pretty sure i let out the smoke

[RNoble]

Paul has already given you a bunch of suggestions all through this thread.

yes.
Paul suggested a resistor and diode system that addresses the more expensive problem of killing Teensy boards which i thoroughly explored multiple ideas including the Shottky diode design i found when AI referenced it that i thought might do better with a hybrid solution. i'm grateful for that new knowledge.

i then looked at the thought in conjunction with the flickering and concern that my 12V nominal system (which is actually 12.5V to 15V) would require regulation (thus the MOSFET circuit) and i started asking new questions to add to my already overwhelming number of questions which i asked for help to answer.

so now i found another idea that will do the same thing with a lot less time and cost invested in experimentation that i don't have the resources for that should die immediately if 12V+ is sent through it.

feel free to augment my already lacking knowledge on this subject.
if there is a better chip to use, let me know, but i've moved on with these thoughts because i believe that the level shifter is a more holistic approach.

 

[PaulStoffregen]

I'll repeat again, if you use the schematic from msg #43, I believe you should add 6 schottky diodes. Each should have its anode side connected to the wire between the 74HC245 pin and 330 resistor. All cathodes should connect to +5V.

I believe you should also add 1 zener diode between +5V and GND, with the cathode (side with the stripe) connected to +5V.

If you're making a PCB, you could add these 7 parts to the design and later decide whether or not to actually solder them. But if you don't add them to the PCB design now, adding them later would be a difficult and kludgy matter, or another PCB.

If you decide to go with the AI suggestion from msg #47 "the chip is ~$1 and does both the the bump in signal voltage and acts as a sacrificial part", you still should add the zener diode between +5V and GND. And of course, use a DIP package part so you can solder a socket to the PCB.

 

[RNoble]

I'll repeat again, if you use the schematic from msg #43, I believe you should add 6 schottky diodes. Each should have its anode side connected to the wire between the 74HC245 pin and 330 resistor. All cathodes should connect to +5V.

I believe you should also add 1 zener diode between +5V and GND, with the cathode (side with the stripe) connected to +5V.

If you're making a PCB, you could add these 7 parts to the design and later decide whether or not to actually solder them. But if you don't add them to the PCB design now, adding them later would be a difficult and kludgy matter, or another PCB.

If you decide to go with the AI suggestion from msg #47 "the chip is ~$1 and does both the the bump in signal voltage and acts as a sacrificial part", you still should add the zener diode between +5V and GND. And of course, use a DIP package part so you can solder a socket to the PCB.

i'm using prototype boards.
and a socket is absolutely part of the design.

are you saying that the IC won't stop a 12V incursion to the Teensy?

 

[PaulStoffregen]

The only way to really know is to actually test. I tried to show how to use a solderless breadboard to do this sort of experimentation quickly and easily. Or it can be done with computer simulation, but getting accurate simulation models for this stuff is far more trouble than just trying it quick on a breadboard.

We can guess here over and over, and even worse you can ask AI chat bots, but that's all just blind guessing.

But if you want a blind guess anyway, if you accidentally drive 12 volts onto an output pin, the 74HCT245 could fail in a number of different ways. My guess for the most likely would be conducting the 12 volt power to the 5 volt power input pin, which would drive 12 volts onto both other boards at their power inputs. That's why I suggested adding a zener diode.

However, the way that 74HCT245 acts is a big unknown. It could hypothetically end up doing anything. I do not know. Whatever AI you're using almost certainly does not know either, but those chat bots are designed to always give an answer and make you feel good about it, even if available info is spotty or the AI just makes it up.

Testing the actual part is the way to really find out. That's why solderless breadboards exist, so you can quickly plug the parts and wires in and experiment to find out what really happens.

 

[PaulStoffregen]

However, if you were to use my suggestion to add 7 diodes, I do have a pretty good idea of what will really happen. It'll very likely be similar to the test I tried where the zener diode wasn't doing much. The 1 extra zener diode from 5V to GND is meant to deal with the case where total current from 12V goes through those schottky diodes ends up more than Teensy and the ESP board are consuming. I also tried to explain that earlier.

That's my advice for the 74HCT245 level shifter, as well as generic advice to actually test as I tried to show.

I'm feeling like I've poured way too much time into this conversation and I'm getting the impression you don't want to follow my advice, which is your choice, but whatever you choose I hope the project works out well and you avoid more damaged hardware.

 

[RNoble]

The only way to really know is to actually test. I tried to show how to use a solderless breadboard to do this sort of experimentation quickly and easily. Or it can be done with computer simulation, but getting accurate simulation models for this stuff is far more trouble than just trying it quick on a breadboard.

We can guess here over and over, and even worse you can ask AI chat bots, but that's all just blind guessing.

But if you want a blind guess anyway, if you accidentally drive 12 volts onto an output pin, the 74HCT245 could fail in a number of different ways. My guess for the most likely would be conducting the 12 volt power to the 5 volt power input pin, which would drive 12 volts onto both other boards at their power inputs. That's why I suggested adding a zener diode.

However, the way that 74HCT245 acts is a big unknown. It could hypothetically end up doing anything. I do not know. Whatever AI you're using almost certainly does not know either, but those chat bots are designed to always give an answer and make you feel good about it, even if available info is spotty or the AI just makes it up.

Testing the actual part is the way to really find out. That's why solderless breadboards exist, so you can quickly plug the parts and wires in and experiment to find out what really happens.

i have a breadboard. it came with the Uno kit.

i don't have the hundreds to thousands of dollars of resources your test bench has and i've asked questions that have gone not only unanswered, but straight up ignored, so i ask other sources. you cannot fault me for that.

However, if you were to use my suggestion to add 7 diodes, I do have a pretty good idea of what will really happen. It'll very likely be similar to the test I tried where the zener diode wasn't doing much. The 1 extra zener diode from 5V to GND is meant to deal with the case where total current from 12V goes through those schottky diodes ends up more than Teensy and the ESP board are consuming. I also tried to explain that earlier.

That's my advice for the 74HCT245 level shifter, as well as generic advice to actually test as I tried to show.

I'm feeling like I've poured way too much time into this conversation and I'm getting the impression you don't want to follow my advice, which is your choice, but whatever you choose I hope the project works out well and you avoid more damaged hardware.

this just adds to the frustration of not having questions addressed.

i appreciate what information you have given me and i want to follow the advice, but half of the equation is being ignored.

i don't understand the reasoning for not addressing the level shifter question at the beginning of the conversation. i'm not a blind follower and i'm working through all of this testing alone, without the resources of a test bench or the decades of knowledge needed to completely understand what is happening.

for instance...
every source i can find says that the WS2811 COB LED strips and WS2815 LED strips i am using are suppose to have no connection to the power and data lines.

however, since i was curious, i took my multimeter to task and tested the resistance of the strips that haven't been put on the project. on the WS2811 COB, the power to ground has a resistance of just under 2K, data to ground has a resistance just over 2K, and power to data has a resistance at 8K; the WS2815 is similar with 7K power to ground, 8K for both back up and data to ground, and 25K power to back up and 30k power to data.

to top it off, i have a some sort of phantom ground when i connect my Uno test package, keeping the LEDs lit when i take the ground off of the 12V power. none of which can be explained.


what i am trying to accomplish is the complete isolation of all 12V sources from the control system.
it sounds like that isn't possible. if it's not possible, i want a hard, cheap fail safe to cut the circuit, hard, so that i can trace the fault and fix it.

so, here is the question:
will either of these solutions do that?
if neither solution does that, what will?

 

[h4yn0nnym0u5e]

what i am trying to accomplish is the complete isolation of all 12V sources from the control system.
it sounds like that isn't possible.

Well … apart from the suggestion of opto isolators, made over a week ago in post #18, and expanded on with specific part numbers a few posts later. For what it’s worth, that’s absolutely the way I’d go.

 

[RNoble]

Well … apart from the suggestion of opto isolators, made over a week ago in post #18, and expanded on with specific part numbers a few posts later. For what it’s worth, that’s absolutely the way I’d go.

this mention?

Perhaps opto-isolators would be a better option. Not sure if their speed is fast enough to run LEDs but at least they provide constant isolation.

that's going to come down to which ones.
supposedly, the circuits are running somewhere around 800kHz. i'd probably have to use a digital one.

 

[h4yn0nnym0u5e]

that's going to come down to which ones.

Which were suggested in post #24:

Most optocouplers are too slow for WS2812 / WS2815 data. If you're going to go that route, you probably want a higher speed RF-based isolator, like ISO6420 or CMT8020N0 or similar. These isolators can also serve to convert logic levels.

Your response (after a digression into isolation transformers - fine for audio, no use for 800kHz digital) being

that's a whole new can of worms...

Well, yes. You need a new can, the old ones are smoking your Teensys. Several flavours of worm have been suggested, pick the one you think you might like

how many of those would i have to have to handle all the circuits?

I’ll grant you, that didn’t get a response. But following the provided links would have told you how many channels each isolator has, then you could have done the maths.

You need two things, level shifting and protection for the Teensy. The 74HCT245 (or similar) gives you the former; the various diode / resistor arrangements give you the latter - that’s one route. Opto isolators do both at once - another route, with complete isolation up to several kV. The Teensy then needs a separate power source, of course, that’s what happens with complete isolation…

Any protection circuit can fail under sufficient stress, with failure defined in the range from “circuit stops working“ to “allows the protected device to be destroyed“. My reason for picking an opto is that the stress level required for destruction is pretty high, by design. But it’s always going to be at least a bit of a sacrificial element - ideally one that’s easier and cheaper than a Teensy to replace.

