LED array clock--power questions

[Corporate666]

Is there any real advantage to using these,other than less soldering to do? I don't need that much current, even at 8 mA the leds were quite bright, and the 595s can handle that. The 595s daisy chain as well.

Upsides would be...

-With the 595, you're at 64mA on the LED's (plus perhaps a little for the logic), it's rated at 70mA so you're close to it's limit.  I never like running things near their limit.

-With the chips I mentioned, you have higher overall brightness available to you, more than double what you get with the 595

-With the STP08 or TLC5916, you don't need the resistors.  All they do is generate heat and are an additional cost and take up board space.


The STP or TLC are basically drop in replacements for your 595 in terms of how they are driven and the software.  The only difference is you ditch the resistors, flip your LED's the other way around and run them common anode instead of common cathode.

I guess I'd turn the question around... what would the reason be to use the 595's instead of the STP or TLC?  Of course do whatever suits you and if you already have the circuit built or don't want to wait for free samples of the STP/TLC, then your circuit will work - I just think it has some drawbacks compared to a chip designed to drive LED's.

[Corporate666]

Ah, okay. In that case, I'll just use the 595s. I only plan on making 2, and they're likelygoing to be wall clocks anyways, so dimming isn't really a big deal.

You might take a look at the WS2803.  18 pwm outputs,  no resistors, cheap and can be chained. Two wires needed.

Have you used that chip in any designs?  I looked at the datasheet and it seems really convenient that it has 18 ouputs and more importantly clock out!  Why don't any of the other CC LED driver chips do that... it makes laying out the board SOOO much easier without traces that have to go "up and around" each chip.

If you've used these, where did you source them?  I can't find any pricing or availability - does WS sell direct?

[xdbx]

-snip-
I guess I'd turn the question around... what would the reason be to use the 595's instead of the STP or TLC?  Of course do whatever suits you and if you already have the circuit built or don't want to wait for free samples of the STP/TLC, then your circuit will work - I just think it has some drawbacks compared to a chip designed to drive LED's.

My only reason to use the 595s is because I can get all 50 of them for 2 boards for ~$10, and the alternatives are just more expensive. Also, in my experience, samples take a long time, and I couldn't get enough to do even 1 board, so the costs don't really balance out. I know it is a lot of soldering to do, but that's fine.

[edavid]

-With the STP08 or TLC5916, you don't need the resistors.  All they do is generate heat and are an additional cost and take up board space.

What difference does it make if the heat is generated by resistors or the driver chips?

The cost of resistors is much lower than the additional cost of expensive driver chips.

If he cared about board space, he would multiplex

[edavid]

My only reason to use the 595s is because I can get all 50 of them for 2 boards for ~$10, and the alternatives are just more expensive.

Wouldn't Chinese MAX7219s be even cheaper?

[xdbx]

Wouldn't Chinese MAX7219s be even cheaper?

Not that I see, they're somewhat more.

[edavid]

Wouldn't Chinese MAX7219s be even cheaper?

Not that I see, they're somewhat more.

But you can get 10 of them for $5.25, and you only need 3 per clock 

[xdbx]

Oh, i thought it also only had 8 outputs, not 64.

[Corporate666]

-With the STP08 or TLC5916, you don't need the resistors.  All they do is generate heat and are an additional cost and take up board space.

What difference does it make if the heat is generated by resistors or the driver chips?

The cost of resistors is much lower than the additional cost of expensive driver chips.

If he cared about board space, he would multiplex

The heat generated in the driver chip is giving you a constant current LED control.  With the plain shift register, you're guessing and running on the low end because you can't account for variations in LED Vf manufacturing tolerance and temperature variation. 

Expensive is a relative term... STP08 is $0.44 @ Farnell.  75HC595 is $0.20 at Digikey both in qty 100.  $10 more for 50 to use the right part for the job?  I guess expensive is relative.... but ST and TI will both sample for free and at least in the case of ST, they send them next-day.  If size is an issue, it would be better to go with WS2812 and get the shift registers and current control built in.  Multiplexing always loses a substantial amount of brightness - not many (if any) LED's are rated for similar pulsed power compared to constant drive.

But those are just the approaches I'd take - to each his own.

