SDR comparison on Skimmer Server JF2IWL , Dai Nagakura ( revised on 2017/08/06 )

QS1R ( Spec. )
Red Pitaya ( Spec. )
S9-C ( Spec. )

QS1R = 192kHz x 7 bands It can run CW and RTTY SkimServ simultaneously.
Red Pitaya = 192kHz x 6 bands and can run CW and RTTY SkimServ simultaneously by modifying CWSL_Tee
S9-C = 192kHz x 8 bands It can run CW and RTTY SkimServ at the same time.


NEW 2017/08/06
Three Red Pitayas decorder number comparison on IARU HF Championship
It is inferred that capture ability is positive function of decorder number. On this test , I compared three Red Pitaya which decorder number ratio = 1 : 2 : 3

Testing condition

On IARU HF Chanpionship held on 2017/07/08, I set same three Red Pitaya-10's decorder number ratio is always 1 :2 : 3 by tuning in front of each Preamp-ATT.

It was very difficult because decorder number varies from 100 to 1000 every moment, then ratio is in the rough.

Each capture data is processed by almost same spec. PCs , Core i7 Win7 64 bit.

I use three IDs JF2IWL-# , JF2IWL/2-# and JS1JRZ/2-#(My another callsign)

24 hours Capture local Spots.txt is analyzed by SH5 Contest analyzer.

Three Red Pitayas ( 10 bit ) during test

Red Pitaya No.1 ( as the standard = Decorder number = 1)

6,616 Capture

Red Pitaya No.2 ( Decorder number is twice of No.1 )
21,277 Capture

Red Pitaya No.3 ( Decorder number is triple of No.1 )
26,343 Capture

Plot graphics

If Decorder number is zero , Capture number must be zero.

So , I add X=0 , Y=0 point.

*Note This graphics has no mean for mathematically even R^2 linear regression shows good relation.

This chart shows only that cupture number is positive function of decorder number .

( Curve line may be logarithmic of decorder number.)

It is so easy to increase decorder number by pushing high gain pre-amp . But the subject is how to emerge " Clean Decorders " as much as possible. Well known that SDR needs good front end , then To do is improve front end of SDR and reduce noise. Please note Skimmer's difficulty is that we must hear broad bands , HAM or BCLer listen only one signal on the other side.

Taking mesures
(1) The best

K3LR's Skimmer System (Photo from K3LR Website )

The most important front end is Antenna. One antenna tuned to amateur band to One SDR is the best way .

K3LR uses six mono-band antennas for six SDRs.

( Gorgeous but Expensive)

(2) Mono band Antenna combine

OL5Q Skimmer uses combined mono-band Yagis to one QS1R .
( Drawing from OL5Q website )

I use three SDRs three antenna to make 10 bands Skimmer.
See my Skimmer page.

I think broad band antenna like small-loop is not good for Skimmer . Multi-band GP may be better .

(3) Broadcast band rejection ( HPF )
To avoid strong MW broadcasting signals , HPF is must .

I use W3NQN type BCB reject filter. If receive only Skimmer , Cheap one is enough .

(4) Good NF Pre-Amp
Red pitaya is lack of front gain , It needs Pre-Amp .
( around 20dB )

I do not recommend bauble one ( sold on eBay 10$ ) because these spread noise and spurious in the band .

There are many good pre-amp kits for HF band like this
( SV1AFN's )

( Photo is Diamond Antenna's DMAX50 )

(5) 1 : n Galvanic step-up transformer
Red Pitaya's imput impedance is high
( It is said 1k Ohm ) .

To match to anntena impedance , 1 : n step up transformer is recommended .( n = 9-16 )

It's easy to make using toroidal core . if you want 1 : 9 transformer , wind coils 1 : 3 or use Mini-Circuits'.

Bob , N6TV provides Genuine Red Pitaya 14:1 Impedance Transformer Optimized for connecting HF antennas to the STEMlab 125 series
Jumps to Bob's page

Galvanic isolator is said it reduces common mode noise.

