Receive capabilities - Skimming

783 views
Skip to first unread message

st...@radio-kits.co.uk

unread,
Oct 18, 2020, 6:22:21 AM10/18/20
to Hermes-Lite
Just in case any prospective builders are in doubt of the receivers capabilities - I've been running the 6 receiver gateway with SparkSDR on FT8 for a few days now. I typically receive 140+ countries in a 24 Hour period - the last 24 Hr period was 150 and 158 for the week.

Antenna is a low height G5RV, HL2 doesn't have the HPF fitted but no overloading seen yet.

For details put my call sign into https://www.pskreporter.info/pskmap.html

Just skimming may seem pointless but it shows where the propagation is and at what time - I need to get up at 4am!

G6ALU - FT8 24hrs 18-10-20.jpg

Steve G6ALU

Steve Haynal

unread,
Oct 18, 2020, 4:35:18 PM10/18/20
to Hermes-Lite
Hi Steve,

It is great to see you up and running with SparkSDR and skimming. I also use SparkSDR for skimming. You can find "Hermes-Lite" on the top 50 list at pskreporter. It does very well compared to other skimmers on the west coast. There are one or two stations on the east coast which do better, I think because of the closer proximity to Europe, and then the top stations always seem to be in the very active European region. I am using the 10 receiver gateware, although that can be pretty taxing on a host computer as each receiver is 384kHz without final FIR filtering. I've started playing around with Pavel's decoder which takes 4kHz IQ data directly. I want to be able to do real time FT8 decode with data from 2 coherent HL2s added together for 8 or more unique streams.

73,

Steve
kf7o

J P Watters

unread,
Oct 23, 2020, 8:08:21 PM10/23/20
to Hermes-Lite
I have a pair of multiband multiwire resonant HF antennas in a vertical inverted "V" configuration directionally oriented 90 degrees apart. I have a pair of HL2 build 9 to experiment with electronically steerable beamforming antennas.

I would like to add a third HL2 to use with transverters and would like to share one of the HF antennas using a coupler/combiners.

In searching it seems that 30db isolation is normal between ports.

Are there examples of coupler/combiners that you could recommend?

..jpw J P Watters
KC9KKO
Morris, IL USA






On Tue, Oct 20, 2020 at 12:07 AM st...@radio-kits.co.uk <st...@radio-kits.co.uk> wrote:
A neat little box... Let me know how you get on. Have you found the 10 receiver gateway?
Steve G6ALU

My CLT-L09. On O2.


-------- Original message --------
From: J P Watters <jpwa...@gmail.com>
Date: Mon, 19 Oct 2020, 23:48
To: st...@radio-kits.co.uk
Subject: Re: Receive capabilities - Skimming
Steve,

I have a small form factor I7 with a 1GB ethernet, https://www.amazon.com/gp/product/B0788DJQR9
The onboard 32gb NME isn't enough for a windows machine, upsized the NME to 512GB and RAM memory to 16GB.
I am thinking that the processor will be fast enough. 

..jpw J P Watters
KC9KKO
Morris, IL USA


On Mon, Oct 19, 2020 at 10:04 AM st...@radio-kits.co.uk <st...@radio-kits.co.uk> wrote:
Morris, not sure a PI4 has enough grunt to do it, I have a recent mid to high end PC which gets to about 50% of its capability every 15 seconds. There is some 10 receive software only it needs a more powerful PC and 1Gb connection as uses all 384KHz bandwidth.
Steve G6ALU

My CLT-L09. On O2.


-------- Original message --------
From: J P Watters <jpwa...@gmail.com>
Date: Mon, 19 Oct 2020, 15:01
To: st...@radio-kits.co.uk
Subject: Re: Receive capabilities - Skimming
Steve,

I will have to get a Raspberry Pi4 loaded up with SparkSDR, so I can free up my computer.
Following your information, I loaded the new Gateware, Configured Spark and turned on the PSKreporter.
I will have to see if there is another gateware that will allow another couple of receivers.

Thanks for the motivation.

..jpw J P Watters
KC9KKO
Morris, IL USA
image.png


On Mon, Oct 19, 2020 at 4:31 AM st...@radio-kits.co.uk <st...@radio-kits.co.uk> wrote:
Hi Morris,
Your setup will be far superior to mine!

