Strange effect: Big step of RX noise floor when moving through band

[s.hi...@gmx.net]

Hi OM's

I am Sven (DH1SV), and managed to build up my Radioberry V2. The project documentation is great, and after setup it worked at first try.

@pa3gsb : This project is really great job!

At the moment I have a strange behavior. Without connected antenna I have typically normal noise floor without signal (expected).

After some time of "scrolling" through the band on Thetis the noise floor steps up (see pictures)...

I double checked this in pihpsdr, and can see the same noise floor, so Thetis seems not to be the root cause.

For recovery a restart of the raspberry is needed (just stop service in console and restart ($ sudo systemctl stop radioberry / $ sudo radioberry) is not enough to solve it...

At the moment I am clueless how to analyze this issue... does anybody have an idea, or maybe same issue?

My setup: Radioberry V2 beta4 (C256 shortcut fixed), CL016 FPGA, RPI4, "current" development release of Radioberry installed. Heatsinks and fan attached.

Are there other HW issues on beta4, which should be fixed?

Reply would be great!

Thanks in advance.

Sven - DH1SV

73'

[pa3gsb]

Hi Sven,

Nice you did make your radioberry going.

I do not have seen this behaviour.

If the noise level is high.... and you start pihpsdr it is also having a high noise level.

The problem starts in Thetis... if you are tuning in pihpsdr .... are you than also to reproduce the problem?

About the hardware; you als soldered the pad to ground for AD9866 and FPGA.?  

If pihpsdr is not setting the noise floor high... we can log the SPI commands when you are running Thetis and see if there some strange command...

Please let us know.

73 Johan

Op vrijdag 30 april 2021 om 14:01:55 UTC+2 schreef s.hi...@gmx.net:

[s.hi...@gmx.net]

Hi Johan,

thanks for reply. You are right, to exclude Thetis as root cause is overhasty.

Regarding your question:

- I checked tuning in (local) pihpsdr. There it was not possible to get into the noise problem, so this points more to Thetis.

- The thermal pads on FPGA and AD9866 are soldered 

I tried multiple scenarios now:

- Setup second PC with clean install of Thetis (same current version). => Issue also reproducible on second PC, so no config topic caused by special settings.

- Tried PowerSDR (same "roots" as Thetis) => Not reproducible in PowerSDR.

=> So issue seams Thetis related

Other new interesting observation. Until now the only recovery in this "error" case was reboot of Raspberry. Now I found a second recovery path to get back to normal noise level.

This works sometimes (50%):

Precondition:

- System in "error" state (high noise floor)  
- Thetis connected

Action:  

- Start pihpsdr (in parallel to Thetis) 

- Connect pihpsdr

Result:

- pihpsdr starts up, but does not work because Thetis is still connected.

- Sometimes (50%) the noise floor recovers to normal operation.

Note: If pihpsdr is started in "error" case without connected Thetis there is no recovery, so init from pihpsdr by itself is not enough..

=> But in general: There is a possible SW command/sequence available to get out of that noise state, and there is no stuck.

Note: I put error in quotation marks, because it is an issue at the moment, and not clear if really an error.

Ist it possible to trace the SPI by debug options, or tool on raspberry?

I would also have a saleae logic analyzer, but sometimes it is difficult to trace SPI because of speed.

Thank you

Sven (DH1SV)

73'

[pa3gsb]

Sven,

SPI commands send for controlling the radio can be easily switched on:

https://github.com/pa3gsb/Radioberry-2.x/blob/master/SBC/rpi-4/device_driver/firmware/radioberry.c

uncomment the following line:

//fprintf(stderr, "RB-Command = %02X Command = %02X command_data = %08X\n", rb_info.rb_command, command, command_data);

and excute a build by using the make command.

sudo ./radioberry and watch the commands....

(sudo systemctl stop radioberry must be performed first)

Hope you are able to find the find problem.

Have fun 

73 Johan

PA3GSB

Op dinsdag 4 mei 2021 om 10:16:59 UTC+2 schreef s.hi...@gmx.net:

[s.hi...@gmx.net]

Hi Johan,

thanks a lot. Now it becomes interesting.

I was able to trace the SPI while its steps into the noise problem (trace attached).

I will try again, and compare the traces. Maybe I can isolate the critical command.

Is there a documentation about the SPI protocol available, or maybe reference to code, what could help?

Best Regards

Sven - DH1SV

73'

[pa3gsb]

Sven,

The radioberry is supporting the openHpsdr Protocol-1 

Protocol can be found at: https://github.com/TAPR/OpenHPSDR-SVN/tree/master/Documentation

The protocol is implemented (for the radioberry) in the firmware by C code ... the radioberry uses the FPGA as a sort of iq-server for rx and tx.

Somewhere in the development time i decided to join the gateware of the Hermes Lite -2 ; The HL-2 uses the command structuur as defined in the protocol.... and it was easy to connect to the HL-2 command structure using SPI.

The HL-2 did make modification in the protocol; described in https://github.com/softerhardware/Hermes-Lite2/wiki/Protocol

Some background:

For the radioberry i use a SPI command structure based on the P-1 ; i only added an additonal 8 bits (the RB-Command); this additon makes it a total of 48 bits; exactly the length of an i and q sample; which was used in the project to transfer the samples and the commands; off loading the work from CPU to SPI .... which was needed at RPI-2 .... for the RPI-4 bitbanging was achievable.

I kept this 48bits for controlling.

So based on the command and command data it should be easily to make a simple interpeter to see what actions are send.

Of course you are able to skip to send certain commands to the radioberry gateware....

I am curious if a command from Thetis can be found which makes the noise floor come up.... seems not logical....

Have fun in your search.

73 Johan

PA3GSB

Op dinsdag 4 mei 2021 om 20:09:50 UTC+2 schreef s.hi...@gmx.net: