The 6M-160M HPSDR Transceiver Project Begins

[K1JJ]

Frank:

Create a folder and dump this into it via SVN. It was just updated on Jan 29th?.  KD5TFD. You may see your CPU use go down, though I like Joe's screen skin much better.

svn://184.81.173.3/svn/repos_sdr_windows/PowerSDR/branches/kd5tfd/PennyMerge



Just talked with Phil. He said another VK is working on the line in .bat. He said they forgot to include a gain block command, so shud be available shortly.

I am going to reload the stock Merc firmware, since I don't need diversity firmware until later. They both work, but the KD5 software may like stock better.

I will look into keying OZY.  Both Phil and Joe said they do NOT have that roger beep up 2kc on AM.  Joe said he has s meter spikes and Phil said he did not.  I have no spike issue on ssb, just AM.

T

[Pete, WA2CWA]

I would suggest buying the computer piece-meal rather then from one of the major vendors that pre-installs Windows and a ton of other stuff that you don't need.  Some of it always runs in the background, eats up memory, and can increase CPU usage, latency, spikes, burbs, etc. With Tast Manager, you can see all the processes that are running in the background. Sometimes you can tell what they are; others have non-descriptor file names and descriptions.

[WA1GFZ]

Tom,
Just downloaded it a couple days ago so I have it. I'm going to read metis manual before I make any changes so I only have to do it once.
I'm sure there will be a couple bugs to fix but this thing should scream and I have the power supply to go full bore.
I'm sure Joe likes your feedback but it could take a while to sort things out. There were many times when software was released just to find some users having communication problems. they get sorted out but the guys writing the code tend to have newer machines so they don't see problems until firmware is distributed.

[K1JJ]

Thanks for the tips, Pete.  I'll look around and see what's out there.  I will definately need a 4-core for diversity later on.

With stock Mercury firmware installed and the stock KD5TFD PSDR, the CPU is now down at 10-15% !! whether the pan adp is on or not. Looks like the Merc firmware used up maybe another 10% when the pan was on before with the diversity.  The diversity software took another 30% when the pan was on. So I will look for a 4-core later on.. :-)

FINALLY, I have a working rig! But I can still see the spike on the panadapter on AM only. It is 2kc up and about 1/2 the size of the carrier.  Still not sure what's causing that. I'll try Frank's OZY key idea and then I'm out of ideas. I've now changed all firmware for both boards and have new version PSDR installed. Swapped boards around and tried all the toggles in software. The other guys sometimes see the S meter spike but not the roger beep. Imagine that - a build in roger beep that won't go away, CAW MAWN.

T

[Pete, WA2CWA]

Check this out. Was just posted by a member on the Flex reflector. Claims CPU usage 1-3 %. Running it with SDR-1000.
http://www.tigerdirect.com/applications/SearchTools/item-details.asp?EdpNo=7108906&CatId=3508#
You'll also need keyboard, mouse, OS

[WA1GFZ]

Nice looking machine. I would make sure to get a bit of extra ram.
The bus speed is key. These new ram chips scream. Good price also.

[K1JJ]

Wow - good to know that kind of performance is out there at a reasonable price. I will probably get one later on.


Frank, I jacked up the audio sampling to 92K and 2K buffer - the CPU is now at about 14-22% MAX with the pan adapter on. Happy camper with that, being a single core machine.

The spike problem - I'm starting to think I got clipped with a bad board. If you look back at my first post in this thread, you will see one ESD bag sticker ripped open. It was sold to me as a used board. The others were unopened.  Bet some gorilla man-handled it. I'm running out of ideas to fix it.  Got a few more and that's it.  Other than that the rig is working FB, OM.