I installed the 6 receiver gateware, used SparkSDR (latest version which is only mentioned on the SparkSDR Google group). Bands were 80m, 60m, 40m, 30m, 20m and 17m. Used FT8 and receive bandwidth was 48kHz so it went down a 100Mb connection OK. I did also try adding Wspr but found contacts on FT8 were better.

Is this enough info?

Steve G6ALU

My CLT-L09. On O2.


-------- Original message --------
From: J P Watters <jpwa...@gmail.com>
Date: Mon, 19 Oct 2020, 06:37
To: st...@radio-kits.co.uk
Subject: Re: Receive capabilities - Skimming
Steve (G6ALU),

I am writing to you outside of the group because I would keep my tinkering private until things are somewhat working. 

I haven't configured SparkSDR to use multiple receivers, and would like to mimic your multi virtual receiver configuration. 

I chose a mult band, parallel resonant wire antenna that covers as many bands as possible.All in the spirit of skimming as an endgame. I have the hardware to install a pair of like antennas 90 degrees offset from each other. Then I could one day, configure a pair of HL2 units in coherent mode and electronically spin their directional properties. It is a dream. I have one of the antennas installed, just to start with. 

So that I walk before I try running a marathon, I want to configure SparkSDR to skim a little(more like a lot).
Can you tell me the bands, frequencies that you configured in SpartkSDR that you are running on your super skimming machine. 
I would be nice to see if I can duplication your skimming to see what my perspective would be from the USA midwest, seventy miles southwest of Chicago. 

..jpw J P Watters
KC9KKO
Morris, IL USA






--
You received this message because you are subscribed to the Google Groups "Hermes-Lite" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hermes-lite...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/hermes-lite/8b3787e1-a4b0-4344-a35e-a86d6e63a515n%40googlegroups.com.

Pibo - S58WW

unread,
Oct 27, 2020, 8:37:19 AM10/27/20
to Hermes-Lite

Hermes lite 2 is realy good receiver!

73 to all Pibo S58WW

Mark Wild

unread,
Oct 27, 2020, 4:50:15 PM10/27/20
to Hermes-Lite
I fully agree.  HL2 with 133ft long wire a couple of places above you

Mark
73 de G6DDX

philip.j.s...@gmail.com

unread,
Oct 27, 2020, 6:07:37 PM10/27/20
to Hermes-Lite
I currently run a Red Pitaya and I'll see if I can compare it directly with my HL2 (both running off the same antenna). That way we can get a like-for-like comparison. 

What setup do people run for skimming?

Philip

Mark Wild

unread,
Oct 27, 2020, 7:27:16 PM10/27/20
to Hermes-Lite
Hi Philip,
I am currently running the 10 receiver gateware along with SparkSDR on Win10.

Steve Haynal

unread,
Oct 28, 2020, 1:25:58 AM10/28/20
to Hermes-Lite
Hi Philip,

That would be an interesting experiment. For multiband skimming, most HL2 users use SparkSDR. This integrates with jt9 from the latest WSJT-X. You must set a path to your jt9 in SparkSDR if not in the default path. Although there is a link on the main SparkSDR page, you will want to use one of the more recent previews. SparkSDR also has its own google group

There are 4 receivers and 1 transmitter in the stock HL2 gateware. We have two variants that skimmers like to use: 6RX has 6 receivers no transmit with selectable bandwidth of 48,96,196 or 384kHz. 10RX has 10 receivers no transmit with only one bandwidth of 384kHz and no final FIR filter. It depends on software for extra decimation. It places demands on your network and computer but is what hardcore skimmers use. This wiki page describes how to update the gateware.

Two common "gotchas" to be aware of are the HL2 LNA gain setting and filter settings. You want to set the LNA gain between +12 to +24 dB. SparkSDR also has an AGC. Set the max LNA gain to something between +20 and +24 dB and turn on the AGC. Select n2adr+hpf to ensure that the filters are not blocking any band from 80M to 10M. If you want to skim 160M too, then you must disable the hpf and accept potential QRM or overload from nearby AM stations. I think SparkSDR will disable the hpf is a receiver is open on the 160M band.

73,

Steve
kf7o


  



philip.j.s...@gmail.com

unread,
Nov 3, 2020, 7:17:38 PM11/3/20
to Hermes-Lite
I'm trying to run this experiment, but I find that after a few hours, the communication stops between the hl2 and sparksdr (I'm using windows). I don't know what makes it stop. Just toggling the power button in sparksdr brings it right back to life. When it stops working, there is no data flow from the radio to the windows box. 

Any suggestions?

