KiwiSDR not powering on after power off

Comments

• smg

  November 12 edited November 12

  What are you using for your external power supply? How much current does it deliver? Have you tried another supply? The KiwiSDR can draw up to 1.5A when starting and you should account for a headroom of up to 2A

  http://kiwisdr.com/info/#id-power
• jks

  November 12 edited November 12

  When the single blue LED flashes quickly at power on it's not over current protection. It's inrush low voltage or slow waveform rise time protection.

  Try the power cable from one of your working Kiwi and it will probably work. Figure out what's different with your other cable causing the voltage sag (e.g. wire gauge too small, cable too long).
• ChrisSmolinski

  November 13

  I took the power supply from this KiwiSDR (which worked fine before all this started) and connected it to known good KiwiSDR, and it comes up fine.

  I took the power supply from that working KiwiSDR, and used it with this one, the BBG booted, but when the KiwiSDR software came up, the webpage reported "No Kiwi hardware detected".

  I got yet another power supply, one of the white Apple chargers that works great with KiwiSDRs, and it does the same thing - "No Kiwi hardware detected"

  5V IN: 4.86

  5VE: 4.78

  5V: 4.70

  3.3V: 3.40

  1V: 1.00

  1.8V: 1.80

  3.3A: 3.29

  3.3G: 3.30

  3.3GPS: 3.29

  I got another USB charger with a similar power capacity, and the voltages are pretty much identical.

  I got a 5V 4A power supply I've previously used with a KiwiSDR as well as a Raspberry Pi, and I get the short blue LED flash.

  Something seems to be wrong with this KiwiSDR board? It seems like something is loading down 5V more than is typical.
• jks

  November 13

  All the voltages are normal except for 5VE, coming from your supply, and 5V, which comes back from the Beagle after going through a simple FET switch from 5VE. 4.78/4.70 is just too low.

  Since your power supplies seem to have enough current capability it's got to be a cable with too much I^2R drop. Too much resistance and/or small wire gauge causing voltage at 5VE/5V to be too low. What are these voltages on a Kiwi that works using the same supply/cable?
• ChrisSmolinski

  November 13

  I took the power supply/cable from the bad (serial 3615) Kiwi and using it with the good (serial 4759) Kiwi:

  VIn:4.88

  5VE: 4.82

  5V: 4.76 (all other voltages fine/not relevant so for brevity I will skip them)

  The Kiwi still works fine.

  
  

  Extra data:

  I took the good KiwiSDR (serial 4759). With it's power supply and integrated cable I get:

  5VIn: 5.00

  5VE: 4.93

  5V: 4.88. (all other voltages fine/not relevant so for brevity I will skip them)

  That *same power supply/cable* on the bad KiwiSDR (serial 3615) results in a "No Kiwi hardware detected." message when trying to access it (of course the Admin page comes up fine). Here are the 5 volt voltages:

  5Vin: 5.00

  5VE: 4.94

  5V: 4.86. All 3.3, and the 1.8 and 1.0 volt supplies are as you'd expect.
• ChrisSmolinski

  November 16

  Any more suggestions of things to check?
• studentkra

  November 16

  So... It looks like FPGA chip dammage. Can you read logs. Find something "panic FPGA".
• ChrisSmolinski

  November 16

  Rummaging around my junk boxes, I found yet another 5V power supply, rated 2.4A with a very beefy looking cable.

  5V IN is 4.999 V

  5VE 4.920 V

  5V: 4.840 V

  Using it, I once again got the No Kiwi hardware error message, the log file does indeed have the FPGA panic message:

  Sun Nov 16 18:48:04 00:00:31.222 ....   L *** FPGA panic: code=1 FPGA_INIT never went LOW

  Sun Nov 16 18:48:04 00:00:31.224 ....   L *** NO KIWI HARDWARE DETECTED

  
  

  I then decided to do a Kiwi Server Restart from the admin panel, and this time, no FPGA panic:

  Sun Nov 16 18:50:05 00:00:01.412 ....   L FPGA firmware: /usr/local/bin/KiwiSDR.rx4.wf4.bit 30141

  Sun Nov 16 18:50:06 00:00:02.177 ....    e_cpu firmware: /usr/local/bin/kiwid.aout 19082

  
  

  And if I go to a receiver tab in the browser, it tries to work. It comes up, there's a waterfall, but it's updating very slowly, with just a distorted buzzing sound for audio, and the waterfall doesn't "look right". I've attached a screenshot (I hope)

  

  
  
• studentkra

  November 16

  Try cleaning the contacts between the two boards. Use alcohol or pure gasoline. Simply add the cleaning liquid to the contact area and connect and disconnect the boards several times. Then let the cleaning liquid dry. If this doesn't help, try reflashing the BeagleBone board. The FPGA firmware file may be corrupted. If this doesn't help, I suspect there's a problem with the FPGA. Either the chip is faulty or the BGA solder balls are broken.
• ChrisSmolinski

  November 18

  It's surprising the FPGA could fail after something as basic as cycling power.
• studentkra

  November 18

  I don't know exactly how the FPGA works. It's really very complicated for me. But perhaps the FPGA continued to function properly with the loaded firmware, even though it was already "slightly broken." And when the power was turned back on, it couldn't initialize. I don't know...
• studentkra

  November 18 edited November 18

  I don't know what John will say about this, but I know of several cases, where simply heating the BGA balls, until they melt and wiggling the FPGA with a good soldering flux, brought KiwiSDR and its clones back to life. It's easier than reballing. Reballing is more reliable, though much more difficult.
• jks

