That ferrite bead is normally for EMI rather than "noise" as these components are about 120 ohms at 100Mhz but that is not your problem. I am not confident though that your pin 18 is hooked up as the artwork shows it is not connected and you didn't say you connected a wire to it either. If you did an ohms check then you might see a misleading internal chip connection but the time spent doing that could have been spent just fixing it. (when we say "do this" and you just do it then you might get results)
BTW, I don't think you would have any problem with the 10k+100k "divider" although I like to think of the 100k as more of a pulldown which it is. So the important one is the 10k and even at 12V it still only "leaks" 12-3.6/10k <= 0.84ma which is way less than many of the 5V current limit resistor values I've seen used. However even with the 100k pulldown you could still use a 100k series resistor instead of the 10k as that would still work quite reliably at 5V or 50V.
I took a magnifying glass and inspected it under bright lighting at several angles and it is for sure connected unlike the Diptrace file showed. I had edited that file several times after I ordered the boards because I found other issues that are already fixed. I may have changed the clearance of the ground plane on the board which caused this issue.
On bench testing, the entire circuit works without glitching. I even pumped up the RPM to 15,000 (using another Prop and a transistor to supply the 5V output) and it never lost USB connection with the same laptop as used in the vehicle. This leads me to the main power filtering and other decoupling caps next to each IC. I feel like I am chasing my own tail
I took a magnifying glass and inspected it under bright lighting at several angles and it is for sure connected unlike the Diptrace file showed. I had edited that file several times after I ordered the boards because I found other issues that are already fixed. I may have changed the clearance of the ground plane on the board which caused this issue.
On bench testing, the entire circuit works without glitching. I even pumped up the RPM to 15,000 (using another Prop and a transistor to supply the 5V output) and it never lost USB connection with the same laptop as used in the vehicle. This leads me to the main power filtering and other decoupling caps next to each IC. I feel like I am chasing my own tail
Can you post the proper DIptrace files otherwise I am chasing my tail too. Do you have a scope that you can connect and check for normal noise levels at least?
(added an edit to the last post)
Here is the original file before I changed the ground plane clearance.
I do have a hand held O-Scope that I have used one a very few occasions since most of the settings on there are confusing :P I also can't afford a larger better one, so I am stuck between a rock and a hard place.
Is that the exact same one as the board you have? If though the noise is coming in through the power supply and quite likely is then the common-mode choke I mentioned may help but I find full isolation easier. An isolated DC-DC converter solves all your noise and ground problems easily and although it is still possible to make your current design work reliably you need to be able to see what's going on which requires a decent scope and also knowing how to use this properly, such as connecting the ground probe close to the point you are measuring and making sure the scope ground is not interfering with the test ground. Most DSOs have a glitch mode that shows you things that are happening that you mightn't be able to see normally or not at all on an old analog scope.
Yes, that is identical to the board I have now except for R24 and R22 are now connected to the ground plane as they should be, and R43 and R44 are connected to the input voltage (12 to 15V).
The "except" bit is always a worry but I will take your word for it as I am very cautious even with my own design as one minor discrepancy might not be the only one, unintentional or not.
I'll get back to you shortly after looking at the pcb, thanks.
You have been doing quite well with laying out your pcb but here are some little tips to help you improve your next design. I've attached a little screen capture and have labeled certain points of attention.
A. You want a ground or guard ring FROM the CPU to isolate the crystal from I/O lines. However the ground almost connects but the I/O lines which have room to move out of the way are instead cutting close to the most sensitive pin in the whole circuit the XI crystal input. The library pattern could be updated too to use a round pad as these are not polarity sensitive parts. Sharp corners are a bit like "antennas".
B. The decoupling capacitor here is nice and physically close but still a long way away electrically. The ground of this cap is not intimately connected to the relevant ground on the CPU. If you can't route it on a double-sided then see the next item V.
V. Vias vias vias! Okay, normally I avoid using a via when I can do a clean route but in the case of double-sided and grounds (even power) I would place them strategically to lower the impedance of a fragmented ground place. You have one in the middle of the CPU but please don't stop there for starters. Just think, though not technically correct that if you were an electron what hoops and mazes would you have to go through to complete not just a circuit but your circuit? I think you would be surprised at how easy we can make it for these "electrons" with a little push shove here, a little reroute there, trim an edge there, and drop vias down as if we are giving them a shortcut.
Ok, I will go through and make those changes and repost the updated PCB file tomorrow. Do you think changing these would fix the USB disconnect issue or should I also change my PCB to match your schematic you posted before with the ferrite bead?
I have also been looking for an isolated DC to DC converter and have noticed these things are quite large. Since this is being used in an automotive application, the voltage range will have to be around 9V min to whatever the highest voltage I can afford I suppose. On Digikey, there are several 9V to 36V input ranges, but the size is 32mm which is slightly over 1.25 inches. That is quite large and would require a BUNCH of work moving everything on the PCB to accommodate the new larger 5V power supply.
I can't decide on which direction to go... going for the DC to DC converter or using the common mode choke. The common mode choke seems to be a much smaller component and is easily available at a much lower price than the DC to DC converter, but, the DC to DC converter seems to be the best way to go due to the isolation, but is extremely expensive
Here is a little 4-pin SIP isolated DC/DC converter that supplies 200ma at 5V. The 5V is not regulated but if you do need some regulated 5V I would just add an LDO for this as the outputs of these unregulated converters tends to be at least 5.3V. Are they expensive? Hell no! unless you think $3.49 is too much.
BTW, don't worry too much about doing a new PCB, just try the one you've got with the isolated supply for the logic. The ferrite bead is not a necessity either, just good practice.
If you use a DC to DC convertor to isolate the power, you will then have to isolate all the inputs too, as the grounds will no longer be common. Your voltage divider won't work without a common ground.
It may be that is what's required, but something you can test first is to power your circuit from a clean source, say a battery. Then use an opto-isolator for your RPM (and any other inputs) to see if the USB resets. If it doesn't reset, you probably need to isolate everything. If is still resets, isolated power isn't the (only) problem.
If you use a DC to DC convertor to isolate the power, you will then have to isolate all the inputs too, as the grounds will no longer be common. Your voltage divider won't work without a common ground.
It may be that is what's required, but something you can test first is to power your circuit from a clean source, say a battery. Then use an opto-isolator for your RPM (and any other inputs) to see if the USB resets. If it doesn't reset, you probably need to isolate everything. If is still resets, isolated power isn't the (only) problem.
There isn't actually a need to isolate the whole circuit as I'm sure that load and signal grounds would not rely on the supply anyway. The main thing is to eliminate that the supply is a cause of this problem and the converter is the easiest way to do it for this application. The other thing is I'm trying to keep it real simple and a single solution or problem to work on at a time as it seems that with every little issue that is brought up that the OP doesn't know which one to focus on and I'm just trying to help.
I am at work right now, but I just pulled up the data sheet on the DC to DC converter you listed and it does not have a max voltage input listed. It just shows 12V input. Just about every thread I have made about connecting my circuits to a vehicle, the topic comes up about spike voltages that can damage components. If 12V is all the converter can handle, I would figure that converter would not be a good pick. Have I been misunderstanding the other posts?
I just don’t get it. I have analyzed the whole circuit board and I have found lots of errors.
Take a look at Drawing.dch page ‘POWER’.
If the voltage regulators are of type 1117 then something is not correct.
And if you are using type 78xx then you must twist 2 and 3 to get it right.
Drive of the N-channel mosfet’s. Why 2 resistors in series?
Jumpers 3.3V and 5V attached to U2 can make a mess if activated at the same time.
For the moment I am making a drawing from scratch using the pcb layout.
Will publish it as soon as I’m finish with it.
Would be a great help for us all if you tell us all the missing part names.
It's how you read datasheets, yes it is designed to be run from 12V to give you 5V but if you put 15V into it it will probably have 6V out. The absolute maximum ratings may be much higher than this too but you need to understand the component, the circuit, and the application.
So anyway, what's the solution? Well in fact I would just use a 12 to 5V switcher that then fed a 5 to 5V isolated converter and then there isn't an issue with voltages etc. So it's a bit like what you have now, you can insert a 5 to 5V isolated converter in there from your 5V power to 5V on the board. Use the same type but 5:5, (ROE0505S). I've used these same technique in a recent project to switch from 7-35V down to 5V then across 5:5 isolated converters.
Ok, that makes a bunch more sense! Now, as far as having a separate ground plane, would that mean I would have to join the grounds back together somehow? Maybe a small trace between them?
You solved your ground problem by going to the trouble of isolating your supply so signals are happy with their signal ground and the loads are happy with their ground, and none of it can go via the supply since it's isolated.
So you want: SIGNAL GROUNDS---->[THE BIG GROUND]<---- LOAD GROUNDS
but not: LOAD GROUNDS---->SIGNAL GROUNDS---->{THE BIG GROUND] bad bad bad
Oh, one more thing is that if you need 5V to be 5V for analog etc then you should add an LDO to the output of the DC/DC as the output could be as high as 5.8V unloaded.
So if I isolate the main ground which I will call chassis ground, then try to detect the RPM signal on the isolated ground, would it work even if there is no common ground?