Philip 




Steve Haynal

unread,
Nov 3, 2020, 11:18:10 PM11/3/20
to Hermes-Lite
Hi Philip,

Are you seeing lots of EP6 errors? These are reported on the main screen in the top right. They occur with dropped our out of sequence packets from the HL2 based on incrementing sequence numbers. They indicate a stressed network. If there is anything on your network which may cause data to delay by several seconds, then a watchdog timer turns off the HL2 to prevent runaway RX or TX operation. 

73,

Steve
kf7o

samuel kallmeyer

unread,
Nov 4, 2020, 4:00:17 AM11/4/20
to Hermes-Lite
Hi Steve & All,
I'm also experimenting on my side WSPR.
I'm using the 10 RX gateware with SparkSDR under a Rpi3B+ with 9RX active.
Workload sounds pretty good:
image.png
73, Samuel F8ACB

--
You received this message because you are subscribed to the Google Groups "Hermes-Lite" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hermes-lite...@googlegroups.com.

Alan Hopper

unread,
Nov 4, 2020, 4:13:47 AM11/4/20
to Hermes-Lite
Hi Philip and Steve,
this does sound like the watchdog triggering. Somewhere on my todo list I have a note about preventing this by sending start commands every 0.5s or so to create a simple auto restart. I think It will be optional as It could have the side effect that a remote pc won't let go when you want to use a local pc, currently you can effectively stop a remote pc by turning off the radio, I do this to use my shack pc without going into the house to stop the other pc.  The alternative is a longer watchdog or an option to disable it.  I believe the p2 protocol has a bit to turn off the watchdog, I think this was to ease debugging software but I've never had a need for this use.
73 Alan M0NNB

philip.j.s...@gmail.com

unread,
Nov 4, 2020, 7:20:03 PM11/4/20
to Hermes-Lite
The ep6 number does not increment significantly (it did initially, but I fiddled with ethernet drivers and took one switch out of the path, and it is pretty static now).

Over the last five hours, these are the stats for the number of different callsigns decoded between a RedPitaya and an HL2 -- both connected to the same antenna through a 1:2 splitter.

+------+----------+------------------------------+
| band | callsign | count(distinct senderInfoId) |
+------+----------+------------------------------+
| 15m  | N1DQ     |                          242 |
| 15m  | N1DQ-HL  |                          241 |
| 160m | N1DQ     |                           70 |
| 160m | N1DQ-HL  |                           68 |
| 17m  | N1DQ     |                          285 |
| 17m  | N1DQ-HL  |                          273 |
| 20m  | N1DQ     |                          827 |
| 20m  | N1DQ-HL  |                          879 |
| 30m  | N1DQ     |                          436 |
| 30m  | N1DQ-HL  |                          470 |
| 40m  | N1DQ     |                          853 |
| 40m  | N1DQ-HL  |                          950 |
| 60m  | N1DQ     |                          120 |
| 60m  | N1DQ-HL  |                          123 |
| 80m  | N1DQ     |                          229 |
| 80m  | N1DQ-HL  |                          242 |
+------+----------+------------------------------+

The N1DQ is the RP, and the N1DQ-HL is the HL2. The HL2 appears to do slightly better, but I don't know what the decoding depth is (by default) on the RP. [The RP was also skimming on 10m, 12m and 6m -- I'm running the 16 band receiver, but only skimming 11 bands]

The raw spot counts are not directly comparable as both systems use slightly different local deduplication strategies.
Unfortunately, the sparksdr ft8 integration does not appear to send the snr values, so I couldn't compare those either.

If you can think of other ways to slice the data, I can have a go at doing that.

Philip





Steve Haynal

unread,
Nov 5, 2020, 1:11:42 AM11/5/20
to Hermes-Lite
Hi Philip,

Thanks for sharing your data. Did you fix your issue with SparkSDR/HL2 disconnecting after a few hours? 

Since your Red Pitaya supports more than 8-bands, I assume you have the more expensive STEM-LAB 16-bit version?

Since you are skimming on 8 bands with the HL2, I am assuming you are using the 10-RX gateware? If so, why not skim on 10 bands? Were there issues?

I think SparkSDR defaults to decode depth 3. There is a global setting in the global menu, and then a per band setting on each receiver menu if you want custom band settings.

The ft8d used in Red Pitaya defaults to decode depth of 1. Pavel told me it is because the Zynq processor can't handle more. See this link:

DC7DS who is often at the top of the leader board, told me that he uses a separate computer with his Red Pitaya for the decode so that he can run a depth 3.