[edavid]

The heat generated in the driver chip is giving you a constant current LED control.  With the plain shift register, you're guessing and running on the low end because you can't account for variations in LED Vf manufacturing tolerance and temperature variation. 

The STP08 only has +/- 8% matching... I doubt he would see more variation than that anyway, if he buys all the LEDs at once.
He can tweak the LED supply voltage to get the current right.

Expensive is a relative term... STP08 is $0.44 @ Farnell.  75HC595 is $0.20 at Digikey both in qty 100.  $10 more for 50 to use the right part for the job?

The cheapest STP08 part I see at farnell.com is STP08DP05XTTR for $0.75 qty 100 (and it's UK stock).
Meanwhile, 74HC595s are $0.10 at Arrow, qty 1.

[Rolo]

Ah, okay. In that case, I'll just use the 595s. I only plan on making 2, and they're likelygoing to be wall clocks anyways, so dimming isn't really a big deal.

You might take a look at the WS2803.  18 pwm outputs,  no resistors, cheap and can be chained. Two wires needed.

Have you used that chip in any designs?  I looked at the datasheet and it seems really convenient that it has 18 ouputs and more importantly clock out!  Why don't any of the other CC LED driver chips do that... it makes laying out the board SOOO much easier without traces that have to go "up and around" each chip.

If you've used these, where did you source them?  I can't find any pricing or availability - does WS sell direct?

I have used the WS2803 in the DIP version, works perfect. I have chained up to 9 to control starburst led display's.
Search for "Lumitron", that's the name of the module. You do not need resistors and V+ LED can be higher than Vcc because it's driving to gnd. So you can place led's in series on one output for bigger display segments. Also your controller does not have to spend time on multiplexing.
The downside is that the power consumption will be higher with the WS2803 because you do not multiplex.
In 2013 I bought these WS2803 in DIP package on Ebay, form China. Paid about $15 for ten pieces. But now it does look the DIP is getting out of fashion, I did a quick search after reading your question and indeed, only the SOP package shows up now. Should you decide to use the SOP please test it first on a breakout/breadboard before production of PCB's. There is an error in the datasheets with the SOP. What I have read on other forums is that the SOP has the same pinout as the DIP. THe datasheet says it is diffferent. I did not use the SOP myself.
I see them fore sale now on Ebay Germany (my neighbour country) and China. It's not that common as some other WS chips.
If you want more info please ask.

[Corporate666]

The STP08 only has +/- 8% matching... I doubt he would see more variation than that anyway, if he buys all the LEDs at once.
He can tweak the LED supply voltage to get the current right.

The 8% is the worst case on one specific drive level that he would not be using.  At the level he would be using, it's 3% max, 1.5% typical.  An LED is going to have a LOT more variation in Vf than that, even just due to temperature variation - no way is he going to get close to 3% or even 8% with voltage tweaking.  Looking at a typical 5mm Avago LED, the Vf shift is about 0.1 volt per 20 degrees.  He'll want to do as little dissipation as possible, so if the Vf of his LED's is 1.9@10mA and he wants to string 6 of them in series (just as an example), and run them off 12V... that's a 60 ohm resistor.  Over a temperature change of 20 degrees, that's 0.1V per LED which doubles his current to 20mA - a huge difference.  And it doesn't account for inherent differences in LED forward voltage from manufacturing.  Sure, you can tweak each string and pick resistors for each if you want - but who has time or motivation for that?  But the light output difference between 10 and 20mA will be significant. 

The cheapest STP08 part I see at farnell.com is STP08DP05XTTR for $0.75 qty 100 (and it's UK stock).
Meanwhile, 74HC595s are $0.10 at Arrow, qty 1.

STP08DP05MTR - 28p (about $0.44)/100.  Or no reason not to go with the 16 channel version like STP16 or similar for 60-70 cents/ea @ Verical/Mouser/Digikey... then he only needs 25 of 'em.  Or if price is an issue, they can be had for 30 cents/ea w/free shipping on eBay.  And not to be pedantic, but just to be pedantic  the 74HC595 is $0.10 for 50pcs - and that's actually a really good price, I checked Octopart and for some reason Arrow doesn't show up - next best price I see is Avnet $0.26/50.  Lesson learned - always check the distributors directly before trusting Octopart or Findchips.

Hey, I'm not saying the CC shift register is going to be cheaper than a 74HC595... but it seems a really small price difference (what, a few bucks?) for doing it the right way with several added benefits to boot. 

