T-Network Remote ATU       

This is fairly new project and it is in active progress now (as of Dec 2020 / Mar 2021). Intended to replace bulky and heavy (weight) of my current remote ATU for DXpeditions.

Target design principles - mounted on mast, attached to antenna input, fully remote-controlled via WiFi and MQTT with SDR integration via TCI, lightweight (as much as possible), less power consumption (as much as possible) and less wind resistant construction (as possible) - all "as possibles" assuming best achievable without sacrifising quality of operations.

Additional source of information regarding this project can be found at EE forum. As current page does not provide any feedback form, you may use forum to ask any project-related questions (alternatively contact me directly via email). Also the news regarding project progress are published on this forum at the same time as on this page.

A bit of warning: below is the area of programming. And the prerequisite is basic level of C++ (either familiar or willing to learn). Because of project is in active phase, I will be filing this page on ad-hoc manner. It you have any questions at this stage, please email me.


Important: The following section is for real nerds (like myself) only. For those of you, who want to know exactly "what is inside, before I buy the idea".

Disclaimer: At any point of time I do not insist that constuction, elements or software concept is "best" of "final". It is exactly opposite. The team approaches to this project in a mood of experiment, based on previous successful works.

Expectations: At any point of time I do not expect this construction will be close (or even near) on its performance/tuning capability to my current remote ATU (called by myself "140" (photo) in a name of P-140 old soviet radio station, from which the variable coil is taken). What our team target for current project is to get "new replica" as close as possible on tuning parameters to "140", while significantly reducing weight, power requirements and size. No one is expecting, that fix capacitors and coils will be same selective and descrete as variable. However, I have strong believe in ATU; on my experience it saved more than one DXpedition, which at certain point seem to be "lost", because of broken antenna.

Who is the user? DXpeditioner (as primary). I am opposite to "antenna snobs", saying that "antenna has to be 1:1 without ATU, otherwise it is not an antenna". For IOTA DXpeditioner, his "antenna snobism" evaporates after first disembarking in 4-ball waves, or after first gully landed on antenna during 1KW transmission. After that, usually, DXpeditioner has to use "finger" as an antenna, to give you your IOTA QSO. Withdraw and fail DXpedition is an alternative.


Design overview
The original idea of "ATU board" (coils, capacitors, relays) design is the combination of "ATU-100 Extended board by N7DDC" and my T-match P-140 based ATU. With the requirement to meet Australian Advanced license and to ensure continuous operations for DXpeditions, the materials selected to provide x2.5 safety factor: Amidon Т130-2, 3kV+ NPO type capacitors and high voltage relays.

Please note that there are two different ATU board designs supported: :
      T-match (or T-Network) 8x8x8 design (this project in active development/production/testing phase). Below explanations and functional/construction descriptions are mostly related to this T-match ATU design.
and
     L-Network 7x7 ATU board is also supported by same software. For functionality performance and comparative measurements I am using board version produced by UR3IQH (orderable via his web site), but with completely replaced new design Control board. For future support of 7x7 ATU the new control board design will be tested and published.

Features highlights
Both supported ATU types are operated via conbination of ESP32-based control board and universal cross-platform application (called ATU connect). Below are the highlights of what is "special" about this kind of control:

     1. Unified software mainline for both ATU types. The only real difference (in control) between those two is number of controlled relays. Apart from that both ATU types are sharing same tuning algorythms, integration and feature set. There are some tuning algorithm features, which will be available only in 8x8x8, however (surely) it is obvious for anyone who undertands the difference between T- and L-match.
     2. TCI integration with SunSDR (2/Pro/DX) line of transceivers (by Expret Electronics). At this stage TUNE, DRIVE are used to automate tuning process and VFO helps ATU to "understand" which band this is all about (predominantly this function is embedded for speed-up the tuning process).
     3. WiFi (as primaty communication media) and (in fiture)CAN protocol (TJA1051) (for situations where WiFi cannot be used). WiFi is simply connecting via either - access point or hotspot - to the network where ESDR PC is, For CAN there is a need of another small box, where receiving CAN board will be.
     4. MQTT as communication server (this runs in background on Win/Mac/Linux OS on same PC with ESDR).
     5. Possibly achievable lightweight design. We are dealling with bulky relays and toroids ... come on ... it cannot fit in a matchbox. However the target is to fit it into 150mm diameter and 320mm length PVC pipe. Why pipe? We are talking about remote ATU, right? Which should be mounted on mast? Will less wind resistance? You got it. Fully assembled boards total weight is 1150 gr. Add extra weight for enclosure (for example pipe, as described above, is about 1.5 kg extra with both sides caps and RF connectos/cables).
     6. Flexible tuning options - it will come with more than two different tuning algorithms, including "classic T-match". Accordong to the experience, different types of portable antennas may require different king of tuning algorithms - hence, here we are.
     7. 5V operations. All relays are 5V, all components are low power consumption. Everything composed to save battery power for field operations. Very usefull for DXpeditions.
     8. 1W to 100W tuning power. More - better, obviously. However, fully suitable for QRP operations.