Of course, nothing is going to protect against user error during development. If you accidentally connect 12V directly to the Teensy by mistake, it’ll die. But if you develop on the bench and only introduce the protected side to the target system later, via a connection which can’t possibly interact with the Teensy, you minimise the risk.

 

[RNoble]

Which were suggested in post #24:

Your response (after a digression into isolation transformers - fine for audio, no use for 800kHz digital) being

Well, yes. You need a new can, the old ones are smoking your Teensys. Several flavours of worm have been suggested, pick the one you think you might like


I’ll grant you, that didn’t get a response. But following the provided links would have told you how many channels each isolator has, then you could have done the maths.

You need two things, level shifting and protection for the Teensy. The 74HCT245 (or similar) gives you the former; the various diode / resistor arrangements give you the latter - that’s one route. Opto isolators do both at once - another route, with complete isolation up to several kV. The Teensy then needs a separate power source, of course, that’s what happens with complete isolation…

Any protection circuit can fail under sufficient stress, with failure defined in the range from “circuit stops working“ to “allows the protected device to be destroyed“. My reason for picking an opto is that the stress level required for destruction is pretty high, by design. But it’s always going to be at least a bit of a sacrificial element - ideally one that’s easier and cheaper than a Teensy to replace.

Of course, nothing is going to protect against user error during development. If you accidentally connect 12V directly to the Teensy by mistake, it’ll die. But if you develop on the bench and only introduce the protected side to the target system later, via a connection which can’t possibly interact with the Teensy, you minimise the risk.

i got the impression this wasn't viable and moved on.
it would probably help if i had far more knowledge than i do. now that i'm looking harder, i'm seeing six channel opto isolators.

so i'll now ask this question:
if i use an opto isolator, to i need to shift the levels at all?

if i'm understanding the technology correctly, the signal is transmitted across the boundary through light rather than through electricity. theoretically, the opto isolators should send the correct signal voltage downstream.

 

[MarkT]

There are 3 main types of isolator, optical, capacitive and magnetic. Of the optical sort there are standard phototransistor types (all very slow), and faster ones using photo-diodes and amplification. Capacitive and magnetic ones are usually fast (logic speeds) but read the datasheet - some have different speeds on different pins even. Almost all have different delays for rising and falling edges, so some waveform distortion is going to happen, again the datasheet will detail this.

 

[PaulStoffregen]

if i use an opto isolator, to i need to shift the levels at all?

I specifically addressed this back in msg #24.

 

[RNoble]

I specifically addressed this back in msg #24.

clearly, i missed that.
i'm going to attribute it to the way i process information.

thank you for pointing that out.

There are 3 main types of isolator, optical, capacitive and magnetic. Of the optical sort there are standard phototransistor types (all very slow), and faster ones using photo-diodes and amplification. Capacitive and magnetic ones are usually fast (logic speeds) but read the datasheet - some have different speeds on different pins even. Almost all have different delays for rising and falling edges, so some waveform distortion is going to happen, again the datasheet will detail this.

it looks like i could use ISO7710, ISO6760-Q1, or ISO6760L pretty easily. the ISO6760-Q1 seems to be the best option for this. but looking at the Si866x series and decided i have more questions...

what does "isolated power" mean in terms of these chips? (yes, i googled it. i'm looking for confirmation because i'm not sure i understand.)

 

[MarkT]

It should mean the transmitter and receiver side are galvanically isolated. Typically TX and RX both need supplies, often denoted Vcc1 and Vcc2.

 

[RNoble]

It should mean the transmitter and receiver side are galvanically isolated. Typically TX and RX both need supplies, often denoted Vcc1 and Vcc2.

that's what i thought.

basically, i would need a 12V-5V buck converter to power the isolator on the 12V side or a DC/DC isolator. the DC/DC isolator is probably a bad idea because the existing power method coming off the ESP32 USB could be overtaxed.

so, if i'm going that route, i might as well do both and power everything off the buck converter with a DC/DC isolator powering the boards.

 

[RNoble]

here's where i'm at.

 

[PaulStoffregen]

Have you considered what happens to the SI8660 if any of the blue output wires accidentally touch the +12V wires on those LEDs?

Is SI8660 meant to be a sacrificial part like the chatbot suggested for 74HCT245?

Here's page 26 from the SI8660 datasheet.

 

[Pio]

TEA 1 0505 won’t be able to power Teensy4.1 + ESP32. 200mA max output current is not enough.

 

[RNoble]

TEA 1 0505 won’t be able to power Teensy4.1 + ESP32. 200mA max output current is not enough.

thanks.
i didn't see that.

i can use the original power method for the boards until i find a better option.

Have you considered what happens to the SI8660 if any of the blue output wires accidentally touch the +12V wires on those LEDs?

Is SI8660 meant to be a sacrificial part like the chatbot suggested for 74HCT245?

Here's page 26 from the SI8660 datasheet.

yes.
i expect this to be sacrificial.
i believe i found the fault that torched the last two boards.