T

[WA1GFZ]

Well, I just sent you the "line in" L & R connections use analog ground as the shield termination. I guess nobody uses it so Phil may have forgot it was there. I'm thinking the DSP is the issue. You might monitor line out to see if the DSP is generating a transient.
What is the input voltage into LPU.
It should be a couple volts above 12 volts so there is some head room for the 12 volt regulator to work. Can you set the power supply to 14 volts. if you put a bridge in series with the power leads you just lost 1.3 volts of headroom. I prefer a big diode across the line to blow a series fuse. The 12 volt regulator will see a transient when you key and unkey due to load on the supply changing. The output amplifier is class A so it may draw a couple watts. The TX chain is DSP input to do audio processing and conversion to digital. Then into the FPGA to do digital up conversion. The FPGA drives the output D/A to generate the waveform. Then the output amplifier. Since the signal is clean on TX the D/A is good. Since the output is on the right frequency the fPGA must be good. (they usually die hard when they die) The only thing left is the DSP and power supply. There are some internal voltage regulators that could be an issue but they would also be a problem during TX.
I bet your problem is power supply related. It appears you are generating some sort of weird modulation when you unkey and it gets through the whole chain. Try hooking penny directly to an antenna with no relays and listen to the signal as you try different things. Put the FT1000 on another antenna to monitor yourself.

[K1JJ]

Frank,

Good idea about the line in on the 25 pin. No one on the list saw that. I'll try it next.  I hope that I can use chassis ground into the audio ground input, right?

On the spike:  I checked all the board jumpers against the manual and they were correct.   I then looked at the 13V input and see it is at 14.1v and holds steady during PTT keying.  On the scope it has a slight bounce of a few millivolts on key up/dn.

The outputs of the -12, +12 and 5V regs are rock solid on the scope with the AC coupling on and sensitivity turned way up. I can see a little hash on the -12V as expected. So the supplies seem stable.

I did notice that when I first power the rig up that the pulse is either not there or weaker on the first key/unkey. But on the next ones it is back to every time and strong.

I hooked the Penny directly to the dummyload, no relays, and listened in on the FT-1000D - the pulse is loud and clear up 2kc.

On the waterfall display, I can see it after the normal carrier goes by. It looks very narrow.

I see it on all bands. The computer latency has no effect after trying tests.

I'm running out of ideas to try...  sigh

T

[WA1GFZ]

my guess the DSP is generating a transient. You are welcome to borrow my HD power supply breadboard I used before Pandora came out.

[WA1GFZ]

Tom and I spent an hour last night running tests on our HPSDR rigs. I confirm the roger beep on unkey but my rig the transient is at least 90 dB down and very short. He has a higher level. Even at 40 dB down that is 1/10 of a watt at 1 KW out. I suspect there is a register house keeping issue on unkey.
He was concerned that the S meter flies way up on unkey. In reality there is no RF connection between mercury and penny so again a register issue.
Tom will report that minor bug to the Borg.

Now for the big one that all Flex and Softrock users should be aware of.
Never shut down Power SDR by Xing out. Always hit the stop button first.
As P SDR shuts down a file is updated and saved. When you just X out random data gets stored so the next time you power up P SDR could do weird things. Tom was getting into some strange conditions of no transmitt and stuck in transmitt.
I learned this the hard way when I would have a locked up P SDR on power up and would need to reload the program. This appears to be Tom's biggest issue. I also got into trouble having too many versions of Power SDR loaded on my machine.

[K1JJ]

OK on the old version of software causing interaction problems.  I always clean out the old PSDR versions as I load new ones. But they do not have uninstall routines, so who knows what dingleberries are hanging on by doing a simple delete file.

Yes, the keying spike/transcient  problem appears to be in software timing. I was glad to see Frank's rig had it, though to a lesser degree.  I posted about it on the HPSDR board and a couple guys are checking it out.

As is, I can probably live with the unkey spike. My big concern was damage to Mercury RX and Penny losing its mind. But as you said Frank, maybe my "Xing" out of PSDR caused the lock problems later. And the spike is down about -30db on my rig, so that's not as bad as I thought.

I think the bottom line is I can be a perfectionist when it comes to shaking down new rigs. Most guys wouldn't give the spikes a second thought. But after years of experience with sequencing big rigs and seeing the damage they can do when not working perfectly, I usually push on to cure them.