(6) Notch filter
There are many strong signals beside amateur bands. Multiple LC notch filter is effective to erase these signals.

It can calculate by some simulator .
( Chart is 4 Notch filter for 1.1MHz strong Nippon Broadcasting System. , 7.3 MHz Radio Beijing , 13.6MHz Japan defence forces and 19MHz broadcast band. )

(7) Band laundering
In SDR age , BPF board is very popular. Normally, this is used single-band by switching relay .

For Skimmer , Use it in parallel multi-bands by by-pass relay to obtain clean amateur bands .

(8) Clean power supply

AC/DC switching power adapter for SDR spreads noise.

Use classical transformer regulator power supply .

(9) LAN isolation
To avoid noise from LAN , The best is use LAN isolator. But It's too much expensive .

Use good LAN cable resistant to noise and wind to Ferrite core .
To handle Red Pitaya propery and run Skimmer , See
CW Reverse Beacon Network How-To Guide and
Red Pitaya Combine CW & RTTY Skimmer & HDSDR by Bob N6TV

QS1R , S9-C and Red Pitaya-10bit CW and RTTY SkimServ one week testing

Testing condition

The antenna is Diamond Antenna's CP-6 , 6bands GP, three splitted to each SDRs .

After RF guards , Diamond antenna's DMAX50 pre-amps are added front of each SDRs .

CW SkimServ was tested by three SDRs, QS1R , S9-C and Red Pitaya-10bit.

RTTY SkimServ was by two SDRs QS1R and Red Pitaya-10bit using CWSL_Tee.

At first , I carefully adjust each pre-amp gain that SDRs decorder number is almost same.

Pre-Amp for Red Pitaya +20+15dB is to much !!. the second pre-amp should be replaced to 1: n step-up transformer.

I used Red Pitaya-10 by Normal Low ATT jumper.

On CW SkimServ , I set ini file capture " ALL SPOTS " to see IBP Beacon reception.

I ran SkimServ from 06/26 00Z to 07/02 2359Z one week.

After testing , I analyzed local Spots.txt files by SH5 contest analyzer. These includes callsign busted and broken..

JS1JRZ is my another callsign.

Capture frequency segment
QS1R ( JF2IWL-# )

I set capture frequency segment all the same.

S9-C could run RTTY SkimServ but it is not good at it. Then I use S9-C for only CW SkimServ.

Red Pitaya-10 ( JF2IWL/2-#)

S9-C ( JS1JRZ/2-# )


CW SkimServ on QS1R
18,386 times , 125 Countres , 4,924 Uniques , km/QSO = 3,494 km

CW SkimServ on Red Pitaya-10
19,508 times 133 Countries , 5,142 Uniques , km/QSO = 3,650 km

CW SkimServ on S9-C
17,583 times 123 Countries , 4,776 Uniques , km/QSO - 3,501 km
* SkimServ on S9-C sometimes stop decoding may be because anti-virus software interrupt it . See break time.

Mean capture SNR
QS1R = 16.7dB
S9-C = 15.6dB
Red Pitaya-10 = 16.4dB

RTTY SkimServ on QS1R
2,110 times , 54 Countries , 338 Uniques , km/QSO = 4,092 km

RTTY SkimServ on Red Pitaya-10
2,357 times , 59 Countries , 366 Uniques , km/QSO = 4,322 km

Mean capture SNR
QS1R = 19.5dB
Red Pitaya-10 = 20.0dB

The present time , QS1R and S9-C is discontinued , Choices for multi-band skimmer is only Red Pitaya.
From this testing, Red Pitaya-10 has good ability for SkimServ if handled properly , Good Pre-amp , Impedance matching.
We must carefully check intermodulation of Red Pitaya to avoid miss-decode.

QS1R , S9-C and two Red Pitaya-10s CW and RTTY SkimmerServer testing on RDXC and BARTG HF RTTY Contest 2017
Testing condition

The antenna is same four splitted to each SDRs .

