|
A few of you are probably familiar with my experiences with the DE version of the sBitx. There have been quite a number of different things happening which, by Occam’s razor, are in my opinion for the most part likely explained by unexpected spurious oscillations, some of which I have actually documented. My most recent trouble was solved by putting in a new Raspberry Pi 4. The previous one toggled the PTT line when pressed (pin 23 on the digital board connector) but there was no voltage on the TX line when PTT was pressed (pin 27). Putting in a new RPi 4 solved the problem. I now have 26 volts P-P on the 2N2219a driver collectors (Q15) and now I need to install new IRFZ24N finals (one had shorted).
This is the third Raspberry Pi that has failed during normal operation. The Pi was seated in the socket, the digital board as well, and so that error can be ruled out. Two others previously had also failed, and each of those had problems with the I2C lines (bit banged, I know) communicating with the Si5351a. I use one of these now as an OctoPrint server for 3D printing, so at least that one functions even though I can’t use it for the sBitx.?
At this point I have to assume that for whatever reason in my particular sBitx I run the risk in the future of damaging the TX line from a working Raspberry Pi, or the SDA/SCL lines to the Si5351a, or perhaps some other GPIO line on the Raspberry Pi. How would one protect the vulnerable lines from voltage transients? Back to back 1N4148 diodes? Bypass capacitors? Optical isolators? I am running on my 4th Raspberry Pi, the fifth if you include one with the onboard 3.3V regulator fried by an error on my part. Does anyone have any suggestions? I’m awaiting the delivery of a new sBitx v2 that will hopefully be less eventful, but I have put too much effort into this Developer’s Edition to leave it alone. As always, thanks everyone for your wonderful help.
Jack
N6LN
|
John,
I have been thinking along the same lines to protect the most expensive component in the sBITX.? Here is what I found in a search on the internet:
I like this one, as it has both current and voltage limiting in the suggestion:
This solution uses a 3.3 volt Zener for voltage protection and a 330 ohm resistor for current limiting.? The disadvantage is any delays the resistor could add to the turn on or off of the external circuit, like in TX/RX transitions.
Until we know the cause, it is hard to come up with the best design.? If we could be sure it is voltage, then a Zener to protect any voltages over 3.3 volts would be enough.? The problem is that it could be current, causing the failure.? The GPIO pins are rated for only 16ma.? If, in oscillation, the equivalent impedance of a capacitor is 50ohms, the current required would exceed the Raspberry Pi rating.? This might be a good simulation exercise to see the average current draw on the TX pin.? I do not know how this would happen on the bit-banged I2C lines.??
The other possibility would be induced voltage on the 3.3volt power rail picked up by the relatively long I2C run for the RTC.? Maybe that should be done with a shielded cable.
The above are guesses and should not be implemented without further research.? We have seen at least two owners of the sBITX DE that have experienced RPi failures.? Each of them has had more than one RPi fail.? Out of 150 shipped, that is still a tiny percentage.? We can track the v2 failures to see if it might be a layout issue rather than a fundamental design issue.? The two systems are different enough in design that the feedback possibility should change.
73
Evan
AC9TU
|
Thanks Evan. 3.3v zeners did occur to me, ?Incidentally I did have Zeners across the finals and still one of them shorted. It’s strange.?
Jack
N6LN
|
Evan, since I have two RPi4s with damaged SDA and SCL GPIO pins and another with a damaged GPIO4 (TX) pin I think I should put 3.3 volt Zeners on those lines, with series 330 ohm resistors. I assume the LC changes won’t be significant enough to interfere with the edge detection for the I2C bus. The best place would be at the exposed right angle bend in the 40-pin brass connector J2 mounted on the digital board. A cut there would allow for a series resistor insertion, and there is an adjacent ground plane on the digital board for the Zener diodes. Incidentally the WM8731 codec pins, 19 RLINEIN and 20 LLINEIN already have 3.9V Zeners! I wonder how delicate those inputs are.
Jack
N6LN
|
I don't know how my two RPi 4's were destroyed?but neither will even boot now.? ?I suspect RF on the power pins.? ? The power measured fine on a DC voltmeter.? ? I have not been able yet to recover from that given how scarce and overpriced the RPi's are.? ?And I'm not willing to power another RPi from my DE box given having watched it fry two!? ?So independently powered via the little usb connector on the Rpi is what I will be trying when I'm able again....? ? ? ?Such is life!