[Corporate666]

Ah, okay. In that case, I'll just use the 595s. I only plan on making 2, and they're likelygoing to be wall clocks anyways, so dimming isn't really a big deal.

You might take a look at the WS2803.  18 pwm outputs,  no resistors, cheap and can be chained. Two wires needed.

Have you used that chip in any designs?  I looked at the datasheet and it seems really convenient that it has 18 ouputs and more importantly clock out!  Why don't any of the other CC LED driver chips do that... it makes laying out the board SOOO much easier without traces that have to go "up and around" each chip.

If you've used these, where did you source them?  I can't find any pricing or availability - does WS sell direct?

I have used the WS2803 in the DIP version, works perfect. I have chained up to 9 to control starburst led display's.
Search for "Lumitron", that's the name of the module. You do not need resistors and V+ LED can be higher than Vcc because it's driving to gnd. So you can place led's in series on one output for bigger display segments. Also your controller does not have to spend time on multiplexing.
The downside is that the power consumption will be higher with the WS2803 because you do not multiplex.
In 2013 I bought these WS2803 in DIP package on Ebay, form China. Paid about $15 for ten pieces. But now it does look the DIP is getting out of fashion, I did a quick search after reading your question and indeed, only the SOP package shows up now. Should you decide to use the SOP please test it first on a breakout/breadboard before production of PCB's. There is an error in the datasheets with the SOP. What I have read on other forums is that the SOP has the same pinout as the DIP. THe datasheet says it is diffferent. I did not use the SOP myself.
I see them fore sale now on Ebay Germany (my neighbour country) and China. It's not that common as some other WS chips.
If you want more info please ask.

Thanks for the info!

I've used literally tens of thousands of the ST Micro and Toshiba parts, and had been using the Allegro Micro parts for higher pin counts, but I've noticed Allegro has discontinued many of their parts.  I'd never heard of World Semi before, but I have used their WS2812 LED's on a few projects with success, so I guess they are becoming mainstream enough for me to trust designing their parts into something.  That's funny about the datasheet - (excuse me if it sounds rude to anyone), but that's typical Chinese supplier stuff - having a datasheet with an erroneous pinout!  Too funny!

I'd be using SOP parts anyway... I was mostly just curious about suppliers.  For as huge as the WS2812 LED's have become in the west, I don't see any of the distributors carrying World Semi parts... I can get some samples on eBay but I'll maybe contact WS direct and see what their volume requirements are like to buy for production work.

Thanks again!

[edavid]

The 8% is the worst case on one specific drive level that he would not be using.  At the level he would be using, it's 3% max, 1.5% typical.  An LED is going to have a LOT more variation in Vf than that, even just due to temperature variation - no way is he going to get close to 3% or even 8% with voltage tweaking.  Looking at a typical 5mm Avago LED, the Vf shift is about 0.1 volt per 20 degrees.  He'll want to do as little dissipation as possible, so if the Vf of his LED's is 1.9@10mA and he wants to string 6 of them in series (just as an example), and run them off 12V... that's a 60 ohm resistor.  Over a temperature change of 20 degrees, that's 0.1V per LED which doubles his current to 20mA - a huge difference.  And it doesn't account for inherent differences in LED forward voltage from manufacturing.  Sure, you can tweak each string and pick resistors for each if you want - but who has time or motivation for that?  But the light output difference between 10 and 20mA will be significant. 

I don't understand this analysis.  For the OP's clock, there won't be any LEDs in series.  All the LEDs will be at the same temperature.  Even with 5% resistors, there won't be any significant brightness variation.

Constant current drivers are neat, and I have used the STP04 and STP08 parts, but I don't see any advantage here.

The cheapest STP08 part I see at farnell.com is STP08DP05XTTR for $0.75 qty 100 (and it's UK stock).
Meanwhile, 74HC595s are $0.10 at Arrow, qty 1.

STP08DP05MTR - 28p (about $0.44)/100.

