MARTIN - G8JNJ

ECLECTIC AETHER - ADVENTURES WITH AMATEUR RADIO

OBTAINING MORE BITS FOR YOUR MONEY

(http://www.g8jnj.net/softwaredefinedradio.htm#970952309)

Most SDR's provide fairly good receive sensitivity, but the main factor that differentiates the performance is the Dynamic
Range. This is a measure of how well you can receive very weak signals in the presence of other very strong signals.

The two main parameters that define the dynamic range of an SDR are the number of bits used by the Analogue to Digital
Convertor (ADC) and the sample rate and filtered bandwidth that the ADC is exposed to.

If an SDR uses a high sampling rate but doesn't have an effective band pass filter ahead of the ADC, then any strong
signals within the pass-band will contribute towards the overall signal level that the ADC has to digitise. So under some
circumstances it's possible for an 8 bit RTL dongle with a 2 Mbit sample rate to out perform a 12 Bit device with a 65 Mbit
sample rate but poor input filtering. This is simply because the 8 Bit dongle ADC is not exposed to as many signals within
it's receive pass band as the 12 Bit device.

I run a 14 bit KiWi WEB SDR (http://southwest.ddns.net:8073/) using a 65 Mbit sample rate ADC with a 30MHz LPF ahead of it.
So although the SDR has quite a good dynamic range due to the 14 bit sampling, form time to time (especially at night) it
tends to suffer from ADC overload due to very strong European Broadcast stations.

I noticed that the problem was actually only due to a very few signals, which where considerably higher in level (>10dB)
than most other stations within the DC-30MHz receive pass band.

One or two other KiWi SDR's (http://www.kiwisdr.com/KiwiSDR/index.html) which are using good antennas (myself included) are
also suffering from ADC overload due to strong Broadcast stations, so here's an inexpensive way to maximise the number
of bits that are available for use by your SDR.

It's basically a very simple filter that introduces a very narrow 10dB deep notch on frequencies where there are problems
with strong signals.

A number of series tuned resonant circuits connected in parallel (shunt) with the KiWi SDR input.

The frequencies and number of tuned circuits can be varied in order to suit your own RF environment. I chose to use four
notches tuned to 900KHz, 5,900KHz, 7,400KHz and 9,400KHz, which corresponded to the strongest signals at my
location.

To help design and fine tune the component values, I used Elsie http://www.tonnesoftware.com/elsie.htm
(http://www.tonnesoftware.com/elsie.html)l (http://www.tonnesoftware.com/elsie.html) and RFSim99 http://www.electroschematics.com
/835/rfsim99-download/ (http://www.electroschematics.com/835/rfsim99-download/) (both free)

The following diagram shows the modeled circuit and frequency response curves.
22uH inductors are used in order to provide the desired bandwidth and attenuation for each frequency, but you can
change the values as required.

Almost any small mounded chokes will do, but I found it important to use mica or polystyrene capacitors.

Small ceramic capacitors may work, but the 'Q' is often very low and they drift slightly in value with changes in
temperature.

Fixed value capacitors with trimmer capacitors connected in parallel were used to fine tune the circuits to the required
frequencies.

If you wish to obtain a deeper notch you just need to change the L/C ratio, or use two or more notches tuned to the
same frequency.

Design the filter for the notch depth that is needed to drop the signal levels below the required level.

For example on 540KHz:-

a 10uH inductor and a 8600pF will provide a notch depth of about 25dB with a bandwidth of 750KHz at the -3dB points

a 100uH inductor with approximately 1812pF will provide a notch depth of about 12dB with a bandwidth of 108KHz at
the -3dB points

In the case of the lower frequencies it is usually easier to use a variable inductor and fixed value of capacitance, as
this makes tuning of the circuit a lot easier.

Old analogue AM radios are a good source of suitable variable inductors. If you can find one of the metal canned
adjustable AM 455KHz IF transformers (remove the small capacitor that can usually be found in the base of the coil)
or the AM local oscillator coil, these make great notch filters.

Note that on the lower frequency bands in particular it is important to use good quality polystyrene capacitors in
order to obtain the best notch depth.

I built my filter into a small metal box with BNC connectors that simply fits in line with the antenna feed.

Here are some plots of the actual filter response curves showing the notch depth and bandwidth.
USING A RTL DONGLE AS A UWAVE SIGNAL TEST SOURCE

In July 2014, OH2FTG (http://www.rtl-sdr.com/update-rtl-sdr-transmitting-1270-mhz/) posted a video where he

demonstrated how to use a RTL dongle as a low power transmitter, by extracting the Local Oscillator (LO)
signal. Following this revelation, IK1XPV (http://www.steila.com/SDR/RFgenmod/index.html) had a similar idea, and
modified an RTL dongle in order to achieve a higher output level.

Whilst discussing methods for testing antennas, I thought it would be useful if a RTL dongle could be used as a
low power test transmitter, in order to perform antenna gain comparisions.

A quick check with a spectrum analyser showed that there was approximately -80dBm of LO leakage from the
RF input connector of an unmodified dongle, which could be tuned over the frequency range 904MHz to
3544MHz. This is very useful as it includes the 1090MHz ADS-B, Inmarsat, GPS, 23cm, 13cm and 9cm
Amateur bands amongst others.

How to configure and tune a RTL 820T or RTL 820T2 dongle to produce an output signal using using
SDR Sharp.

Connect the RF input of the spare dongle to a suitable test antenna such as a wave dipole.

Open SDR Sharp, then:-

Select 'center tuning' mode (this keeps the LO in sync with the tuned frequency)
Set the AGC off (to stop random variation of the tuner circuits)
Set the RF gain to approx 12.5dB (this provides maximum LO leakage)

The LO signal runs at four times the actual tuned frequency, plus the frequency of the first IF stage. Which
is usually either 3.57MHz for the 820T, or 6MHz for the 820T2

This formula is valid for a SDR Sharp tuning range from approx 220MHz to 880MHz (which equates to an
output frequency of 904MHz to 3544MHz)

For example:-

To produce an output near to 1090MHz tune SDR Sharp to either 268.93MHz for a plain vanilla RTL820 dongle
or 266.5MHz for a RTL820T2 dongle (LO divided by four, minus the low first IF frequency which is usually
either 3.57MHz for the 820T or 6MHz for the 820T2, (this depends on the exact tuning commands sent by SDR
Sharp, which is why I suggest using the center tuning option).

e.g.

1090MHz / 4 = 272.5MHz subtract 3.57MHz (for 820T) = 268.93MHz

1090MHz / 4 = 272.5MHz subtract 6MHz (for 820T2) = 266.5MHz

Other useful frequencies include:-

1240MHz = 306.43MHz (820T) or 304.0MHz (820T2)

1296MHz = 320.43MHz (820T) or 318.0MHz (820T2)
 2300MHz = 571.43MHz (820T) or 569.0MHz (820T2)
3410MHz = 571.43MHz (820T) or 846.5MHz (820T2)

Add a cheap broadband amplifier if a higher output level is required.

RECEPTION OF VLF FREQUENCIES USING RTL820T2 DONGLE

When the RTL820T2 dongle is used in direct sampling mode, it is capable of receiving signals on very low
frequencies <1KHz (The modifications required to facilitate this are documented further down this
page). The following notes are a result of work I conducted for use on the SUWS WEB SDR
(http://websdr.suws.org.uk/) in this application the dongle is used in conjunction with an active antenna using a
pair of complimentary FETs. This was originally designed by Chris Trask (http://home.earthlink.net
/~christrask/) and is documented on the Active antennas (http://www.g8jnj.net/activeantennas.htm) page of
this website.

This note focuses on the additional modifications required in order to overcome problems associated with
the dongles 8 bit sampling and the subsequent <50dB dynamic range which is problematic when operating
over a frequency range of 5 KHz to 2MHz. The main issue is overloading due to very strong broadcast
stations operating on the LF and MF bands, which are required to co-exist with very weak amateur stations
operating on the 136, 474 & 1800 KHz bands.

In order to maximise the limited dynamic range of the RTL820-T2 dongle is was necessary to use an
external frequency equalisation network.

The equaliser performs several specific functions:-

4MHz Low pass Filter to minimise alias signals originating at 30MHz

20dB Variable attenuator to set the overall signal level fed into the dongle
Switched LF roll-off to optimise the performance at frequencies around 10KHz in the presence of
strong lightning surges
 -10dB notch at 198KHz to reduce level of BBC R4 broadcast station in the LF band
-10dB notch at 800KHz - to reduce level of local broadcast stations in the MF band

Circuit diagram and control panel layout

Graph showing typical frequency response of equaliser

Configuration

4MHz Low pass Filter

No adjustment required

198 KHz -10dB notch

Turn tuning adjustment to minimise level of signal on 198 KHz

800 KHz -10dB notch

Turn tuning adjustment to minimise and balance level of the strongest broadcast signals on
693 KHz/ 909KHz / 1089KHz

At the SUWS WEB SDR site, the two strongest signals are on 693 & 909 KHz, but it should
be possible to get all three reduced to approximately the same level by careful adjustment of
the centre frequency.
 Switched LF roll-off

This is designed to provide various levels of LF roll off at the lowest frequencies.