  November 18 edited November 18

  Something here doesn't make sense. I looked at the kiwisdr.com logs for any clues. What I found was strange. The periodic public registration status query was working fine until 11 Nov. Afterwards here is the record of that plus the info sent for the my.kiwisdr.com feature beginning 16 Nov including an indication which is sent that only works if the FPGA is responding:

  16/Nov/2025:17:19:56 fpga ok

  16/Nov/2025:17:35:01 fpga ok

  16/Nov/2025:18:04:05 fpga ok

  16/Nov/2025:18:14:00 fpga BAD

  16/Nov/2025:18:19:48 fpga ok

  16/Nov/2025:18:48:10 fpga BAD

  16/Nov/2025:18:50:14 fpga ok

  16/Nov/2025:19:03:33 fpga ok

  Now these are all for serial number 3615, the bad board you mention. Since the serial is in an EEPROM on the board this can't (shouldn't) be duplicated across multiple boards. Obviously this doesn't cover the time you worked on things without an Internet connection.

  So I can't explain why the FPGA would alternate between good and bad in a relatively short amount of time, unless this is power supply substitution. But you said you got the "FPGA not responding" message in all cases. "fpga ok" here is a limited indication that the FPGA programmed correctly and the SPI interface between Beagle and FPGA is working (code for the embedded e_cpu was downloaded correctly and data with it exchanged correctly). But it doesn't necessarily mean anything beyond that was okay.

  I guess I'd like to connect as admin and try rebooting/restarting a few times and look at log messages. Maybe something will turn up. Please email details to support@kiwisdr.com
• nitroengine

  November 18

  I assume you've already done this, but I'll post it just in case. Double check that you don't have the GPS antenna and receive antennas swapped. A guy in town here who also has a Kiwi accidentally swapped those antennas and his Kiwi was acting in a similar way. With his antenna transformer (which almost appears as a short at DC) connected to the GPS antenna input, it was essentially shorting out the 3.3 volt DC bias on that jack. The Kiwi would power up, he could get into the admin page, but the waterfall was slow, had a bunch of spurs with no actual signals, and it would randomly freeze up.

  After some remote troubleshooting (I wasn't there), we realized the antennas were swapped. After correcting that, it came back online like normal.

  If they are not swapped, try powering it up without the GPS antenna attached, in the event there is a fault with the GPS antenna.

  
• jks

  November 18 edited November 19

  Could you please check the P9 header, pins 3 & 4 again? (marked 3.3 on PCB)

  I had forgotten that the EEPROM has its own I2C interface separate from the SPI used by the FPGA. So the EEPROM should really be working and delivering the serial number.

  If you use the admin console and do a cdp to get to the build directory, then a m eeprom_dump Debian says "Connection timed out" (I2C device address 0x54). The Beagle EEPROM is on the same I2C bus (address 0x50) and works if you say m eeprom_dump_all (second I2C device shown). So we know the Beagle I2C bus/driver is okay.

  So either the 3.3V is not present on the EEPROM (see Kiwi-1 schematic) or the EEPROM is broken. All the other regulators on the Kiwi board (3.3A, 3.3G, 1.8V, 1V) are fed from the raw 5V from the Beagle via P9 pins 7 & 8. The 3.3V from the Beagle only goes to the EEPROM and FPGA. So that's why I'm suspicious of it.
• G8JNJ

  November 19

  I'd still suspect the power supply & cable.

  I found that my KiWi's were more reliable if fed with 5.4V at the barrel connector when under load.

  A slight voltage drop across the KiWi board results in 5V at the Beagle header.

  The voltage drop on some boards is greater than others, and this additional margin helps with the more reluctant ones.

  Regards,

  Martin
• ChrisSmolinski

  November 19

  I checked the P9 header, pins 3 and 4 both have 3.4 V.

  I also checked voltages on the EEPROM, U301:

  3.40 volts on Vcc pin 8, and the A2 and WP pins. Closer to 3.35 volts on the SDA and SCL pins.
• jks

  November 19

  Did you ever try swapping the Kiwi board onto another previously working Beagle? It bothers me the 3.3V from the Beagle is at 3.4V. I've never seen one run high like that. There is a separate regulator on the Beagle used for that 3.3V which goes to the Beagle cape header, eMMC, sd card slot, serial level converter, JTAG and Ethernet. So it must be relatively okay.
• ChrisSmolinski

  November 20

  As it turns out, in my testing I did the opposite in fact, I put another (working) Kiwi board on this BBG, and it worked fine.
• jks

  November 20

  Okay, that tells us something. But trying the trouble board on another BBG would be a useful verification.
• studentkra

  November 20 edited November 20

  Now i am complitly confused! Which board is not working? Kiwi or beagle?
• jks

  November 20

  Kiwi it seems given the last swap Chris mentioned. The BBG used in the original stack with the FPGA problems runs another Kiwi board fine. I'd like to know if the FPGA problem continues when the Kiwi board is moved to the BBG of a known working Kiwi (it almost certainly will, but it's worth a check).
• studentkra

  November 20 edited November 20

  So.. It something incompability specific kiwi and beagle boards? Its strange.
• ChrisSmolinski

  November 21

  I moved the bad 3615 KiwiSDR board to a known good BBG, and it's the same series of FPGA messages in the log at startup.