OP has a US flag... here's the link for STP08DP05MTR at Newark/element14/Farnell: http://www.newark.com/stmicroelectronics/stp08dp05mtr/led-driver-8bit-30mhz-nsoic-16/dp/38Y7955?ost=38Y7955&CMP=AFC-SF-T11-US

The price is $0.797 ea plus $20 handling charge for UK stock

And not to be pedantic, but just to be pedantic  the 74HC595 is $0.10 for 50pcs - and that's actually a really good price, I checked Octopart and for some reason Arrow doesn't show up - next best price I see is Avnet $0.26/50.  Lesson learned - always check the distributors directly before trusting Octopart or Findchips.

You are right about the multiple... they fooled me by having an odd number in stock.  Findchips does include Arrow.

Here they are at Newark for $0.13 qty 1:  http://www.newark.com/nxp/74hc595d-112/8-bit-sipo-siso-shift-register/dp/71R2679

[Corporate666]

I don't understand this analysis.  For the OP's clock, there won't be any LEDs in series.  All the LEDs will be at the same temperature.  Even with 5% resistors, there won't be any significant brightness variation.

Constant current drivers are neat, and I have used the STP04 and STP08 parts, but I don't see any advantage here.

OK, so with single LED's per output, around the same 60 ohm resistor, with temperature variation of 20 degrees he's getting a 20% difference in brightness.  Noticeable?  Probably.  That's an order of magnitude worse than an LED driver chip.  That's why LED driver chips exist, because it's the right way to drive LED's, because treating them as CV devices leads to current swings which affect brightness substantially.  The other advantages are that he gets rid of the resistors, he has simple brightness control, he can achieve much higher brightness (not to say it must be used, but it's there).  It's much more tolerant of ripple voltage or any other change in LED supply voltage.  And he's not running the chip right at it's max rated spec.  The only singular downside is the price difference.

OP has a US flag... here's the link for STP08DP05MTR at Newark/element14/Farnell: http://www.newark.com/stmicroelectronics/stp08dp05mtr/led-driver-8bit-30mhz-nsoic-16/dp/38Y7955?ost=38Y7955&CMP=AFC-SF-T11-US

The price is $0.797 ea plus $20 handling charge for UK stock

http://uk.farnell.com/stmicroelectronics/stp08dp05mtr/led-driver-8bit-30mhz-nsoic-16/dp/2460706?CMP=GRHB-OCTOPART

$0.44/ea in 100's. 

Or just use the 16 output version that Newark sells for $0.414

http://www.newark.com/stmicroelectronics/stp16cpc26ptr/power-conversion/dp/73T2457?CMP=AFC-OP?gross_price=


He said he needed 25 of the 74HC595's... $3.25 @ Newark vs. $6.60.  Three bucks to do it the right way is a bridge too far?  For a cheap commercial product where you're counting pennies and running on batteries fine... but for an at-home product, especially when the company will send you the actual parts for *free*, I just see no reason not to do it the right way, unless the other parts are sitting on a shelf and have an expiration date on them or something.  On the other hand, he said he had the 595's on the shelf, so the debate is rather academic - but CC is absolutely the way to drive LED's.

You are right about the multiple... they fooled me by having an odd number in stock.  Findchips does include Arrow.

Strange... I use Octopart daily, and I ran a couple of other parts I use - CY8C21123 and MBR0520LT1 and arrow shows up on both.  The search results are instant, so Octopart must be caching their database and presumably scraping or somehow getting the data from the distributor sites.  Curious that some parts that are in stock at some distributors simply do not show up on Octopart.

[PICmonster]

I would use a different approach using LEDs that have a built in controller, the WS2812B for example, these LEDs are daisy chained together, simply orient them to create the numericals on your clock.


Here's an example



The advantages are that you can change color of any segment and alter its brightness.

[edavid]

OK, so with single LED's per output, around the same 60 ohm resistor, with temperature variation of 20 degrees he's getting a 20% difference in brightness. 

Where do you get 60 ohm?  5V supply, red LED, 8mA current -> 390 or 360 ohm resistor.
Why would there be any temperature variation?  All the LEDs are in the same place and have the same drive current.

http://uk.farnell.com/stmicroelectronics/stp08dp05mtr/led-driver-8bit-30mhz-nsoic-16/dp/2460706?CMP=GRHB-OCTOPART

$0.44/ea in 100's. 

Why are you quoting UK prices when OP has a US flag set?

He said he needed 25 of the 74HC595's... $3.25 @ Newark vs. $6.60.

Since he's in the US, it's $3.25 vs. $39.93