Frank, I will reload PSDR today and see if it smooths out.  BTW, Gerd thinks it's softwware too and axed me to run a few simple tests for him.

Also, the line in .bat file is working - Phil sent it to me to try.  Said it works at around 2V audio in, no problem.

Tnx for the help last night, OM.  Maybe we can get on and do some chatting HPSDR to HPSDR over the next day or two.

T

[WA1GFZ]

Tom,
Remember there is no RF connection between Penny and Mercury so the S meter thing is just another sw hack. It does not hurt anything. The beep is just a dirty register somewhere. I bet in the DSP. These guys are usually quick to fix issues when they are pointed out. Glad you got the line level thing working. Looks like softwsare releases in th efuture.
I wonder if you saw the response to my loading firmware question using metis. No more big file just to move data. Metis firmware takes care of downloads. So the only issue is the jtag. Bummer they could controlled in with firmware so you don't have to remove boards and swap them around.
I have not updated anything since last spring so not that big a deal. i bet there will be a rash of them when diversity is released.

[Pete, WA2CWA]

There shouldn't be any issues with multiple versions of PowerSDR software on your machine since each one is loaded into a separate directory. I have five different versions of PowerSDR on my machine and there are no issues with that. The big issue is having multiple Firewire or USB drivers still resident and active on your machine or using firmware that isn't compatible with the version of PowerSDR that you are using.

And, yes, it has always been recommended that PowerSDR not be shut down by simply "X'ing" out the program.

[K1JJ]

Some of the guys on the HPSDR forum wanted to take a look at the screen shots and layout to evaluate what's happening with the spike and locked spurious TX problems.



THESE SHOTS ARE FOR DIAGNOSIS of the SPIKE/Transcient Penny problem on PTT UNKEY. Using Penny, Mercury, Ozy, Atlas, KD5TFD PSDR and current firmware off the HPSDR site:


Pic 1: Normal AM signal on transmit

Pic 2: The UNKEY SPIKE (to right of red line at ~ 3805 kc, middle right on screen, with a spike peak at about -68dbm. I snapped it before it reached maximum height at about -40dbm.)

Pic 3:  The spurious images up and down the band when Penny locks up from unkeying spike. Cannot stop it until power is turned off. Software has no control.  PTT is unkeyed at this time as shown by DBM meter and "ADC Overload" message from Mercury.

More below

[K1JJ]

Current Rig Layout for evaluation:


Penny is keyed via #1 pin on 25 pin socket.  

I added 20,000 ufd at the 13.8v input to the LHU regulator board.  The various voltage outputs appear stable with no bounce on keyup and keydown when viewed under high scope magnification.

T

[W1AEX]

Very strange that in the third picture of the sequence PSDR is in LSB mode rather than AM or SAM. When it's locking up due to the un-keying pulse is it also jumping modes?

[K1JJ]

Hi Rob -

No it was already in the LSB mode. That was the first time I saw it take off in LSB cuz AM or FM usually has bigger spikes under unkey due to the carrier I think.

So far, by moving the db gain per band to about 42, I haven't had it take off in spurious again. Though the spike is still there.  I'm trying a few new tests as the HPSDR guys are offering them.   Wish I had a new board to swap in... :-)

BTW, the antennas are loaded pretty heavily with ice, but nothing broke yet. How you doing 25 miles north?

T

[WA1GFZ]

Tom, I have never seen anything like that on my set up. I wonder if you deleted Power SDR and reloaded it fresh since you Xed out so many times.
Something is latching up power SDR to force it into a different mode and stuck in TX. HPSDR is just a slave so control comes from power SDR.

[WA1GFZ]

Do you have a load on the TX and is the cable any good? Try putting a cap across the PTT input to ground say .1 uf. maybe you are getting contact bounce.

