GS8208 based Pixels - testing and evaluation 2022

AussiePhil

Dedicated elf
Administrator
Generous elf
Joined
Jun 20, 2009
Messages
1,606
Location
Canberra, ACT, Australia
This thread will be about 2022 testing of the GS8208 based pixels and like Ryan's excellent thread some comparison numbers.
Please read all the great info in Ryan's thread for more than I will type up.
The thread also contains links to four other threads about these pixels.

To long don't want to read:- these should be you pixel of choice, end debate. The 5v versus 12v debate is dead with these.

So for those that haven't read the other threads yet, what is a GS8208 Pixel.
This is from the Datasheet: My own comments in blue.

  • The GS820 8 is a 3 channel constant current LED drive with redundant data transfers and internal display patterns. Internal test patterns are absent on my strip
  • The range of input power is from +9V to +15 V, and voltage endurance of LED port is +12 V. Minimum voltage will be tested
  • There is a built in 12bits GAMMA correction module. This is an added feature over ws2811 and should provide smoother dimming and better colour at low levels.
  • PWM maximum refresh frequency is 8 kHz. This is higher than listed for a WS2811
  • GS8208 provides two signal data input as redundant control, which ensures the transmission of the signal if any single chip fails. same as ws2813 IC's that implement redundant data inputs
  • Current is constant regardless of colour being displayed including full white.
  • 2811 compatible protocol support for input at least, output to be tested.

Initial review and data on 5M 12v 150 pixels strip.

Strip was ordered from two different vendors, one has arrived so far.

I ordered IP30, ie not covered strip so I could have easy access to measure voltages along the strip and to keep the price cheap :)

12v 30Leds/m 5m total length. Strip came with 4 pin JST connectors. quite honestly this is no different to any other led strip you will find on various sites except it has four wires in and out.

First thing was to power it up and measure the standing/Quiescent current and watch the test mode kick in.
First surprise was, no test mode kicked in - maybe there is a later chip revision that has removed it.
So standing current, this is the current when the strip is powered on doing nothing
Measured - 320mA. This seemed high but as the GS8208 has the equivalent of a 78l05 built in and will be on par with other regulated pixels.
So a look a the datasheet says each pixel is 2.1mA, we have 150 of them so 150 x 2.1 = 315mA.
Seems that we are right on spec at this point but this needs to be taken into account when you use thousands of these the same as any other 12v pixel.

Now it gets interesting:
We are all used to the current draw increasing with mixed colours and white .... these do NOT do this and this makes the difference in planning and optimisation of power.
To borrow a table from one of the other testing threads to emphasize the point.
Note 150 LED's used here on strip, actual voltage drops may be different for strings.

ColourVoltage InVoltage EndCurrentAverage Pixel Current mA
Red1210.71.6310.9
Green1210.71.6310.9
Blue1210.71.6310.9
Purple1210.71.6310.9
Cyan1210.71.6310.9
Yellow1210.71.6310.9
White1210.71.6310.9

So the spec sheets lists 18.5mA... that is a MAX number.
at 12v the spec sheet says 11mA
Seems we again on target and in agreement with the spec sheet

Why is this the giant killer and why should we adopt it.


Since I started in pixels back in 2009 the nirvana pixel would use around 10mA max per colour to save on power, power inject etc, well these use 11ma total full stop.
  • Power requirements are constant regardless of colour allowing better planning of power requirements and any injection.
  • Low current draw means longer runs before injection
  • more pixels per port from your controller before power
  • 12v means less injection needed
  • Pixels per 350w supply jumps to 2,500 from the current guideline of 1000
  • Redundant data.
Cons
  • Hard to find as pixel nodes
  • Pricing is all over the place and can be up to twice the price of 2811 nodes
Ray Wu has GS8208 bullet nodes at similar pricing to 12 resistor based nodes so if that holds and supply can be improved then these really should be a no brainer.

Still to test, voltage input levels and 2811 compatibility on the output.

Now to order some nodes for more testing
 

AussiePhil

Dedicated elf
Administrator
Generous elf
Joined
Jun 20, 2009
Messages
1,606
Location
Canberra, ACT, Australia
Some more data and observations afterwards

This graph shows the voltage seen at the end of 5m of strip for brightness levels from 1% to 30%
the little bumps are just reading errors only grabbing to 2 decimal places
Visually this was very smooth with the LEDS lighting up from 1%. The gamma correction would seem to be effective

1644999213758.png

This graph shows the voltage seen at the end of 5m of strip for brightness levels from 1% to 100% in 5 steps at a time
The two lines meeting is just a coincidence due to the numbers.
This was also very linear.