I suspect decode depth difference is the biggest reason for the differences in your spot counts. Maybe you can compare with SparkSDR set to decode depth of 1?

In the terminal where you start SparkSDR, do you see any "dropped" messages? That would indicate your computer is not keeping up with the decoding tasks.

73,

Steve
kf7o

Steve Haynal

unread,
Nov 5, 2020, 1:14:13 AM11/5/20
to Hermes-Lite
Hi Samuel,

I am amazed that you can run 9RX on a Rpi3B+. This requires more than 100Mbs ethernet connection. Do you have a USB gigabit ethernet adapter on this system?

73,

Steve
kf7o
To unsubscribe from this group and stop receiving emails from it, send an email to hermes-lite+unsubscribe@googlegroups.com.

Steve Haynal

unread,
Nov 5, 2020, 1:15:23 AM11/5/20
to Hermes-Lite
Hi Alan,

There is a way now to disable the watchdog. See the protocol wiki. I also have a plan to increase the watchdog timeout to tens of seconds.

73,

Steve
kf7o

S.ANICHINI

unread,
Nov 5, 2020, 3:25:24 AM11/5/20
to Steve Haynal, Hermes-Lite
Hi All,

I want to share some information with you on the configurations used by the top of PSKreporter, I use:
- 1 stemlab with 10 bands and 2 antennas (low bands and high bands) for the FT8
- 1 retpitaya (modified for a 50 ohm input) with 1 antenna (limited by SkimmerServer) with 8 bands and SkimmerServer for the CW
- 1 HL2 (10 gateware bands) for 8 FT4 bands and 3 WSPR bands (I am limited by the number of simultaneous decoding with my I5 computer> 90% during the decoding)
-kiwiSDR for 8 WSPR tapes
And so it is the whole that is reported in the statistical table of pskreporter
At the top there is DLOPF which uses a second decoding downstream of the pitaya and uses a special script to reassemble the whole on pskreporter :-))

73 from F1EYG


--
You received this message because you are subscribed to the Google Groups "Hermes-Lite" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hermes-lite...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/hermes-lite/ad20d3f4-2396-42fd-a022-10abc6de52f7o%40googlegroups.com.

Steve Haynal

unread,
Nov 5, 2020, 11:14:15 PM11/5/20
to Hermes-Lite
Hi F1EYG,

Thanks for the information! You and others have some very sophisticated skimming setups.

73,

Steve
kf7o
To unsubscribe from this group and stop receiving emails from it, send an email to hermes-lite+unsubscribe@googlegroups.com.

ron.ni...@gmail.com

unread,
Nov 6, 2020, 12:25:09 PM11/6/20
to Hermes-Lite
When a band is open, what's the typical range for the number of FT8 decodes per 15 second time slot per band slice?
I'm wondering if the fact I only get 15 to 25 or so decodes per time slot on 20M mid-day tells me something about the RF noise floor at my operating site.
73,
Ron
n6ywu
To unsubscribe from this group and stop receiving emails from it, send an email to hermes-lite...@googlegroups.com.

st...@radio-kits.co.uk

unread,
Nov 6, 2020, 12:51:51 PM11/6/20
to ron.ni...@gmail.com, Hermes-Lite
Have a look at my decodes this afternoon. This was 4 bands with a PI 4.
Steve G6ALU


My CLT-L09. On O2.


-------- Original message --------
From: "ron.ni...@gmail.com" <ron.ni...@gmail.com>
Date: Fri, 6 Nov 2020, 17:25
To: Hermes-Lite <herme...@googlegroups.com>
Subject: Re: Receive capabilities - Skimming
When a band is open, what's the typical range for the number of FT8 decodes per 15 second time slot per band slice?