[WA1GFZ]

You could have a solder short on Atlas. Odd you see A/D overload in TX

[W1AEX]

...Wish I had a new board to swap in... :-)

BTW, the antennas are loaded pretty heavily with ice, but nothing broke yet. How you doing 25 miles north?

Yah, it would be nice to be able to easily isolate if it is hardware or software. I have to agree with Frank that PSDR may be the biggest suspect here. There have been enough complaints about the "sloppy" transition from TX to RX with the Flex users that it makes you wonder. Some installations never experience it, but there are enough who do that it's an acknowledged problem.

The first link below at the FlexEdge reflector is one example. Note the comment by Tim Elision indicating that Flex is dinking around with the transition to improve the turn-around time from TX to RX. (Note: He is referring to the Quadrature Sampling Detector when he mentions QSD.)

"The noise (I call it a thurp, a combination of a thump ad a chirp) is a
consequence of keeping the QSD on while transmitting.  This was done to
decrease TX->RX transition times.  They are still working on it to make it
better."

One user mentions that adjusting the Phone RX buffer changes the behavior. Might be worth checking that out.

http://www.mail-archive.com/flexedge@flex-radio.biz/msg00717.html

Another thread here about sloppy TX/RX:

http://www.mail-archive.com/flexedge@flex-radio.biz/msg01957.html

The thread below talks about this issue emerging again with the latest beta version 2.0.16:

http://www.mail-archive.com/flexedge@flex-radio.biz/msg04947.html

I haven't read about anything as severe as what you are seeing, but it might be worth trying a variety of PSDR versions that the HPSDR guys are offering. As Pete CWA mentioned, there is no issue at all with multiple versions co-existing. I have 8 or 9 installed at last count, and as he said, they hold all their data in their own directories.

We lucked out here and the ice did not arrive! Just 6 inches of wet and heavy snow. I'm hoping the new ropes and wire on my antennas will hold up!

:O)

[K1JJ]

Frank,

Yes, I have changed out PSDR many times. I did it twice tonight. It seems when I stop the program and exit via the X, I load it back and my power output is way down and the receive box is always checked "receive only".  Sometimes the audio is way low and will not adjust back up.  I then reload new PSDR thru SVN and it works normally again.  I am definately hitting "stop" before exiting.  I think that spike may be corrupting the software, but that's hard to believe. In fact, my audio external EQ lost it's mind and had weird digits showing until I reset it. I never saw so much spike effects from such a little PW box.

With the PA set above 40db, the spurious doesn't happen now.   The mode does not change - I already had it in LSB for that screen shot above.

The ADC overload is when Penny is locked on without being keyed. It is stuck in TX and Merc is receiving at the same time.  It seems to happen when Penny is set at 38db gain.

I will try some caps across the relays, however, the problem spike occurs with no cables attached, no relays clicking, just the mox mouse.

 Rob, that's interesting info on the T/R transistion. If Flex is hearing a thurp, then it must be a problem for some. That's what I hear - a thurp   and a backlash on unkey.  At least Penny is not taking off when set higher. Frank, you saw the post about keeping your final set about 41.5 or so, right?

T

[WA1GFZ]

Why do you check receive only when you have Penny?
I will check my set up tonight
The transient does nothing to my Software. I did not see the post about adjusting the gain to kill the transient. I'll give it a try but I'm almost 100 dB down so non event.

[K1JJ]

Why do you check receive only when you have Penny?
I will check my set up tonight
The transient does nothing to my Software. I did not see the post about adjusting the gain to kill the transient. I'll give it a try but I'm almost 100 dB down so non event.

What I mean is I always manually uncheck "RX only" when using the rig, of course.  But when I hit "stop" and exit then I reboot back into the PSDR later, it is always checked. The program is not saving the file for some reason.

This is software related, but I don't see anyone else mentioning it, so it must be my computer.


*** BTW, Gerd in Germany found the same spike on one of his Penny boards. I sent you a pic of it, Frank. It's .BMP, so can't post it here.  He's looking into it.



T