or 4 digits x 8 segments ) : https://www.lcsc.com/datasheet/lcsc_datasheet_2209231730_TM-Shenzhen-Titan-Micro-Elec-TM1630_C5174529.pdf




There's also dot matrix" led driver chips.  For example, have a look at

IS31FL3738 : https://www.digikey.com/en/products/detail/lumissil-microsystems/IS31FL3738-ZLS4-TR/14308389

It's a 8 x 6 matrix driver, so you could easily do a 7x5 or 8x5  or 8 x 6 matrix - you may want 8 pixels because it makes it easier to make custom "fonts" / "characters"

Compared to the previous drivers, this driver lets you compare the maximum current through a single resistor, then you can set the brightness of each individual led to whatever level you want. So you can for example set the maximum current to something like 40mA because the actual average current on a led will be 1/6 or 1/8 of that)

[kanaron]

Well, I have never heard of LED drivers, you always learn something new I will for sure take a look at this tomorrow, thanks a lot

[mariush]

Yeah, I should add ... there's dot matrix drivers that can go up to 350+ single color leds, I just gave you one that can do more than 7x5 leds and is easier to solder.

There's for example IS31FL3733B that can do 192 leds in a 16 x 12 matrix (but you can rearrange it as 8 x 24 or other patterns) : https://www.digikey.com/en/products/detail/lumissil-microsystems/IS31FL3733B-TQLS4-TR/12675547


If you want to keep it more basic, instead of individual transistors, with your 7x5 arrangement, you have the option of using chips like ULN2003A  which has 7 darlington transistors inside along with the base resistors. 
So basically, you put a voltage on one of the 7 input pins, and that channel is connected to ground. 

Because you're dealing with darlington transistors there's around 1v drop in the chip, so you have to account for that when you calculate the current for the leds.

For example, let's say you power with 5v , your red leds have a 1.8v forward voltage, and you want 20mA (0.02A) per led... then you'd set the current to  5v - 1.8v (drop on led) - 1v (drop in transistor)  = 0.02A x Resistor  ... so Resistor = 2.2v / 0.02A = 110 ohm

You can get these ULN2003A for pennies in DIP packages : https://www.lcsc.com/search?q=uln2003a

If you don't mind surface mount, you can go with the mosfet version ULN2003V12  which replaces the darlington transistors with mosfets, so it doesn't have that 1v drop, otherwise works the same : https://www.lcsc.com/search?q=uln2003v12

It also contains all the gate resistors, esd protections, everything inside.

so you can use uln2003A / uln2003v12 to connect to ground 7 leds, which means you'd rotate through 5 columns giving power to one column at a time.

This means each column may need up to 7 x 20mA = 140mA if you want to limit the single leds to 20mA, so you'd want pnp transistors or p-channel mosfets capable of at least 200mA.


There's loads that can do more than 200mA : https://www.digikey.com/short/mrvj1jb5

For example BC807 can do up to 500mA, and has a hFe of at least 100, so with just 3-5mA you're more or less fully open the transistor.  Basically, I'd just use a 100 ohm resistor on the base, but probably could go up to 330 or so without any issues.


There's dual pnp transistors or dual p-channel mosfets if you want to reduce component count.

[PGPG]

Well, these leds are green and in datasheet it says that voltage drop is 2.6V - 3.4V...

In past all red LEDs worked with about 1.8V drop and green with about 2.0V. Green LED needed higher current to light comparably with red once.
Some years ago new green LEDs appeared. Working with much higher voltage and needing much less current. For signalling I use such green LEDs powering them with around 1mA while those old needed about 10mA.

I prefer to drive LEDs with current source so I don't care if voltage drop is 2.6V or 3.4V. Where from is such big voltage drop difference I don't know. Is it temperature change or manufacturing serie versus serie.

[ledtester]

So... multiplexing by each LED one by one? Then each row will take 7 times longer and entire matrix will takie 35 times longer to multiplex.

With the way you have set up the resistors you should scan the display row by row, not column by column.

Also, if if your LEDs are bright enough with just 1mA of current you can get rid of both NPN and PNP transistors.