Alan Hopper

unread,
Nov 6, 2020, 12:56:13 PM11/6/20
to Hermes-Lite
Hi Ron,
a good test is to upload to pskreporter and then compare your results with those close by, both quantity and snr, you can't easily compare local spots with others on psk reporter because of the ~20mins de duplication.
73 Alan M0NNB

S.ANICHINI

unread,
Nov 7, 2020, 10:12:11 AM11/7/20
to Alan Hopper, Hermes-Lite
Hello all
From 2:00 p.m. UT today, I connected to the same antenna (hexbeam):
- stemlab with Pavel's FT8 software (F1EYG)
- IC7800 with WSJT (deep decoding) (F1EYG / IC)
-HL2 with SparkSDR (deep 3) (F1EYG / HL)
WSJT and Spark are on the same I5 4-core PCThe comparison is currently just over 20m
Be careful when reading the results on PSK reporter, the interval between each UPLOAD is different depending on the software !!! it's impossible to compare over short periods, I think over a few hours it starts to be possible, the problem is that there is a limitation in pskreporter after 1900 spots, the result is always the same?
73 from Sylvain F1EYG


Steve Haynal

unread,
Nov 7, 2020, 7:28:23 PM11/7/20
to Hermes-Lite
Hi Sylvain,

PSKreporter shows you are using SparkSDR 2.0.2.3. There is a bug in that version which reports spots off by 15 seconds. This causes PSKreporter to then block spots. Before this bug was fixed, my HL2 was 30% lower than Red Pitaya in my region. After this bug was fixed, my HL2 is always better than the Red Pitaya in my region. Please use at leat 2.0.2.7 from:


73,

Steve
kf7o

To unsubscribe from this group and stop receiving emails from it, send an email to hermes-lite+unsubscribe@googlegroups.com.

Steve Haynal

unread,
Nov 7, 2020, 7:33:54 PM11/7/20
to Hermes-Lite
Hi Philip,

PSKReporter shows you are using the oldest version of SparkSDR for any user: SparkSDR v2.0.0.9

This older version may explain why things stopped for you after a few hours. Also, this older version has the bug of spots being reported off by 15 seconds which you helped track down. This bug caused a substantial drop in my FT8 spots since PSKreporter would block these spots since they disagreed with other reporters. Please use at least v2.0.2.7 from:


or v2.0.3.5 from:


73,

Steve
kf7o

Alan Hopper

unread,
Nov 8, 2020, 2:46:59 AM11/8/20
to Hermes-Lite
Hi Steve, Phillip, Sylvain,
thanks for sorting that out. Just for background - WSJTX uses the computer clock time of when it has finished decoding for the psk upload timestamp rather than the displayed times of the spots, older versions of Spark used the spot time which caused pskreporter to drop some. Spark now adds an offset to try and put the time in the centre of the somewhat random WSJTX times, thanks to Steve and Philip's help on this.
For anyone who wants to analyse spots there is an experimental version of spark here https://groups.google.com/g/sparksdr/c/zCPYaqpSyvE that sends all spots out over a websocket, these spots have the unajusted time.
I shall add a control to disable the watchdog in the next release.
73 and happy skimming Alan M0NNB

S.ANICHINI

unread,
Nov 8, 2020, 8:38:17 AM11/8/20
to Alan Hopper, Hermes-Lite
Hello all
I update the test configuration from 10h00 TU, same antenna Hexbeam 
- StemLab with Pavel softwareFT8 and 11 bandes as F1EYG (my reference system)
-HL2 with 9 bandes Rx only gateware 384 Khz  (no 10m and 6m) and SparkSDR V .2.0.2.7 with Deep 3 decoding level  as F1EYG/HL
For the test i need to stop all application on my I5 computer !! (Skimmerserver, SDRServer,antimalware,.... only Spark and i have some pic to 100% !!! )
now wait and see :-) 
73 to all from SylvainF1EYG


--
You received this message because you are subscribed to the Google Groups "Hermes-Lite" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hermes-lite...@googlegroups.com.

S.ANICHINI

unread,
Nov 8, 2020, 8:45:57 AM11/8/20
to Alan Hopper, Hermes-Lite
test starting at 13h00 TU (i forget to start again stem lab after change of configuration ;-(
73 F1EYG

sylvain....@gmail.com

unread,
Nov 9, 2020, 10:11:13 AM11/9/20
to Hermes-Lite
HI
After 24 hours of testing between the stemLab and HL2, the HL2 gives 57442 reports and the piyaya 49579 so no doubt the HL2 + SparkSDR set is more efficient than the StemLab + FT8Pavel, even if the Stemlab treats 11 bands and the HL2 only 9 It is true that the spread over 10M and 6M is very low right now here. 
On the other hand the snr are better on average from 5 to 10 db on the StemLab (same spot, same time).
 I start a new test with the stemLab + sparkSDR in deep 3 (only 8 bands spark recognize two IP addresses for the Stemlab but only one works?) and HL2 + sparkSDR in deep 3
Next result in 24H
On psk reporter : F1EYG/Py for stemLab and F1EYG/HL for Hermes
73 de Sylvain
F1EYG

Mark Wild

unread,
Nov 9, 2020, 10:19:40 AM11/9/20
to Hermes-Lite
Very interesting test Sylvain.  Look forward to the next results.