73 Gordon KX4Z
On Tue, Apr 18, 2023 at 2:32?AM John Terrell, N6LN < N6LN@...> wrote: Evan, since I have two RPi4s with damaged SDA and SCL GPIO pins and another with a damaged GPIO4 (TX) pin I think I should put 3.3 volt Zeners on those lines, with series 330 ohm resistors. I assume the LC changes won’t be significant enough to interfere with the edge detection for the I2C bus. The best place would be at the exposed right angle bend in the 40-pin brass connector J2 mounted on the digital board. A cut there would allow for a series resistor insertion, and there is an adjacent ground plane on the digital board for the Zener diodes. Incidentally the WM8731 codec pins, 19 RLINEIN and 20 LLINEIN already have 3.9V Zeners! I wonder how delicate those inputs are.
Jack
N6LN
|
The I2C lines on the raspberry pi can be blown in many ways: 1. The lm1117 of the si5351 shorts and puts 12v on the pins. 2.The lm1117 of the wm8731 shorts and puts 5v on the pins. 3. Si5351 or the wm8731 blow up and short the data/clock lines to ground. All these can only happen when the LM1117s blow out. It might be best to put a 12v zener on the powersupply line. - f
I don't know how my two RPi 4's were destroyed?but neither will even boot now.? ?I suspect RF on the power pins.? ? The power measured fine on a DC voltmeter.? ? I have not been able yet to recover from that given how scarce and overpriced the RPi's are.? ?And I'm not willing to power another RPi from my DE box given having watched it fry two!? ?So independently powered via the little usb connector on the Rpi is what I will be trying when I'm able again....? ? ? ?Such is life!
73 Gordon KX4Z
On Tue, Apr 18, 2023 at 2:32?AM John Terrell, N6LN < N6LN@...> wrote: Evan, since I have two RPi4s with damaged SDA and SCL GPIO pins and another with a damaged GPIO4 (TX) pin I think I should put 3.3 volt Zeners on those lines, with series 330 ohm resistors. I assume the LC changes won’t be significant enough to interfere with the edge detection for the I2C bus. The best place would be at the exposed right angle bend in the 40-pin brass connector J2 mounted on the digital board. A cut there would allow for a series resistor insertion, and there is an adjacent ground plane on the digital board for the Zener diodes. Incidentally the WM8731 codec pins, 19 RLINEIN and 20 LLINEIN already have 3.9V Zeners! I wonder how delicate those inputs are.
Jack
N6LN
|
A 12V Zener might be well advised on the power line, and I bought a bunch anyway for the gates of the finals, but since my supply puts out 13.7V I wonder whether I should also put two or three regular diodes in series with it. I have seen the 5351’s lm1117s fail twice, once simply shorting with smoke and simply dropping all power output to the 5351 and another time failing and sending 13.7v straight through to the 5351 which glowed orange in the chip’s center before melting two pads, the latter which was fixed by removing and replacing it and the crystal oscillator with a 5351 breakout board. It’s not as stable frequency wise but it’ll do for now. I now protect the new lm1117 by putting a small 5V LDO 78L05 between the 13.7V rail and the new lm1117 which in turn provides the breakout board with 3.3V. The 5V supply to the 1117 also has a 47 uF electrolytic on its input. So far it works.
On the other hand the most recent failure did not involve the bitbanged I2C lines. Instead it involved the TX line output (GPIO4) failing to send a TX signal on PTT. Since this does not directly involve an lm1127 I have to assume that it failed because of a power surge from elsewhere, either a transient from the main power supply (in which case a 12 or 14V Zener on the main supply would be helpful) or else a higher voltage coming from rf somewhere. Either way if it’s transient I doubt I’ll be able to measure it when it happens. If there’s no downside on putting 3.3V Zeners, with or without series resistors, on those three RPi lines, maybe it would help to do so. The right angle connector on the digital board has good exposure for adding small Zeners, and can easily be snapped open to insert a series 330 ohm resistor, and repaired if that interferes with the timing. Couldn’t hurt to try, could it?
Jack
N6LN
|
Could any of these failures be caused by low current supplied to the RPi by the radio? I have noticed that it I try to connect an NvME SSD drive to the USB port, it’s operation is erratic. I have no problems with the same drive connected to a RPi 400 powered by a 3.5A wall wart supply. I believe the sBitx only supplies 2A.
Also, the instructions in install.txt say to disable the low voltage warning, which does display if not disabled. Does disabling the warning also disable any low voltage protection?
Lastly, can I power the RPi by its USB C port while it is in the sBitx or do I need to do something to disable the power coming from the radio?
--
73,
Mark, N8ME
|
Mark, If you monitor the current draw of sbitx, it is well under 1A on receive. There is no way the rpi can draw more. There is some other reason, not low voltage. The low voltage warning happens anyway when you power it from the gpio pins instead of the usb port. - f
Could any of these failures be caused by low current supplied to the RPi by the radio? I have noticed that it I try to connect an NvME SSD drive to the USB port, it’s operation is erratic. I have no problems with the same drive connected to a RPi 400 powered by a 3.5A wall wart supply. I believe the sBitx only supplies 2A.
Also, the instructions in install.txt say to disable the low voltage warning, which does display if not disabled. Does disabling the warning also disable any low voltage protection?
Lastly, can I power the RPi by its USB C port while it is in the sBitx or do I need to do something to disable the power coming from the radio?
--
73,
Mark, N8ME
|
John,
Trying the 3.3 volt (maybe 3.6 volt?)? with or without the resistor could not hurt.? The diode would add capacitance that might make the I2C decode an issue, so I would do the Zener first, then add the resistor.? Adding the 7805 regulators ahead of the LM1117 is a good idea.
Using diodes to drop the input voltage on QRP transceivers works well.? The current required by ORO transceivers, like the Sbitx, may make it a problem.? 10 amps with a .6 volt drop require 6 watts to be dissipated.? A buck or buck/boost converter might be better if the noise is not an issue.
I am waiting to buy a new oscilloscope with I2C decoding.? This may be a good project to test the capability.
We should also look at the long 5 volts run from the 5-volt buck converter to the RPi board.? We may need RF bypassing on it, moving the converter, or replacing the trace with a shielded wire that can handle the current.? I like Gordon's idea of modifying how the RPi gets its power.? There are protections on the USB-C connector that are not on the GPIO connector.? The RPi schematic has a 5-volt Zener on the input from the USB-C connector.? That would be connected to the 5-volt line on the GPIO.? I would think it would keep the voltage at 5 volts regardless of the supply.? It would be interesting to do the forensics on one of the dead RPis to see if that diode is shorted or open.
I also do not like that the power warning had to be turned off when powering from the GPIO connector.? Low voltage may be the issue, with the pulses indicated by the audio when WiFi is turned on.? We are back to Gordon's suggestion to power through the USB-C connector with the voltage warnings on to see if there is an issue.
The above are questions, not suggestions.
73
Evan
AC9TU
|
I don't have the time//ability to? "? It would be interesting to do the forensics on one of the dead RPis to see if that diode is shorted or open."
If someone is really an expert on failed rPi's and someway for me to ship them a DEAD RPi4 --- I'll be happy to ship them one so they can enlighten us all how I managed to destroy these!!!
Thanks, Gordon Kx4Z
On Tue, Apr 18, 2023 at 9:38?AM Evan Hand < elhandjr@...> wrote: John,
Trying the 3.3 volt (maybe 3.6 volt?)? with or without the resistor could not hurt.? The diode would add capacitance that might make the I2C decode an issue, so I would do the Zener first, then add the resistor.? Adding the 7805 regulators ahead of the LM1117 is a good idea.
Using diodes to drop the input voltage on QRP transceivers works well.? The current required by ORO transceivers, like the Sbitx, may make it a problem.? 10 amps with a .6 volt drop require 6 watts to be dissipated.? A buck or buck/boost converter might be better if the noise is not an issue.
I am waiting to buy a new oscilloscope with I2C decoding.? This may be a good project to test the capability.
We should also look at the long 5 volts run from the 5-volt buck converter to the RPi board.? We may need RF bypassing on it, moving the converter, or replacing the trace with a shielded wire that can handle the current.? I like Gordon's idea of modifying how the RPi gets its power.? There are protections on the USB-C connector that are not on the GPIO connector.? The RPi schematic has a 5-volt Zener on the input from the USB-C connector.? That would be connected to the 5-volt line on the GPIO.? I would think it would keep the voltage at 5 volts regardless of the supply.? It would be interesting to do the forensics on one of the dead RPis to see if that diode is shorted or open.