After RF guards , Diamond antenna's DMAX50 pre-amps are added front of each SDRs .

CW SkimServ was tested by three SDRs, QS1R , S9-C and Red Pitaya-10bit.

RTTY SkimServ was by two SDRs QS1R and Red Pitaya-10bit .

At first , I carefully adjust each pre-amp gain that SDRs capture SNRs are same ( within +-2dB ) . But when contest is started , this coordination is wrecked . Then I simply add maximum gain +20dB to all SDRs .

On core i7 PCs , CPU load was around 5% at CW SkimServ. 10% on Core i5 PC , RTTY SkimServ CPU load was both 60% on Core i7 PCs in the contests.

After contests , I analyzed local Spots.txt files by SH5 contest analyzer. These includes some normal QSO and callsign busted and broken..

JS1JRZ and 7K1NUY is my another callsigns.

QS1R v.s. S9-C v.s. Red Pitaya-10 on RDXC 2017 CW SkimmerServer

12,182CQs , 90cty , Unique=1,247 Kilometer/QSO=6,478km

12.382CQs , 93cty , Unique= 1,311 Kilometer/QSO=6,438km

Red Pitaya-10
8,646CQs 83cty , Unique=1,025 Kilometer/QSO=6,288km
* It lost three hours from 19:41Z by PC power shut down .

QS1R v.s. Red Pitaya-10 on BARTG HF contest 2017 RTTY SkimmerServer

3,310CQs 59cty , Unique=383 Kilometer/QSO=5,820km Busted=0.5% ( by N2QT Mark )

Red Pitaya-10
3,713CQs 67cty , Unique=411 Kilometer/QSO=6,073km Busted=0.5%

Red Pitaya drove good performance the first half of RDXC. but after PC fairue , It may " Catch a cold ", sometimes stop decoding . S9-C is good next generation SDR. On RTTY , Red Pitaya is better than QS1R .

Capture amount or rate may be positive correlation of Decorder number of SkimmerServer . For good decoding , It needs " Juggring " of Front Gain and S/N of system .

My conclusion is that alll three SDRs is good for multi-band Skimmer if handled properly .

QS1R v.s. Red Pitaya
RTTY Skimmer Server on JARTS RTTY Contest 2016
Testing condition
On RP ,it is recomended "Direct capactor Jumper ". But then RP stops decoding in 20~30 min. so I set RP " Normal ATT Jumper " and added 15dB Pre-Amp to RP's front end.

I used " CW Reporter " by N1DQ Philip on each PCs to send captured spots to PSKR and drew ADIF data from PSKR .These data are converted to Cabriilo log format and analyzed by SH5 Contest analyzer. Data may include busted and broken.
Pse note these result is just only on my testing environment. *JS1JRZ is my another callsign.

5,818 spots , 77 ctys in 48 hours Kilometers per QSO = 4,912km

Red Pitaya
5,911spots , 81ctys in 48 hours Kilometers per QSO = 4,935 km

CW Skimmer Server , 24 hours ( same testing condition as above )
574spots , 29ctys Kilometers per QSO = 1,997 km

Red Pitaya
589 spots , 32cty Kilometers per QSO = 1,969 km

QS1R v.s. S9-C
Testing condition
PC for QS1R is Core2Duo , Corei7 for S9-C , 4 bands Using splitted same antenna and 20dB Pre-amp is inserted before splitter
CW Skimmer Server in same 24 hours
2,978spots , 77ctys Kilometers per QSO = 4,310 km

3,236spots , 81ctys Kilometers per QSO = 4,384 km

RTTY Skimmer Server on Russian WW RTTY Contest
Testing condition
Both PC is Corei7, Using plitted same antenna and 20dB Pre-amp is inserted before splitter.
1,768spots 44ctys Kilometers per QSO = 5,801 km

805spots , 28ctys Kilometers per QSO = 5,490 km

<-Back to JF2IWL home