I summarised your results across the bands to see if there was any correlation.  

F1EYG comparison.png

73 de Mark, G6DDX

S.ANICHINI

unread,
Nov 9, 2020, 1:53:11 PM11/9/20
to Mark Wild, Hermes-Lite
Great Mark !!
Yes i have the same analysis , interpretation is more complex, i think the charge of CPU with my I5 computer used for sparkSDR can explain some difference ., i starting 
from 14h TU a new configuration. i keep the StemLab (racing pitaya :-))with now F1EYG/PY but i use sparSDR for decoding with deep 3 (i read that pavelFT8 software use deep 1)
See tomorrow :-))
73 de Syvain
F1EYG



S.ANICHINI

unread,
Nov 10, 2020, 7:47:55 AM11/10/20
to Mark Wild, Hermes-Lite
Hi
here are the last results of the comparison between the HL2 and the StemLab at 11:30 UT (pskreporter statistics)
-StemLab + SparkSDR = 56,446 spots
-HL2 + SparkSDR = 60,426
The difference is smaller(than stemlab+FT8 Pavel software) but the HL2 with sparKSDR remains more efficient than a StemLab with the same sparSDR software .Given the difference in cost, the HL2 is therefore the best choice for this use if you can use an existing PC, if you have to add the cost of a pc capable of simultaneously decoding the FT8 on 9 bands (a 4-core I5 overclocked to 3Ghz is just sufficient) the cost of the HL2 and the PC remains higher than the stem lab or redpitaya solution and the FT8 software from Pavel: - )
73 to All
Sylvain
F1EYG

Mark Wild

unread,
Nov 10, 2020, 9:18:29 AM11/10/20
to Hermes-Lite
Thanks Sylvain that's been a really good comparison so really appreciate you taking the time to do this.

Some of the difference is on 12m as the HL2 has 718 spots and that band isn't running on the RP.  I don't think that changes your conclusion though :-)

73 de Mark, G6DDX

Alan Hopper

unread,
Nov 10, 2020, 9:50:40 AM11/10/20
to Hermes-Lite
Hi Sylvain,
this is all very interesting, I wonder what causes the difference.  It is good to hear Spark works with your StemLab, I don't have one so rely on feedback to know it works, there maybe things that could work better. I don't believe the rf agc (auto lna gain control) works with the redpitya/stemlab as it relies on the ep4wideband packet, this may give the HL2 an advantage.  One thing I found made a difference in back to back testing is frequency calibration to avoid capturing different spots at the edges of the band, this was one of the motivations for ntp frequency calibration in Spark.  
73 Alan M0NNB

Steve Haynal

unread,
Nov 10, 2020, 12:09:48 PM11/10/20
to Hermes-Lite
Hi Sylvain,

What analog amplification does the Red Pitaya have in front of the ADC? I suspect that the adjustable LNA on the HL2 makes a difference, especially when skimming multiple bands with one antenna since there are bad impedance matches for some bands.

I run 9 band skimming with SparkSDR on a low end atom-based PC similar to the $40 atomic pi. It does take a bit of tuning to get everything running correctly. Using an inexpensive PC like this would bring the total cost down considerably. Also, in the future I would like to make some changes to reduce the PC workload, such as supporting 192kHz bandwidth in the 9 and 10 RX gateware variants.

73,

Steve
kf7o

Steve Haynal

unread,
Nov 11, 2020, 12:09:27 AM11/11/20
to S.ANICHINI, herme...@googlegroups.com
Hi Sylvain,

Thanks for the info. In case you haven't tried it, there is a 10 receiver gateware available. See: 

73,

Steve
kf7o


On Tue, Nov 10, 2020 at 1:37 PM S.ANICHINI <sylvain....@gmail.com> wrote:
Hi Steve

I believe that there is no amplifier in front of the ADC just an amplifier of gain 1 which isolates the input. I changed my configuration and I have now on the same sparkSDR instance:
- the stemlab with which I decode 8 bands in FT4 and PSK and 2 bands in FT8 (spark only allows me 8 receivers as for a pitaya 125-14 instead of 11 with the stemlab ?? )
- the HL2 with 9 bands in FT8,
- the second pitaya is connected to skimmerserver
the whole works on my I5, I did the cleaning in W10 I still have to remove the antimalware :-)) 
I will see the result tomorrow ,for instance (after 4 hours) system still running
Would be great if we could run sparkSDR with a 10 band FT8 gateware on a racing arm SBC like ODROID N2 for example :-)
The HL2 is really a great board !!