The ADC's would also need a common ground to the chassis if the power provided to the ADC input pin is from the vehicle's ECU. The throttle output and RPM is 0 to 5V from the ECU. If the ECU and my circuit do not have a common ground, would that not cause a "complete" circuit and that would be like testing a battery with only 1 lead of a meter?
@feng : The regulators I originally designed for are not going to be used anymore, but they did work properly, just got warm to the touch and I don't like unwanted heat. I would have to search for which ones I picked if you need to know which ones they are. Also... I did not see the 2 resistors in series until you pointed them out. I must have placed the Prop pin connection on the wrong side of the first resistor I have not tested that part of the circuit yet either.
I am at work and can't view the file until I get home. The R6J6 power jumper is designed to be a voltage selector for the output to whatever is plugged into it. This will probably be changed to a SPDT SMD switch once I get all these other bugs worked out.
The solenoid driver has a ground that is connected with a 150 Ohm resistor as the datasheet of the IC calls for if I want to use short circuit protection.
Once I get home, I will look at the sheet 6 and see which MOSFETs you are talking about. The larger mosfets work and handle the current I need plus some, so I don't want to mess with that circuit if I don't have to.
@Peter Jakacki : Did you see my previous question about isolating the ground and trying to read RPM, Throttle, and other ADC input from the vehicle ECU which would be a separate ground?
Can anyone help me on the separation of the common grounds? It seems like a good idea, but I want to know what I am working with before ordering parts.
I can't see the board and the way it has been wired up which would give me a better idea about what to tell you. However the power supply ground has been isolated then you only have the load ground and signal ground and basically there is no real problem wiring it all back to one point, like at the power supply or battery or maybe on the board perhaps but I doing a bit of guessing on your board here you must realize. Just hook it up so that it works for now then see if it plays up.
I previously asked whether you had a scope but I don't recall your reply on that. Designing equipment and not having basic tools to check what is happening is hard enough for the builder but you can probably only expect guidelines from anybody then in those circumstances. Can you imagine if I kept asking the forum for help on my RS485 glitch problem but I couldn't capture and show what was going on? It would be nigh well impossible and utterly frustrating for someone else to fix it remotely with guesses.
I am confused on how to isolate the power and ground, then bring the ground back together. I thought it needed to be isolated? The terms Load ground and Signal ground make some sense since I believe I have this already on my current board. The Load ground (Black wire run into the header connected to the vehicle's ground or chassis) is connected to the high voltage controlling components (MOSFETS), then, with a small trace and the "signal" ground plane is connected to the "load ground". Is this what you are referring to that I should do with the DC to DC converter ground?
EDIT >>>> I do have a hand held O-scope but I don't think it is good enough to detect whatever issue I am having.
Comments
BTW, I don't think you would have any problem with the 10k+100k "divider" although I like to think of the 100k as more of a pulldown which it is. So the important one is the 10k and even at 12V it still only "leaks" 12-3.6/10k <= 0.84ma which is way less than many of the 5V current limit resistor values I've seen used. However even with the 100k pulldown you could still use a 100k series resistor instead of the 10k as that would still work quite reliably at 5V or 50V.
On bench testing, the entire circuit works without glitching. I even pumped up the RPM to 15,000 (using another Prop and a transistor to supply the 5V output) and it never lost USB connection with the same laptop as used in the vehicle. This leads me to the main power filtering and other decoupling caps next to each IC. I feel like I am chasing my own tail
https://goo.gl/photos/fimob3hgtcGcyD3U7
Can you post the proper DIptrace files otherwise I am chasing my tail too. Do you have a scope that you can connect and check for normal noise levels at least?
(added an edit to the last post)
I do have a hand held O-Scope that I have used one a very few occasions since most of the settings on there are confusing :P I also can't afford a larger better one, so I am stuck between a rock and a hard place.
https://www.digikey.com/product-search/en/power-supplies-board-mount/dc-dc-converters/4325599?k=DC+converter&k=&pkeyword=DC+converter&pv1525=62&pv1526=24&FV=fff40042,fff800df,15c0003&mnonly=0&newproducts=0&ColumnSort=0&page=1&stock=1&quantity=0&ptm=0&fid=0&pageSize=25
Holy Smile!
I'll get back to you shortly after looking at the pcb, thanks.
A. You want a ground or guard ring FROM the CPU to isolate the crystal from I/O lines. However the ground almost connects but the I/O lines which have room to move out of the way are instead cutting close to the most sensitive pin in the whole circuit the XI crystal input. The library pattern could be updated too to use a round pad as these are not polarity sensitive parts. Sharp corners are a bit like "antennas".
B. The decoupling capacitor here is nice and physically close but still a long way away electrically. The ground of this cap is not intimately connected to the relevant ground on the CPU. If you can't route it on a double-sided then see the next item V.