* For L-network ATU (produced by 3rd party) parameters, including physical dimensions, please contact UR3IQH (link above).


Boards schematics
Schematics for T-network:


KiCad schematics for L-C1 and C2 boards in zip file>.

The photos
below are real boards pictures (FR4 / TG150, 1.6mm, 1oz, dual layer):
L/C1 board front and back 320mm x 150mm (click on pic to see full size in new window):



C2 board front and back 320mm x 60mm:



Power board front 89mm x 86mm:


Firmware and Software
As mentioned, there are two pieces of code required:

Firmware for ESP32 is C++ based open source code (currently work in progress) and will be available at GitHub when release is ready.


ATU connect application
It is the same application used in my other remote ATU projects. Intended to run on same PC with ESDR. Integrated with SunSDR/ESDR via TCI. Some example screenshots of the SW is below.

Dashboard provides main operation:


ATU type selector (on left) allows to select between different ATU type for current setup..Tuning algorithm (on right) provides selection of desired tuning algorithm.

I.e. Algorithm 1 is T-match with L tuning first, then C2 (hot), then C1 (cold). Algorithm 2 engages C2 first, then L and then C1. Because of major antenna parameters (such as capaticance, resistance, reactance, etc) may vary for the same antenna dependent on antenna's locationfor, hence (from the experience), different algorithms may be needed for same antenna, used at backyard with salty soil or in DXpedition, when same antenna is installed at waterfront.

Each antenna type may require different starting L and C1/C2 values. Those values may be dependent on band, antenna resonanse (and how far from it we are trying to tune) and antenna parameters at and out of resonanse, SW provides virtually unlimited number of antenna memories:
As the physical ATU construction provides kind of overkill at each C1 and C2 network total capacitance, hence, it is highly recommended each antenn to be pre-measured and C1/C2/L starting tuning point to be loaded.

In addition, there is an extra option for one-, two- and three-stage tuning of each algorithm. This is achievable via Tune menu and selection of JSON messages, sent from ATUconnect to ESP32. The SunSDR power level (TCI-based) for each stage is selectable by user in the same menu option:

JSON message format and its reading described below.

Lets say Angorithm 1 - L, then C2(hot), then C1(cold) is selected. The following JSON format:
{"stage":[1],"tune":[25]}
will engage tuning algorithms at 25% of SunSDR TUNE power and algorithm will run once tuning L, then C2, then C1.

JSON message like
{"stage":[3],"tune":[25,50,100]}
will engage same algorithme by rinning stage 1 tuning in 25% of power, them repeating at 50^ and then at 100%.

JSON messages are linked to Dash Tune buttons as shortcuts,

At glance - the combination of user-selectable starting tuning C1/C2/L values and selection of particular tuning algorithm with 1, 2 or 3-stage tuning on different power levels, provide quite powerfull mechanism to tune wide range of untennas.

Please note : application is Qt.5 based cross-platform. The test releases for Mac and Win platform will be published here.

Please note that this software is currently under code cleanup and optimisation, hence the acceess to code is for development team only. Additionally we are working on feature enhancements:adding new features to allow granular control of the solid state remote ATU and provide more flexibility in tuning "unknown" DXpedition's antennas.



To the construction facts:
C1 network and C2 network
5pF / 10pF / 22pF / 39pF / 78pF / 160 pF / 330pF / 660pF
total 1304pF with 255 step tuning combinations and 5pF/step for each network
Measured parasitic capacitance of assembled boards is about 35-40pF.

Parameters and datasheets:
5pf combined out of 2x10pF in series. 78pF is a combination of 2x39pF in parallel. 160pf is the parallel of 150pf and 10pf. And 660 is 2x330 in parallel.
10pF, 22pF, 39pF are Murata Electronics MLCC C0G(NP0) 3.15kVDC with 5% tolerance.
150pF and 330pF are AVX MLCC C0G(NP0) 3kVDC with 5% tolerance.

Coil wire is Single Core 2.03mm (14 SWG) Copper Wire.

L network
0.05uH / 0.1uH / 0.2uH / 0.4uH / 0.8uH / 1.6uH / 3.2uH / 6.4uH
total 12.75uH with 255 step tuning combinations and 0.05uH/step

0.05uH, 0.1uH, 0.2uH, 0.4uH and 0.8uH coils are frameless inductors, made from above wire. Remaining coils are wired with same wire on Amidon T130-2 (1.6 on single toroid, 3.2 - on dual and 6.4 - on 3xT130-2).