73 
Sylvain
F1EYG


S.ANICHINI

unread,
Nov 11, 2020, 9:24:37 AM11/11/20
to Steve Haynal, Hermes-Lite
Hi Steve
Yes  ! I try to follow in the forums and I saw that you published a gateware for 10 bands, I will try it as soon as I have stabilized the PC for spark. I am setting up a linux machine, i think it will work better than W10  on a small config. :-)
Tks
 73
Sylvain
F1EYG

philip.j.s...@gmail.com

unread,
Nov 21, 2020, 10:01:47 PM11/21/20
to Hermes-Lite
I'm now running sparksdr 2.0.3.6 with the 10 band gateware (sparksdr says it is version 72). However, it stops skimming after a few hours (I'm running on windows). Just two quick presses of the green power button in sparksdr brings it back to life. I haven't figured out what the timeout is yet... One of the messages indicated that there might be a watchdog timer going off -- is there any way to disable that?

Thanks

Philip

Steve Haynal

unread,
Nov 22, 2020, 1:22:17 AM11/22/20
to Hermes-Lite
Hi Philip,

There is no way in that gateware to disable the watchdog timer, but it is available in recent gateware. I need to regenerate a 10RX version. I'll try to do that soon.

The watchdog timer is at ~2 seconds. I'm curious if there is anything that might stop traffic for that long on your network.

Are you seeing and EP6 errors? Is the radio running at 384kHz bandwidth?

73,

Steve
kf7o

philip.j.s...@gmail.com

unread,
Nov 22, 2020, 10:43:45 AM11/22/20
to Hermes-Lite
Steve

The radio is running at 384kHz. I do see some ep6 errors -- and assuming that they are the count of dropped packets, then I calculate that I'm dropping around 1 packet in 100,000   (70 packets lost over 5 minutes at 25,000 packets per second).  Is that that a reasonable loss rate or should I investigate the network? 

Philip

Steve Haynal

unread,
Nov 22, 2020, 11:38:29 PM11/22/20
to Hermes-Lite
Hi Philip,

I think 0.3% is still an acceptable loss. Alan may be able to provider more insight. The gateware with 10 384kHz receivers turns out to be pretty demanding on home networks. I will work on disabling watchdog timer in that variant, and also see if we can reduce to 10 192kHz receivers for skimming.

73,

Steve
kf7o

Steve Haynal

unread,
Nov 23, 2020, 11:30:55 PM11/23/20
to herme...@googlegroups.com
---------- Forwarded message ---------
From: S.ANICHINI <sylvain....@gmail.com>
Date: Mon, Nov 23, 2020 at 8:19 AM
Subject: Re: Receive capabilities - Skimming
To: Steve Haynal <softerh...@gmail.com>


Hello Steve, Philips
I try to run sparksdr on pi4 (4Gb Gb ethernet attachment) with 9 bands  gateware (i have not yet uptdate to 10 bands)
pi working fine with StemLab and 8 bands (96kbits)  FT8 decoding (you see on pskreporter F1EYG-PI) but when i try to
test with HL2 the EP6 counter panic and turns into a packet counter ;-(  ??
there are two switches (gigaethernet and 25m cat6e cable) between the pi and the HL2, I think to do a test with the pi connected to the same switch as the HL2 to see ?

73
Sylvain
F1EYG

Steve Haynal

unread,
Nov 27, 2020, 2:03:57 AM11/27/20
to herme...@googlegroups.com
---------- Forwarded message ---------
From: S.ANICHINI <sylvain....@gmail.com>
Date: Mon, Nov 23, 2020 at 8:19 AM
Subject: Re: Receive capabilities - Skimming
To: Steve Haynal <softerh...@gmail.com>


Hello Steve, Philips
I try to run sparksdr on pi4 (4Gb Gb ethernet attachment) with 9 bands  gateware (i have not yet uptdate to 10 bands)
pi working fine with StemLab and 8 bands (96kbits)  FT8 decoding (you see on pskreporter F1EYG-PI) but when i try to
test with HL2 the EP6 counter panic and turns into a packet counter ;-(  ??
there are two switches (gigaethernet and 25m cat6e cable) between the pi and the HL2, I think to do a test with the pi connected to the same switch as the HL2 to see ?