It is required because of the presence of very strong lightning pulses on frequencies

around 10 KHz, which can overload the dongle and interfere with reception of the overall
frequency range. As it is difficult to predict the performance of the antenna and dongle at
a particular location, a series of switched capacitors can be used to set the best
compromise between the performance on very low frequencies and the overall immunity
to strong lightning pulses.

The exact amount of roll-off is determined by the setting of 5 DIP switches, which select a
number of different series capacitors that are roughly arranged in a binary sequence.

Graph showing various LF roll-off settings

The following table shows the switch positions and attenuation value measured at 10 KHz.

Switch 1 is furthest to the left and switch 5 is furthest to the right.

Note that Switch 1 completely by-passes the capacitor bank.

1 2 3 4 5 dB @ 10KHz
0 0 0 0 0 -25
0 0 0 0 1 -20
0 0 0 1 0 -16
 0 0 0 1 1 -14
0 0 1 0 0 -7
0 0 1 0 1 -6
0 0 1 1 0 -5.5
0 0 1 1 1 -5
0 1 0 0 0 -3.5
0 1 0 0 1 -3.2
0 1 0 1 0 -3.1
0 1 0 1 1 -3.1
0 1 1 0 0 -2.5
0 1 1 0 1 -2.4
0 1 1 1 0 -2.3
0 1 1 1 1 -2.2
1 X X X X -0.6

Screen grab from HDSDR showing 9KHz lightning pulses with different values of LF roll-off

Whilst trying to minimise the effect of lightning pulses it is also desirable to attempt to
maximise the dynamic range on these very low frequencies.
One such example is the series of weak Alpha navigation signals on frequencies around
12 and 14 KHz.

20dB Variable attenuator

This can be used to define the overall signal level being fed into the dongle. Set strongest
signal to the top of the SDR spectrum analyser display

The 300 to 500 KHz aeronautical and maritime navigation band is a further example of
weak signals that can be heard if the overall dynamic range of the dongle is optimised.
For reference - Plot of existing SUWS WEB SDR LF Spectrum display using
temporary antenna
RTL 820T2 DONGLE MODIFICATIONS

This modification adds a direct sampling and 15MHz diplexer to the existing input
connector and provides frequency coverage from less than 1KHz to 1.8GHz.

Note that HDSDR will not receive signals in its normal mode on frequencies below
24MHz, but by using a combination of direct sampling and normal modes it is
possible to cover a very wide frequency range.

In order to achieve full coverage, I would suggest using HDSDR in direct sampling
mode to receive from 1KH to 14MHz (good IF notch filters) and SDR Sharp to
receive from 14MHz to 1800MHz (good scanning facilities).
Top of unmodified RTL 820T2 board

Bottom of unmodified board

Circuit diagram showing new input diplexer and Balun transformer.

The transformer core material I originally tested was type 73 ferrite with 10
trifilar turns, this gives fairly good performance over the required frequency
range.

The measured low frequency response of the modified dongle with a 220 Ohm
terminating load is:-

20MHz 0dB Reference

10MHz 0dB
20KHz 0dB
10KHz 0dB
5KHz -3.5dB
4KHz -5dB
3KHz -7dB
2KHz -10.5dB
1KHz -13dB

Type 43 ferrite material will also work, but RF levels below 100KHz are much
lower (almost 20dB down at 18KHz).

I've measured the input SWR of this circuit and it is less than 1.5:1. The
differential input impedance of the RTL 283 chip is approximately 3,300 Ohms,
so an additional 220 Ohm load resistor needs to be connected across the
transformer secondary in order to provide the correct terminating impedance.

I have tried other ratio transformers, but the 4:1 ratio transformer seems to
give the best compromise between signal level, bandwidth and suppression of
spurious signals. Note that I connected the Balun center tap to ground via a
1uF DC blocking capacitor, this seemed to further reduce some of the
internally generated suprii and unwanted signals produced within the Dongle.
The routing of the wires from the Q inputs to the chip back to the
transformer is fairly critical in order to avoid noise pickup from the RTL 283
on-board DC-DC convertor. Running the thin balun wires close to the board
(moving them around slightly whilst the dongle input is terminated with 50
ohms is a useful exercise). Covering the whole of the top side of the PCB
with Kapton tape and placing self adhesive copper foil screening on top of
the Kapton tape makes a further reduction in the level of unwanted signals.

The component values for the diplexer are not quite correct, but they were
the nearest values I had in the junk box. The HF port is fine but the LF port
would benefit from a few more filter sections in order to further attenuate
signals on frequencies above 14.4MHz.

This is an alternative circuit that provides better rejection of unwanted

images on the LF and HF bands.

Note that I used standard value 100pF capacitors in order to make

construction easier to implement on the PCB.

The 1uF capacitor is a DC block that I have included as a safety precaution in

order to prevent the inductors (which have a very low current rating)
from becoming damaged in the event of the dongle input being incorrectly
connected to the wrong side of the antenna bias tee.
Photo of trifilar wound Balun

Modifications to RTL 820T2 PCB

Additional decoupling capacitors added to underside of board

Track to IR detector circuit disconnected.
100pF Input capacitor added
Unwanted IR components removed
 Balun glued to PCB
Input inductor added to form diplexer circuit
Additional 1uF decoupling capacitor added to DC-DC
convertor inductor
High temperature, self adhesive Kapton tape placed on PCB
near Balun
Self-adhesive copper tape pads added on top of Kapton tape
1uF decoupling capacitor added to PCB near RTL283 chip for
Balun center tap
Balun windings connected to Q input pads on RTL283 chip
Balun input windings connected to ground and 0.68uH
inductor
Wires from Balun held in place with Kapton tape

100pF capacitor connected from junction of 0.68uH inductor

and Balun input to ground (not shown in photo for clarity)
220 Ohm resistor connected across transformer secondary
(not shown in photo for clarity)
Top side components on PCB insulated with Kapton tape
Solder resist scraped away from underside edges of PCB
PCB screened with self-adhesive copper tape which is
connected to PCB on underside edges
Copper foil patch soldered over rear of MCX coax connector

If this is done carefully, it is possible to fit the PCB back into the
original plastic case.
Since originally publishing these notes, I have experimented with
extending the LF range to cover even lower frequencies by
increasing the number of turns on the transformer. If you use fine
enough wire (and have good enough eyesight) it is possible
to thread more than 40 turns through the twin hole ferrite core. In
practice I found that 30 turns provided the best compromise
between good performance at very low frequencies and a flat
frequency response up to and beyond 14MHz.

The graph below shows the secondary impedance of each

transformer when the primary is terminated with a 50 Ohm
resistive load.
The measured low frequency amplitude response of the
modified dongle with the new transformers and 220 Ohm
terminating load is as follows:-

Frequency 30 trifilar turn 20 trifilar turn 10 trifilar

turn (original transformer)

20MHz 0dB Reference 0dB Reference 0dB

Reference
10MHz 0dB 0dB 0dB
20KHz 0dB 0dB 0dB
10KHz 0dB 0dB 0dB
5KHz -1dB -1dB -3.5dB
4KHz -1dB -1.5dB -5dB
3KHz -0.5dB -2dB -7dB
2KHz -1dB -4dB -10.5dB
1KHz -2dB -6dB -13dB

Note that even with 30 turns the high frequency

performance is still adequate for most purposes and the
overall frequency / amplitude response is good enough
for measurement purposes.

RTL2832 input impedance

Another experimenter, Robert Symanek, wrote to me and

described the results he had obtained.

You are right about the RTL 2832 chip, the input impedance
is very high, and most logic chips have a high input
impedance.

I found that Mini-Circuits makes a RF matching transformer

(T16-6T-KK81, part number) that gives about 16db to 19DB
higher signal that T37-43 toroid with 200 ohm on the
secondary .
The mini-Circuits RF matching transformer T16-6T-KK81 is
fifty ohms on the primary and eight hundred ohms on the
secondary.
That matching transformer has the lowest Insertion Loss and
Return Loss of transformers that I looked at Mini-Circuits
information and charts are easiest to read of all other
matching transformer makers, but there may be other
companies selling similar products.

After reading Roberts notes, I managed (by means of an

external resistive measuring bridge) to determine that the
differential input impedance of the RTL 2832U device is
approximately 3,300 Ohms (3K3) when the input is enabled.

Circuit diagrams of the RTL dongle can be found on this

Japanese website http://ggtoshi.at.webry.info/201406
/article_6.html (http://ggtoshi.at.webry.info/201406/article_6.html) (use
Google translate)
I had previously found that even with a 4:1 ratio
transformer, the input signal level is plenty high enough
for my purposes. But in order to improve the input match
and amplitude / frequency response, I have now modified
my circuit to include a 220 Ohm terminating resistor
across the transformer secondary, This has not decreased
the signal level by any significant amount, and the RX S/N
ratio has remained the same.

I think it would be very difficult to build a 66:1 ratio

transformer with sufficient bandwidth for this application,
But based on my experiences, I believe that Roberts
choice of a T16-6T-KK81 transformer is a good one, but
I'd suggest adding a 1K terminating resistor across the
secondary (in parallel with the 3K3 input Z = 800 Ohms) in
order to provide a flatter frequency response and better
input match with no 'peculiar' self-resonant peaks & dips.

Whilst looking at the Mini-Circuits catalogue, I also came

across another transformer with a higher value of
impedance transformation ratio, part number T36-1-KK81
which offers a 36:1 ratio over a frequency range of 30KHz
to 20MHz. This would provide a higher input voltage, but
as before Id strongly suggest connecting a 3,900 Ohm
(3K9) terminating resistor across the secondary. This is in
parallel with the RTL 8232U input impedance of
3,300Ohm (3K3) so the secondary of the transformer is
correctly terminated with a 1,800 Ohm (1K8) load
impedance.

Mini-Circuit's catalogue page http://194.75.38.69

/pdfs/T36-1-KK81.pdf (http://194.75.38.69/pdfs/T36-1-KK81.pdf)
http://194.75.38.69/pdfs/T16-1-KK81.pdf (http://194.75.38.69
/pdfs/T16-1-KK81.pdf) http://194.75.38.69/pdfs/T16-6T.pdf
(http://194.75.38.69/pdfs/T16-6T.pdf)

Note that the T36-1 and T16-1 transformers do not have a

centre tap on the secondary, so the 1uF capacitor shown
in my original modification can be omitted.

Robert very kindly sent me some samples of the Mini-

Circuits transformers to test.

I connected each transformer in turn to a 820T dongle and

measured the signal level and frequency response of
each type.

The results are shown in the following table and graph.

The T16:1 and T16-6 provide a terminated level
which is approx 7dB greater than the 4:1 ratio
transformer.

The T36:1 provides a terminated level which is

approx 11dB greater than the 4:1 ratio transformer.

As expected the upper frequency limit falls away

much faster as the transformer ratio becomes
larger.

In all cases the signal level was a few dB higher

when the transformer secondary was not
terminated, but the overall frequency response was
not as flat, and a much greater variation in signal
amplitude occurred across the desired frequency
range, especially when using the higher ratio
transformers. This variation would be much more
noticeable when connected to a typical antenna as
it would not present a purely resistive 50 Ohm
source impedance.

The 4:1 ratio offers good performance at really low

frequencies, with a very flat frequency response
over the whole range, but at a lower signal level.

The 16:1 offers good signal levels over most of the

useful operating range especially on
frequencies above 100 KHz.

The 16:6 offers offers good performance at low

frequencies and a flat frequency response
with good signal levels over the useful operating
range.

The 36:1 offers moderate performance at low

frequencies, with a compromised high frequency
performance, but a much higher overall signal
level.

Based on these measurements I would concur with

Robert's findings, that if you don't wish to wind your
own transformer, then in the majority of cases the
Mini-Circuits T16-6 Transformer offers the best
compromise between overall frequency coverage
and signal level.
I was slightly surprised to find that the 36-1 and
T16-6 Mini-Circuits transformers had better
performance at the low frequency end of the
operating range in comparision to the T16-1
version. So I carefully opened up one of each
type in order to see what was causing the
difference.
The type T36-1 and T16-6 use larger ferrite
cores and more turns of wire than the
T16-1, which is why they have enhanced low
frequency performance.
One other option, as an alternative to using
a transformer, would be to build a buffer
amplifier with a differential output.

Texas Instruments produce a range of low

noise, high slew rate, differential
input/differential output devices such as the
THS4151, which look ideal for this
purpose.

Texas Instruments data sheet

http://www.ti.com/lit/ds/symlink/ths4151.pdf
(http://www.ti.com/lit/ds/symlink/ths4151.pdf)

Software modification to SDR

sharp to provide extension of
lower frequency limit

This extends the tunable range down to

around 13 and up to 1864 MHz (previously
24 1766 MHz)

Download SDR Sharp from here and

install:-

http://sdrsharp.com/#download

Download unzip and copy this Windows

binary file into the SDR Sharp folder

Download the modified file from

here https://db.tt/0JuVpWBL

More info:-

http://www.rtl-sdr.com/new-experimental-
r820t-rtl-sdr-driver-tunes-13-mhz-lower/

More SDR Sharp plugins available from

here:-

http://www.rtl-sdr.com/sdrsharp-plugins/

http://www.rtl-sdr.ru/ Use Google translate

(plugins are in English)
Add RTL Dongle Driver using
Zadig

This application is supplied with SDR

Sharp, and is required to load relacment
drivers.

Connect the RTL dongle to your PC.

Do not run any drivers that may have

been supplied with the dongle, and
cancel any automatic windows driver
install routine.

Identify the device name allocated to the

RTL dongle by opening device manager
and checking for new USB devices. You
may need to plug and unplug the dongle
to identify which USB device is the
correct one.

Locate Zadig.exe in the SDR Sharp

folder and run.

Select the correct USB device in the

Zadig application.

If you wish to, you can change the name

of the device in order to make it easier to
identify in future.

Finally run Zadig to install the

replacement windows driver.

Software modification to
useHDSDR in direct sampling
mode

Install HDSDR from here and install:-

http://www.hdsdr.de/download
/HDSDR_install.exe

Download the special HDSDR DLL for

the Dongle from here:-
https://app.box.com
/s/7tpiy8r6qo2bbhdxtt4k

and place it in HDSDR's program

folder

Run HDSDR and change the ExtIO-

DLL settings to match

More info:-

http://hdsdr.de
/RTLSDR_with_HDSDR.pdf

Frequencies between DC and 28MHz

can now be received using direct
sampling. However due to aliasing of
the 28.8MHz clock sampling
frequency, some image frequencies
from above 14.4MHz are present
below this frequency and some
frequencies from below 14.4MHz are
present above this frequency, but
these are generally not problematic.

Example of VLF spectrum using

HDSDR in Direct Sampling mode
Simple 100KHz LPF circuit
This can improve sensitivity in
the VLF frequency range by
attenuating strong MF broadcast
signals.

RTL 820T2 dongles are available

from Cosycave UK

https://www.cosycave.co.uk
/product.php?id_product=323
(https://www.cosycave.co.uk
/product.php?id_product=323)

You may also need an MCX

adaptor or fly-lead

https://www.cosycave.co.uk
/product.php?id_product=289
(https://www.cosycave.co.uk
/product.php?id_product=289)

http://www.ebay.co.uk/itm/BNC-
Female-to-Right-Angle-MCX-
Male-Connector-Pigtail-Cable-
SDR-RTL2832U-dongle-
/121600390328?pt=LH_DefaultDomain_3&
hash=item1c4ff2b4b8
(http://www.ebay.co.uk/itm/BNC-Female-
to-Right-Angle-MCX-Male-Connector-
Pigtail-Cable-SDR-RTL2832U-dongle-
/121600390328?pt=LH_DefaultDomain_3&
hash=item1c4ff2b4b8)

REDUCING RTL DONGLE

INTERNAL SPURII AND
NOISE SIGNALS
Ive recently bought quite a
few RTL DVB-T RTL
2832U / Rafael Micro
R820T dongles to use for
various purposes.

Whilst experimenting with

these devices, Ive noticed
that one or two have much
higher levels of internally
generated unwanted
signals and broadband
noise than the others.

Prodding around with a

spectrum analyser Ive
found that there are
three main sources of
internally generated
noise.

1. Carriers at
multiples of the
28MHz clock
oscillator
2. Wideband noise
with spurs from the
USB data lines
3. Wideband noise
with spurs from the
on board 3.3v to
1.2v DC-DC
convertor

Plus received noise on

the connecting cables
which can be from any
of the above, or external
sources.

Looking first at the three

internally generated
noises.

The 28MHz oscillator is

a fairly integral part of
the design, so theres
not a lot that can be
done to reduce these
signals. Soldering the
oval shaped metal can
of the crystal to the PCB
ground plane can
reduce the level of some
harmonics slightly, but I
found that any
improvement was hardly
noticeable.

Wideband noise with

spurs from the USB data
lines and the on board
3.3v to 1.2v DC-DC
convertor could be
dramatically reduced by
adding some metal
screening around the RF
input stage. I used
some very thin brass
sheet that I obtained
from a model shop for
this purpose, but
some metal from a tin
can would work just
as well. I found that
this was easier to
implement if
I removed the IR
receiver, blue LED
and associated
surface mount
resistors from the top
side of the
board. Adding
a small patch of
metal across the
underside of the RF
connector pins on the
reverse side of the
PCB also made a big
difference. Note that
placing the whole
dongle inside a tin
can didn't seem to
help. The noise
seems to be coupled
directly between
components on the
topside of the PCB.

I decided to try fitting

0.1uF chip caps
across all the main
electrolytic capacitors
and DC rails. None of
these seemed to
make much of a
difference to the
remaining spurious
signals that could be
observed with the
antenna
disconnected, but I
thought it was worth
adding them for the
sake of
completeness.
If these
modifications are
done carefully, the
plastic case will
snap back on
without any
problems.

All of these
modifications
made a significant
reduction of
unwanted signals.
The next two
photos show the
before and after
unwanted signal
levels at around
480MHz. Note that
these signal levels
were with the
dongle set for the
maximum RF gain
of 42dB. When
used at a more
reasonable gain
setting of 30dB, all
unwanted signals
(with the
exception of
harmonics of
the crystal
oscillator) were
usually at a
level near or
below the
dongle noise
floor.

Before the
modifications
After the
modifications
I also
noticed
that the
820 sticks
have got a
back to
back SMD
package
strapped
directly
across the
RF input. It
may be
worth
checking
to see if
that has
got
popped
by static
if you
have a
broken
one.

Once I
had
sorted
out the
internal
noise
sources
I took a
look at
what
external
factors
could
be
causing
problems.

I
noticed
that
when I
connected
the
short
RF fly-
lead
that was
supplied
with the
dongle
to the
input, a
lot of
noises
Id just
got rid
of had
returned.
These
were
not
detectable
when
the
input
RF
cable
was
removed.
The
main
problem
was
the
poorly
screened
fly-
lead.

Unfortunately
the
connector
is a
male
MCX,
which I
didn't
have
in my
junk
box.
So in
the
end I
just
ran a
small
length
of
thin
PTFE
50
ohm
coax
with
good
screening
from
the
rear
of the
connector
to a
BNC
plug.
A
quick
melt
of
the
plastic
case
and
the
cable
exits
OK,
and
the
case
will
fit
back
together
again.Fortunately
this
solved
the
problem
almost
completely.

I
decided
to
cut
up
one
of
the
dongle
fly-
leads
in
order
to
see
if I
could
figure
out
why
it
was
so
problematic.
As
soon
as
I
stripped
back
the
coax
and
plastic
moldings
it
was
obvious
that
there
was
hardly
any
copper
in
the
non-
overlapped
screen,
and
a
very
poor
termination
at
the
TV
connector
end.
Fortunately
it's
easy
to
take
apart
the
crimped
connector
at
the
MCX
end
and
fit
to
some
proper
PTFE
coax. So
this
is
what
Ive
now
done
rather
than
wire
directly
to
the
board.

In
order
to
further
improve
the
dongle screening
and
make
them
more
robust.
I
made
up
this
de-
cast
metal
box for
use
on
our WEB
SDR
(http://websdr.suws.org.uk/) .

A
photo
of
inside
of
box
prior
to
fitting ferrite
beads
on
the
coax
between
the
Dongle
and
a
bulkhead
RF
connector.,
in
order
to
help
reduce
the
likelihood
of
RF
current
loops
between
coax
screen
and
chassis.
I
had
to
characterise
suitable
ferrites
for
the
best
VHF
/
UHF
choking
impedance,
as
most
clip-
on
types
or
similar
don't
do
anything
at
these
frequencies.
The
USB
screen
needs
a
good
low
Z
connection
to
the
box
metalwork
and
to
the
RF
connector
on
the
box.
I
found
that
if
I
didn't
do
this
the
screening
was
not
as
effective.
Also
any
AC
potential
difference
between
the
antenna
coax
and
PC
chassis
resulted
in
current
flow
across
the
RTL
PCB.
This
added
multiple
50
/
100Hz
noise
sidebands
to
the
LO.

The
type
of
USB
connector
I
used
was
a
chrome
plated
A
to
B
bulkhead
version
from LCOM
ECF504-
BAS
(http://uk.farnell.com
/l-
com/ecf504-
bas/adapter-
usb-
a-recpt-
usb-
b-recpt
/dp/2293818?Ntt=2293818)
HF
UP-
CONVERTORS
&
RTL
DONGLES

I've
been
experimenting
with
an
HF
up-
convertor
and
RTL
DVB-T RTL
2832U
/
Rafael
Micro
R820T. The
Up-
convertor
consisted
of
a
50MHz
input
low
pass
filter,
high
level
double
balanced
mixer,
100MHz
Local
Oscillator
and
Output
filter.
Total
conversion
loss
was
approximately
7dB
plus
an
additional
10dB
attenuator
on
the
output
to
reduce
the
signal
level
going
into
the
dongle.

I
tested
the
up-
convertor
and
dongle
with
a
variety
of
antennas,
including
a
10m
vertical,
two
different
active
antennas
and
a
100ft
doublet
with
ATU.

I
measured
the
gain
levels
for
different
frequencies
at
which
the
dongle
performed
at
the
optimum
performance
in
terms
of
S/N,
dynamic
range
and
low
levels
of
IMD.
I
also
made
sure
that
there
was
enough
headroom
to
cope
with
short
duration
increases
in
signal
levels
when
operating
at
the
optimum
levels
at
different
times
of
the
day
and
night.

I
then
averaged
all
the
results
and
plotted
them
in
Excel.
The
shape
of
the
general
gain
/
frequency
curves
tend
to
remain
fairly
consistent
between
day
and
night.
The
only
real
difference
is
that
the
overall
gain
has
to
be
reduced
by
abut
10dB
at
night
to
prevent
overload
at
some
frequencies
between
5
and
15MHz.