1644999232844.png

Now here is something you cant do with resistor based pixels.
The blue and orange lines show the input and end of string voltages from 7v to 15v input.
At around 9v in 8v out the current stopped rising fast and largely stayed unchanged to 15v input.
This shows how effective the constant current driver is with the LED.

1644999251397.png

Observations:
  • At least for the IC in LED strip based version they are very tolerant of higher voltage up the max specified in the datasheet and the 9v is really the lower limit.
  • only having a low limit of 9v isn't a real world issue due to the overall lower voltage drop from less current draw.
  • based on the numbers two 5m strips (10m) would be a non issue.
  • heat may be an issue for 10m with over 3A going into the first string (untested yet)
Let me know if anyone wants any other tests.

Cheers
Phil
 

Skymaster

Crazy elf
Generous elf
Joined
Dec 19, 2021
Messages
208
Location
Western Sydney

ryanschristmaslights

Senior elf
Administrator
Joined
Jun 30, 2010
Messages
822
Location
Adelaide, Australia
The GS820 8 is a 3 channel constant current LED drive with redundant data transfers and internal display patterns. Internal test patterns are absent on my strip

Interesting, do you see any transient flashes of red and/or white at the point that you apply power? Example of power on behaviour (0:00 to 0:06):

View: https://www.youtube.com/watch?v=JzdDf8rCmCI

(N.B. The Red/green/white waving animation is a playing xLights sequence and not the built-in test pattern)

GS8208-Ray-Order-2018-30L-per-mtr-strip-white-40-percent-intensity-9-meter-total-length.jpgThe strip-based 30L/m GS8208 that I ordered through Ray in 2018 exhibits the same red flash at power on as the 2021 Shiji bullets in the video above. I use these strips for front door & window outlines plus leaping arches.

My longest run of strip is 9m with four lengths clockwise around a window divided into 3m -> 1.5m -> 3m -> 1.5m sections, with Ray's flat wire 13.5mm pigtails used at the start and end of each section. With power supplied at the first pixel only and with a 40% intensity limit (which is how I've run these strips every season since adding them in 2018), to my eyes the last pixel (#270) looks close to or perhaps even the same brightness as the first pixel.


Now to order some nodes for more testing
Do you have a preference between bullets vs square node pixel strings or not fussed?

As indicated in my earlier thread, Shenzhen Shiji Lighting Co., Ltd via Alibaba is the supplier I ordered the bullet pixels from (though they aren't the only supplier). Could just be a coincidence but Shiji was also mentioned in a separate forum thread by @queena who is associated with Genesis Systech (the IC design company). ScottLED can also make them specially if asked (they do not normally hold stock of the GS8208 IC) however their MOQ for this is 5000 pieces (e.g. 100x 50 pixel strings). Rita Lighting (Ray) and ETOP LED were not able to make them when asked last year. Whether that will change one day, I don't know.
 

AussiePhil

Dedicated elf
Administrator
Generous elf
Joined
Jun 20, 2009
Messages
1,606
Location
Canberra, ACT, Australia
Hi Phil,

Data sheet says this:
Under the internal display mode, the first chip’s SDI/SDI2 needs to be connected to GND.

Could it be because the controller was holding it high and/or floating?

Link to the DS I'm reading: http://www.normandled.com/upload/201805/GS8208 LED Datasheet.pdf
I've lost track of the times i have read that DS and missed that line in it. in the strip i have, both inputs are floating, i'll ground one and see what happens.

Cheers
Phil
 

AussiePhil

Dedicated elf
Administrator
Generous elf
Joined
Jun 20, 2009
Messages
1,606
Location
Canberra, ACT, Australia
Interesting, do you see any transient flashes of red and/or white at the point that you apply power? Example of power on behaviour (0:00 to 0:06):
Nothing no flash, no test pattern, initially thought they were not gs2808
View attachment 19717The strip-based 30L/m GS8208 that I ordered through Ray in 2018 exhibits the same red flash at power on as the 2021 Shiji bullets in the video above. I use these strips for front door & window outlines plus leaping arches.
Suspect we no know the reason will test the theory tomorrow.
My longest run of strip is 9m with four lengths clockwise around a window divided into 3m -> 1.5m -> 3m -> 1.5m sections, with Ray's flat wire 13.5mm pigtails used at the start and end of each section. With power supplied at the first pixel only and with a 40% intensity limit (which is how I've run these strips every season since adding them in 2018), to my eyes the last pixel (#270) looks close to or perhaps even the same brightness as the first pixel.
Matches what i would expect from testing the strip and yeah intensity from first to last seems to hold up well.
Do you have a preference between bullets vs square node pixel strings or not fussed?
Bullet due to the post testing use, but square would work as well
As indicated in my earlier thread, Shenzhen Shiji Lighting Co., Ltd via Alibaba is the supplied I ordered the bullet pixels from (though they aren't the only supplier). Could just be a coincidence but Shiji was also mentioned in a separate forum thread by @queena who is associated with Genesis Systech (the IC design company). ScottLED can also make them specially if asked (they do not normally hold stock of the GS8208 IC) however their MOQ for this is 5000 pieces (e.g. 100x 50 pixel strings). Rita Lighting (Ray) and ETOP LED were not able to make them when asked last year. Whether that will change one day, I don't know.
I had 500 in the cart from Ray but he has removed the listing, not happy at all. looking at alternatives