I also do not like that the power warning had to be turned off when powering from the GPIO connector.? Low voltage may be the issue, with the pulses indicated by the audio when WiFi is turned on.? We are back to Gordon's suggestion to power through the USB-C connector with the voltage warnings on to see if there is an issue.
The above are questions, not suggestions.
73
Evan
AC9TU
|
Gordon,
I have the time and have done some forensics on other devices.? I am NOT a Raspberry Pi expert.? I DO have the SMD tools that I need.
With the above understanding, I would like to look into the cause of your RPi failure.? If there is someone who does have experience with RPi diagnostics, I would rather you send it to them.
Let's wait a few days to see if someone else approaches the challenge.? If not, then my QRZ email and address are good.
73
Evan
AC9TU
|
I found the Rpi output (and inputs) are fragile based on data sheets.
So one of the mods I made and wrote about maybe 5-6 months ago were to buffer all lines out
to the main board and and diodes and resistor to the inputs as well from the main board.
I've also done this with other Pi applications that do not use a direct plug in hat.
to date I've not fired a pi (or Arduino).
--
Allison
------------------
Please use the forum, offline and private will go to bit bucket.
|
Allison I found that post.? Do you have any drawings you can share? 73 Lou KI5FTY
On Tue, Apr 18, 2023 at 11:53?AM ajparent1/kb1gmx < kb1gmx@...> wrote: I found the Rpi output (and inputs) are fragile based on data sheets.
So one of the mods I made and wrote about maybe 5-6 months ago were to buffer all lines out
to the main board and and diodes and resistor to the inputs as well from the main board.
I've also done this with other Pi applications that do not use a direct plug in hat.
to date I've not fired a pi (or Arduino).
--
Allison
------------------
Please use the forum, offline and private will go to bit bucket.
|
Alvin, that is fine, I’m just would be really interested to know what I blew apart. That might help avoid a repeat!
Gordon
toggle quoted message
Show quoted text
On Apr 18, 2023, at 10:34, Evan Hand <elhandjr@...> wrote:
?Gordon,
I have the time and have done some forensics on other devices.? I am NOT a Raspberry Pi expert.? I DO have the SMD tools that I need.
With the above understanding, I would like to look into the cause of your RPi failure.? If there is someone who does have experience with RPi diagnostics, I would rather you send it to them.
Let's wait a few days to see if someone else approaches the challenge.? If not, then my QRZ email and address are good.
73
Evan
AC9TU
|
Lou, I remembered that post from Allison but I can’t find it. What was the number of that post?
Jack?
N6LN
|
|
|
While I wait for my sbitx v2, I have noticed there are no 'backward facing' clamp diodes on the fixed regulators that prevent the regulated output voltage (VOUT) exceeding the input voltage (VIN) on power down conditions. I also think according to the data sheet LM338 a filter capacitor is required from the 'ADJ' pin to ground. The datasheets for both the LM338 and the AMS1117 recommend the? clamp diode and I think the AMS1117 is particularly sensitive in this respect. I remember many years ago my homebrew Z80 computer regularly blew the graphics eeprom until I did this.
|
Paul The v2 does not use LM338, it uses a switching regulator. We also use a 7805 step down that feeds the 1117s. There are 3.3v zeners on all external facing gpio pins
- f
While I wait for my sbitx v2, I have noticed there are no 'backward facing' clamp diodes on the fixed regulators that prevent the regulated output voltage (VOUT) exceeding the input voltage (VIN) on power down conditions. I also think according to the data sheet LM338 a filter capacitor is required from the 'ADJ' pin to ground. The datasheets for both the LM338 and the AMS1117 recommend the? clamp diode and I think the AMS1117 is particularly sensitive in this respect. I remember many years ago my homebrew Z80 computer regularly blew the graphics eeprom until I did this.
|
If sBitx V1 schematic on github is correct,
Main board's C66 (0.1uF) is being driven directly by the pi.? Cap load without an added current limiting series resistor.
Surprised that so many TX controlled circuits on the sBitx main board are being driven directly by the pi, including setting bias of an amp.
I don't see the downside of having the TX node on the main board at 12V (with the appropriate required changes to adjust for the higher voltage).
There's already an inverting buffer on board. Use the 2 unused channels on the ULN2003A to take the 3.3V signal from the pi up to 12V with 500mA of drive.
Still wouldn't drive a cap directly.
Just my observation,
Gary
|
John Terrell, N6LN
Mark Erbaugh