The 74HC164 can easily source 1mA per output pin and the HEF4094B can also easily sink 5mA from its output pins.

Datasheets:

https://www.diodes.com/assets/Datasheets/74HC164.pdf
https://assets.nexperia.com/documents/data-sheet/HEF4094B.pdf

[PGPG]

And I think I will go with individual resistor per led.

From that sentence I think you miss something.
When current in LEDs is switched on each of them should have its individual resistor to make their currents be the same. But this doesn't mean that you need as many resistors as you have LEDs. As they are switched in groups you need only so many resistors as there are LEDs in such group (row or column). The same resistors then are used when the next group is switched on.

[kanaron]

From that sentence I think you miss something.
When current in LEDs is switched on each of them should have its individual resistor to make their currents be the same. But this doesn't mean that you need as many resistors as you have LEDs. As they are switched in groups you need only so many resistors as there are LEDs in such group (row or column). The same resistors then are used when the next group is switched on.

Well I missed that I need resistors only on rows and not on columns. Because of multiplexing by columns on row there will be always just one led on or off

With the way you have set up the resistors you should scan the display row by row, not column by column.

Also, if if your LEDs are bright enough with just 1mA of current you can get rid of both NPN and PNP transistors.

I will rearrange resistors to multiplex by columns. Also I'm not sure if 1mA will be enough. I measured that in current scenario I have about 10% duty cycle so I think it is safe to calculate everything for 20mA

Also I'm thinking if I need NPN row transistors... One row will always have maximum one led turned on so one output pin of HEF4094 would have to drive just one led. But then in datasheet input/output current is +-10mA so... I think it is better if I will leave these transistors here

And also thanks a lot for information about that dot matrix led driver. I will have to calculate everything which approach will be better. Dot matrix led driver sounds like perfect solution for... dot led matrix And I think I could use much less pins with this approach and maybe smaller uC, because well... Right now I'm using everything what I can

I paid for whole uC and I will use whole uC

However I have to calculate costs as I think I will have to ship them from US... It seams like they can't be found in Poland... There are eg. already made matrix boards with IS31FL3741 on them but there are none just IS31FL3741 to buy...
But then it is something new for me and will be fun to explore
On the other hand I have used shift registers before, I'm familiar with them, I have already tested this approach on smaller scale and I have code already prepared
Tough decision

[PGPG]

so one output pin of HEF4094 would have to drive just one led. But then in datasheet input/output current is +-10mA so...

You have to learn to read datasheets.
In:
https://www.mouser.pl/datasheet/2/916/HEF4094B-2937862.pdf
at page 6/15 you read that at temperature of 25°C when IC is powered (VDD) with 5V IOL of 0.5mA will rise output Vo to 0.4V (when VDD=10V, IOL=1.3mA will make Vo=0.5V). As you want Vo being lover than 0.4V you can get less current from this IC. As it is CMOS you can assume that it has some resistance at output so if 0.5mA rises output to 0.4V then 0.25mA will rise it to 0.2V.
When VDD will be lower than 5V (I understand your supply is 3V3) the output resistance will be higher so current you will be able to get will be lower.
10mA you found is the maximum value forced by external circuitry. Exceeding it can damage IC.

HEF4094 is not the best selection for this application. It belongs to digital ICs designed when analog circuits were powered by +-15V and digital by 15V. Its switching times are about 75ns (for VDD=10V) and 150ns (for VDD=5V). So 2x lower voltage 2x slower. For 3.3V you should expect 150*5/3.3 = 230ns.
HC4094 will work better here. About 10ns switching times for VCC=4.5V. And you can find that taking 6mA from its output (VCC=4.5V) will rise VOL to 0.26V. So for HEF 0.5mA makes output being 0.4V while here 6mA makes it 0.26V (when ideal value is 0V).

There are faster/stronger digital ICs families (AHC, LVC are both faster and stronger than HC family, and there are more) but I don't know in which of them 4094 are there (may be you need to search other shift register).

[Zero999]

Pure green LEDs with a higher forward voltage 2.6V to 3.4V are marginal at 3.3V.

Is a 5V supply available?