However as i have typed this got a message back from Shenzen Rita (aka Ray) with apologies and they will become available in Mar.
Usual suspicion that the listing was an old one and they have now gone to one of the others to get them made and the pricing will go up. six weeks to the end of march so will pursue the other options as well and see what happens.

Supply may well be the big negative unless some demand happens.

Cheers
Phil
 

MichaelF5

Apprentice elf
Generous elf
Joined
Nov 23, 2016
Messages
76
Location
West Hoxton
I'd love to sacrifice one to see what the limits of the voltage are, would also love to see some sort of video showing how the Gamma helps (side by side with the alternatives? )
 

AussiePhil

Dedicated elf
Administrator
Generous elf
Joined
Jun 20, 2009
Messages
1,606
Location
Canberra, ACT, Australia
I'd love to sacrifice one to see what the limits of the voltage are,
There is absolutely no need to push these much past 13/14v ever as you run out of port current capacity and the rest of the voltage just get dumps as heat anyway.... it would be better to run these in line other other 12v stuff at 12v, and you still get a longer run before injection and if you need to run local point of use dc-dc convertors.

would also love to see some sort of video showing how the Gamma helps (side by side with the alternatives? )
as soon as i have some nodes in hand i can do the Gamma video comparision
 

ryanschristmaslights

Senior elf
Administrator
Joined
Jun 30, 2010
Messages
822
Location
Adelaide, Australia
as soon as i have some nodes in hand i can do the Gamma video comparision

Should be good alongside @BradsXmasLights strip colour rendering comparison 👍



Brad has a couple of other videos too including one of the built-in test mode, if they are of any interest:
 

AAH

I love blinky lights :)
Community project designer
Joined
Dec 27, 2010
Messages
4,042
Location
Eaglehawk
Excellent read and tests. I'm working on getting these to stock if they can be had for less than the cost of unobtanium. They have so many things going for them.
The chip comes in 2 current settings from memory. I was aiming to stock the 10/11mA or whatever they are rather than the higher current ones as it seems that most people are now running at lower brightness due to the actual brightness being too high especially in high density props. This means that instead of 80-130 dimming levels you'd get to use the full 255.
 

AussiePhil

Dedicated elf
Administrator
Generous elf
Joined
Jun 20, 2009
Messages
1,606
Location
Canberra, ACT, Australia
Excellent read and tests. I'm working on getting these to stock if they can be had for less than the cost of unobtanium.
Great news Alan.
They have so many things going for them.
The chip comes in 2 current settings from memory.
I have found no reference to two different chips. The Normand DS we all seem to be working from is not the best with the wordings in the Electric Characteristics table but based on measurements the IOUT current at VDH12v is measured at 11mA. This would say the normal operating condition is 11mA.
This is in line with the table.
The fact that only a slight rise in current consumption was seen at 15v supports the lock in around 11mA. The 18.5mA maximum rating is no doubt to cover manufacturing tolerances and we really should just ignore this number.

I was aiming to stock the 10/11mA or whatever they are rather than the higher current ones as it seems that most people are now running at lower brightness due to the actual brightness being too high especially in high density props. This means that instead of 80-130 dimming levels you'd get to use the full 255.
I've heard people now say that they never run over 20%, that is 20 brightness steps in software, the first two are generally useless for 2811 anyway so they get 18 grey scale % steps and crappy colour mixing.
Visually to me at least the GS8208 was visible changing all the way up to around 70% and with the gamma curve was effective from 1%.
Even limiting to 70% is a vast improvement.
 
  • Like
Reactions: AAH

AussiePhil

Dedicated elf
Administrator
Generous elf
Joined
Jun 20, 2009
Messages
1,606
Location
Canberra, ACT, Australia
Quick update:
Tested connecting between the GS8208 strip and a 100 cnt 12v ws2811 bullet pixels.
GS8208 and WS2811 both purchased late 2021 so will have >280uS reset timing compatibility.

Falcon f16v2 fw 1.09 as the 2811 protocol source.

Falcon -> GS8208 -> WS2811 - PASS
Falcon -> WS2811 -> GS8208 - PASS

Maybe some of the interconnect issues reported have been with older WS2811 pixels using the short (original) reset timing less than 280uS

Cheers
Phil
 
Top