Relays are Omron Electronics G2R-1-E-DC5. Using 5V relays makes crucial difference in power requirements and overall device power consumption.

Main power input 12V, then reduced to 5V by DC-DC converter (one per network and one for logic) is based on LM2596S-12 stabiliser. Input voltage can be tuned between 6 and 45V if higher power source desirable. Overall device max power consumption (all relays switched on) is around 300W (12V 2.5A). In real operations used power is about 100-150W (depends on how many relays engaged).

Relay switching is performed on positive wire (this gives fair enough RFI protection, compared to ground wire switching). HW version 2 uses SN754410 Quad Half-H bridge drivers to operate the relays. When relay is off, the "positive" is connected by bridge to the ground, preventing relay engaging by RFI.

Bridge drivers are operated by MCP23017 GPIO expander and ESP32 GPIOs. MCP23017 is operated by ESP32 via i2C bus.

ESP32-WROOM32U with external WiFi antenna should provide up to 50-60meters (open space) signal distance coverage with acceptable signal level. For situations where WiFi is not an option, the future designs will be provided with CAN bus module. (However there is the feedback for using CAN media: additional (receiver) device is needed. This can be based on any Arduino (like Nano), or on ESP32D. I have working/tested device in hand, will add construction and code details later).

For ADC the ADS1115 is used. It is 16-bit and gives good enough resolution (compared to 12-bit ESP32 in-build ADC). ADS1115 sencitivity in the given construct is 1mV with total range from 2 to 4096 mV in each A0 and A1. "Lower" 2mV is the parasitic noise with given L/C filter and shunting 100K resistors (R6 and R7 on C2 board). This brings us to capability of guarranteed 1W starting measurement/tuning point. Voltage devider provides with 5mV...4V measurement range; all used ADC channels are clamped to 5V. ADS1115 is set to 16 SPS and tuning algorithm is set to wait 8 full measured cycles before engaging next tuning step.

Measured ADS delay with given T-match construct is 10usec. Important: make sure that C31 and C32 on C1/L board and C3 and C5 on C2 board are not soldered. Those four capacitors are kind of "spare", in case of extra RFI protection needed. If those capacitors solfered in - this will create quite large overall capacitance of combined L/C filter - large enough to extend ADS delay tp 15ms. Obviously, higher delay will create extra delays in starting FWD/REV measurements.

Optimal FWD/REF measurements are expected in range of 40W-75W of tuning power. Lower tuning power (for example for QRP operations) will gradually reduce ADS lower sensitivity limit, hence ATU will reach minimal sensitivity 2mV limit on REF sooner. This limit may still not nesesserally be at optimal tuning step of given antenna setup. 20W of tuning power (native SunSDR2/Pro) is good enough compromise with usual 1:1.2 - 1:1.3 SWR is typicall reached before ADS hits lower 2mV on REF.


There are currently three tuning algorithms for T-match, and two algorithms for L-match. Each algorithm is targeted to minimum achievable SWR (withregards to different step size, step direction, etc. in each algorithm). If CWR 1:1.01 reached at any step - this ends the aldorithm.

L, C1 and C2 start positions are depending on band and antenna. Those parameters are tunable and available for end user via operating SW. Definition of "step deviation" (below) is as following (example):

      Step 1. C1-network starting point 1204pF (0x11001111).
      Step 3. First tuning step (say, Algorithm1(below) in higher capacitance direction) will be 8 steps away from starting point, i.e. 0x01011111 (or 1277pF). At this stage SWR comparisson is performed and if:
      Step 3. - SWR becomes higher, then reverse tuning step (towards lower capacitance direction and 8 steps away from starting point) is enanged.
      Step 4. - SWR becomes lover, then algorithm will continue in same (higher capacitance) direction with +1 step, i.e
0x01011111
0x11011111
0x00111111 and so on, untill SWR comparisson will detects SWR become higher (goto step 5), or untill end of combination reached (goto step 6).
      Step 5. At this stage logic reverses to step 3 and repeats until minimum achievable SWR reached.
      Step 6 (and following further, say, for Algorithm1 below), is to engage C1 (with 8 step deviation) and then L (with 16 step deviation) on described above principle - this will complete Stage 1 and then repeats with double lowered step deviation for Stage 2 and Stage 3 respectively.

Depended on antenna type, antenna parameters (i.e. resistance, reactance, capacitance, etc) and antenna location (soil type or nearby water front) the tuning algorithm can be adjusted by user to select either order: L - C2(hot) - C1(cold) or C2(hot) - L - C1(cold). In addition to tunable initial (or start, before tuning) L, C1 and C2 values this allows wide range of antennas to be tuned.