My
thoughts
are
to
build
up
a
3
section
band
notch
filter
with
a
center
frequency
just
low
of
the
Up-
convertor
LO
frequency,
so
that
I
can
obtain
the
required
30dB
slope
characteristic
and
use
the
dongle
and
up-
convertor
without
having
to
change
the
gain
setting
every
time
I
move
frequency.

Whilst
experimenting
I
discovered
that
there
was
a
major
contest
running
and
that
10m
was
wide
open

Heres
a
screen
grab
of
the
SDR
Sharp
spectrum
display.
A
pretty
impressive
number
of
stations
I
think
have
you
ever
seen
10m
so
busy
?

As
part
of
this assessment I've
also
been
testing
to
see
if
the
limited
dynamic
range
can
provide
adequate
performance
on
the
more
crowded
bands
such
as
7MHz.
For
these
tests
I
have
used
a
passive
broadband
vertical
antenna
in
conjunction
with
an
up-
convertor
and
RTL820
dongle.
I
used
SDR
Sharp
with
various
gain
settings
but
no
AGC
selected. In
order
to
find
a
worst
case scenario with very
high
level
signals
I
chose frequencies near
the
9MHz
AM
broadcast
band.

First
12dB
gain
-
note
the
level
of
noise
floor
relative
to
the maximum
signals
-
looks
clean
with
about
60dB
dynamic
range

Notice
the
two
diagonal
traces
from
Ionospheric
sounders
gradually
sweeping
upwards
in
frequency.

21dB
gain
-
note
the
rising
level of
noise
floor
relative
to
maximum
signals
-
Start
of
severe
intermodulation
and
reduction
to
40dB
dynamic
range
34dB
gain
- note
the
very
high
level
of noise
floor
in
relation
to maximum
signals
-
very
bad
intermodulation
products
and
only
about
25dB
dynamic
range
Finally
44dB
gain
- note
the
levelling
off
of
maximum
signal
levels
and excessive intermodulation
products
Here's
another
trace
of
the
7MHz
amateur
band
with
about
12dB
of
gain.

No
problems this
time with
intermodulation
products,
as the
maximum
signal
levels
are
a
lot
lower
than
those
from
the
AM
broadcast
stations
on
9MHz.
Just
to
prove
that
I
wasn't
cheating
with
the
above
screen
grab,
here's
a
wider
view
of
the
night-
time
spectrum, with
high
power
broadcast
stations
either
side
of
the
40m
amateur
band.
I
also
attempted
to
perform
some
two
tone
IMD
measurements
on
the RTL
DVB-T RTL
2832U
/
Rafael
Micro
R820T dongle.
However
it
should
be
noted
that
SDR's
do
not
behave
in
the
same
way
as
conventional
receivers
with
respect
to
the
production
of
intermodulation
products.
Instead
of distinctive
single
frequency
products,
the
only
observable
products
tend
to
be
indicated
by
a
rise
in
the
noise
floor,
as
the
Analogue
to
Digital
Convertors
become
overloaded
and
'run
out
of
bits'.

The
three
screen
grabs
show
in
successive
order.

1.
3rd
order
products
with
two
carriers each at
-3dB
WRT
0dB
reference
line
on
SDR
Sharp -
note
the
relatively
flat noise
floor.
2.
As
1
but
at
10dB
higher
level
-
note
how
the
noise
floor
is
beginning
to
rise
near
the
base
of
the
two
carrier
signals.

2.
As
1
but
at
20dB
higher
level
-
note
how
the
noise
floor
is
now
quite
high near
the
base
of
the
two
carrier
signals
and
that
other
spurious
signals
are
also
being
produced.
I
believe
this
demonstrates
that
if
the
SDR
dongle
gain
is
set
appropriately,
their
performance
is
just
about
adequate for the
reception
of
signals
on
the
HF
bands
just
using a
simple
block
up-
convertor.

Note
that
in
all
of
the
above
cases
I
was
feeding
the
block
up-
convertor
directly
from
a
large
broadband
antenna
with
no
additional
filtering.
I
have
no
doubt
that
the
performance
could
be
further
improved
by
the
addition
of
suitable
band
pass
filters
before
the
SDR
dongle,
if
a
greater
dynamic
range
is
desired
on
a
particular
band.

Heres
my
suggestion
of
how
to
use
multiple
SDR
dongles
with
a
single
HF
up-
convertor
for
a
WEB
SDR
front
end.
All
content
is
made
freely
available
for
personal
use
only.
Please
acknowledge
the
author
if
content
is
used
elsewhere