If not, I would just use a low value resistor, or none at all, since the internal wiring will be enough to limit the current.

[kanaron]

Thanks, I had HEF4094 lying around for years so I just used them, but thats right, they are not for this application. From faster families there are available 74HC4094, however I browsed a bit and it looks like I could use 74AHC594. I will check both options and decide and I will have to read about newer IC families

Is a 5V supply available?

If not, I would just use a low value resistor, or none at all, since the internal wiring will be enough to limit the current.

Well this project will have a few different boards (about 20 but others have much less complexity and current consumption) that I wanted to power from one 12V power supply. And each board will have their own 3.3V switching regulator. So I could just easly change one feedback resistor so that this one board would have 5V switching regulator and I would just place LDO from 5V to 3.3V for logic and leave 5V for leds

[PGPG]

I would just place LDO from 5V to 3.3V for logic and leave 5V for leds

HCT logic can work at 5V and accept signals from 3V3 logic.

[kanaron]

True, I meant that I will need LDO for STM32

[kanaron]

So I have changed HEF4094 to 74LVC595A
I got rid of NPN transistors as LVC family can drive 26mA so 1 20mA LED shouldn't be a problem... If I'm not mistaken
I changed PNP transistors to P channel mosfets DMP2004K
I replaced current limiting resistors from columns to rows and I made them 10 ohm just to be sure (for now I left 3.3V supply)
And I left 1k ohm resistors on mosfets gates



Does it look correct now? And do I need pull up resistors for mosfets gates?

[ledtester]

Make sure there isn't a limit for the entire chip.

What datasheet are you looking at?

According to this datasheet:

https://assets.nexperia.com/documents/data-sheet/74LVC595A.pdf

the supply and ground currents are limited to 100mA.

[kanaron]

True, I missed that

But now I hope that it is all correct?



I added N channel mosfets 2N7002AQ on rows with 1k ohm gate resistors

[PGPG]

I added N channel mosfets 2N7002AQ on rows with 1k ohm gate resistors

I wanted to suggest it but had no time to write.

2N7002AQ. Rds=3.5Ω at Vgs=5V. For smaller Vgs no data in table so no guarantee. Chart with typical output characteristic tells little more, but you should think that these are only typical. Looking at Gate Threshold Voltage range you can try to imagine this chart for worse case (when buying you have no guarantee to get typical, but only not worse than described worst).

See DMG1012. Rds=0.4Ω at Vgs=2.5V.

[kanaron]

Thanks, changed to DMG1012

Now I just have to design pcb and later solder everything

[xvr]

I added N channel mosfets 2N7002AQ on rows with 1k ohm gate resistors

Just use TPIC6(A/B/C)595 instead of 74LVC595. MOPSFETs not needed

[Zero999]

Why bother with low on resistance transistors? There are 10R resistors in series with the LEDs. Simply reduce the value of the series resistors to compensate for the on resistance of the transistors. Heck, you might even be able to omit the series resistors.

[PGPG]

Why bother with low on resistance transistors?

OP started this thread from using NPN beta parameter to control LED current. Beta can vary 2 times or more from piece to piece. Good transistor project is made the way to be not dependent on it.
You suggest to use Rds to control LED current. 2N7002AQ Rds is defined as typical 3.5Ω and max 6Ω when Vgs=5V. So what is the minimum? I suppose 2Ω. So parameter changes 3 times.

The next question is what happens when Vgs=3.3V (OP clung to this voltage). It is possible that some of transistors will be really not switched on yet. It wouldn't surprise me if Rds for Vgs=3.3V would be from 4Ω to 40Ω (or even more). Depending on it would be not good design.
If the voltage would be such that serial resistors would be 220Ω and transistors driven by 5V then Rds varying from 2..6Ω can be ignored, but if it can have important influence on LED current then such design shouldn't be accepted.

Heck, you might even be able to omit the series resistors.

Driving npn base with known current can be more precision solution than this.

The art of designing transistor circuits is to make everything depend only on the resistors and not the transistor parameters. I learned it long time ago when to have oscilloscope I had to design and build one.  Imagine if the gain of the Y path could vary as much as transistor beta.