V. Vias vias vias! Okay, normally I avoid using a via when I can do a clean route but in the case of double-sided and grounds (even power) I would place them strategically to lower the impedance of a fragmented ground place. You have one in the middle of the CPU but please don't stop there for starters. Just think, though not technically correct that if you were an electron what hoops and mazes would you have to go through to complete not just a circuit but your circuit? I think you would be surprised at how easy we can make it for these "electrons" with a little push shove here, a little reroute there, trim an edge there, and drop vias down as if we are giving them a shortcut.
I have also been looking for an isolated DC to DC converter and have noticed these things are quite large. Since this is being used in an automotive application, the voltage range will have to be around 9V min to whatever the highest voltage I can afford I suppose. On Digikey, there are several 9V to 36V input ranges, but the size is 32mm which is slightly over 1.25 inches. That is quite large and would require a BUNCH of work moving everything on the PCB to accommodate the new larger 5V power supply.
I can't decide on which direction to go... going for the DC to DC converter or using the common mode choke. The common mode choke seems to be a much smaller component and is easily available at a much lower price than the DC to DC converter, but, the DC to DC converter seems to be the best way to go due to the isolation, but is extremely expensive
BTW, don't worry too much about doing a new PCB, just try the one you've got with the isolated supply for the logic. The ferrite bead is not a necessity either, just good practice.
It may be that is what's required, but something you can test first is to power your circuit from a clean source, say a battery. Then use an opto-isolator for your RPM (and any other inputs) to see if the USB resets. If it doesn't reset, you probably need to isolate everything. If is still resets, isolated power isn't the (only) problem.
There isn't actually a need to isolate the whole circuit as I'm sure that load and signal grounds would not rely on the supply anyway. The main thing is to eliminate that the supply is a cause of this problem and the converter is the easiest way to do it for this application. The other thing is I'm trying to keep it real simple and a single solution or problem to work on at a time as it seems that with every little issue that is brought up that the OP doesn't know which one to focus on and I'm just trying to help.
This one better possibly? http://www.mouser.com/ProductDetail/Cincon/EC3AW01/?qs=sGAEpiMZZMvGsmoEFRKS8IScV8SUvsqq4VhnrLAIRUQ=
Take a look at Drawing.dch page ‘POWER’.
If the voltage regulators are of type 1117 then something is not correct.
And if you are using type 78xx then you must twist 2 and 3 to get it right.
Drive of the N-channel mosfet’s. Why 2 resistors in series?
Jumpers 3.3V and 5V attached to U2 can make a mess if activated at the same time.
For the moment I am making a drawing from scratch using the pcb layout.
Will publish it as soon as I’m finish with it.
Would be a great help for us all if you tell us all the missing part names.
So anyway, what's the solution? Well in fact I would just use a 12 to 5V switcher that then fed a 5 to 5V isolated converter and then there isn't an issue with voltages etc. So it's a bit like what you have now, you can insert a 5 to 5V isolated converter in there from your 5V power to 5V on the board. Use the same type but 5:5, (ROE0505S). I've used these same technique in a recent project to switch from 7-35V down to 5V then across 5:5 isolated converters.
So you want: SIGNAL GROUNDS---->[THE BIG GROUND]<---- LOAD GROUNDS
but not: LOAD GROUNDS---->SIGNAL GROUNDS---->{THE BIG GROUND] bad bad bad
The ADC's would also need a common ground to the chassis if the power provided to the ADC input pin is from the vehicle's ECU. The throttle output and RPM is 0 to 5V from the ECU. If the ECU and my circuit do not have a common ground, would that not cause a "complete" circuit and that would be like testing a battery with only 1 lead of a meter?
POWER. Have reconnected the 5V regulator.
Sheet6. Shouldn’t the 1K and 100K resistors shift place? Is 3V enough to drive the MOSFET?
NETWORK CONNECTOR. Bridging J1 and J2 at the same time will make sparks.
SOLENOID DRIVER. Something is missing here. What about a ground connection?
The solenoid driver has a ground that is connected with a 150 Ohm resistor as the datasheet of the IC calls for if I want to use short circuit protection.
Once I get home, I will look at the sheet 6 and see which MOSFETs you are talking about. The larger mosfets work and handle the current I need plus some, so I don't want to mess with that circuit if I don't have to.
I previously asked whether you had a scope but I don't recall your reply on that. Designing equipment and not having basic tools to check what is happening is hard enough for the builder but you can probably only expect guidelines from anybody then in those circumstances. Can you imagine if I kept asking the forum for help on my RS485 glitch problem but I couldn't capture and show what was going on? It would be nigh well impossible and utterly frustrating for someone else to fix it remotely with guesses.
EDIT >>>> I do have a hand held O-scope but I don't think it is good enough to detect whatever issue I am having.