73
Sylvain
F1EYG

Alan Hopper

unread,
Nov 27, 2020, 2:52:26 AM11/27/20
to Hermes-Lite
Hi Sylvain,
this is asking a lot from the pi due to the very high network load of 9 384k receivers. There is a setting to reserve cores in spark which can help, you need to restart it after changing this. Steve mentioned a way to protect the comms thread on linux somewhere on this forum.  Any other software that causes spikes in cpu use will also make the situation worse. Reducing the decode depth can also help.  Running spark with elevated permissions to allow it to set thread priorities may also help, I'm not sure how to do this but know that the calls to raise priority generally fail when run normally. 
73 Alan M0NNB

Steve Haynal

unread,
Nov 27, 2020, 6:58:06 PM11/27/20
to herme...@googlegroups.com
---------- Forwarded message ---------
From: Alan Hopper <al...@samsararesearch.com>
Date: Thu, Nov 26, 2020 at 11:50 PM
Subject: Re: Receive capabilities - Skimming
To: Steve Haynal <st...@softerhardware.com>


Hi Sylvain,
this is asking a lot from the pi due to the very high network load of 9 384k receivers. There is a setting to reserve cores in spark which can help, you need to restart it after changing this. Steve mentioned a way to protect the comms thread on linux somewhere on this forum.  Any other software that causes spikes in cpu use will also make the situation worse. Reducing the decode depth can also help.  Running spark with elevated permissions to allow it to set thread priorities may also help, I'm not sure how to do this but know that the calls to raise priority generally fail when run normally. 
73 Alan M0NNB

DL1YCF

unread,
Nov 29, 2020, 10:49:27 AM11/29/20
to Steve Haynal, herme...@googlegroups.com
Note that the Stemlab as an ARM processor doing the Ethernet connection,
this means, it can do a full GBit/sec even when using P1.

On the other hand, all machines which do the Ethernet with the FPGA
have a fixed speed, 100 MBit/sec for P1 and a GBit/sec for P2.

The bandwdith is about (6n + 2) bytes per sample, where n is the
number of receivers.


So do the figuring yourself . . . .
> Steve G6ALU
>
> --
> You received this message because you are subscribed to the Google Groups "Hermes-Lite" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to hermes-lite...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/hermes-lite/e3bd7407-aad1-44bf-94dd-75f263e281f2n%40googlegroups.com.
>
> --
> You received this message because you are subscribed to the Google Groups "Hermes-Lite" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to hermes-lite...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/hermes-lite/2162a237-64ba-4696-ad76-bd0216c5f1e0n%40googlegroups.com.
>
> --
> You received this message because you are subscribed to the Google Groups "Hermes-Lite" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to hermes-lite...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/hermes-lite/fb39139c-2c17-44fd-8a6c-0b957225fdc4o%40googlegroups.com.
>
> --
> You received this message because you are subscribed to the Google Groups "Hermes-Lite" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to hermes-lite...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/hermes-lite/685c56c3-c4ee-4f12-9205-c99638104db1n%40googlegroups.com.
>
> --
> You received this message because you are subscribed to the Google Groups "Hermes-Lite" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to hermes-lite...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/hermes-lite/CAPmdav8UqxtsUniRUrLYgtQTuzwSCF2tktcrd1Tew_ngLwbn0Q%40mail.gmail.com.
>
> --
> You received this message because you are subscribed to the Google Groups "Hermes-Lite" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to hermes-lite...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/hermes-lite/CAPmdav-2hDAnGFTaMJFM5ri-Xwq%3DB_xxzt5Td6DvVSawAGC6ww%40mail.gmail.com.

Alan Hopper

unread,
Nov 29, 2020, 11:02:45 AM11/29/20
to Hermes-Lite

Not sure of the context of this message but the HL2 happily works at 1GBit/s using P1, it auto detects.
73 Alan M0NNB

Steve Haynal

unread,
Dec 1, 2020, 2:27:31 AM12/1/20
to Hermes-Lite
Hi All,

Just to be clear, the HL2 does support gigabit (1000Mbs) ethernet. You may have problems with skimming 10 bands if your unit is connecting at only 100Mbs. You can tell via the LEDs which speed your HL2  has connected:

73,

Steve
kf7o
Reply all
Reply to author
Forward
0 new messages