• AfuSW-PA-45W

Vorläufig / Preliminary, work in progress

Translate (Übersetzte) with Google

SW PA-45W

This web page documents the commercial 45W PA Power Amplifier kit, labeled DIY 45W SSB HF Linear Power Amplifier ....

It is offered on https://www.banggood.com with the ID number 1018834 for about 15 EUR. See the picture on the right. This shows the board with all modifications.

When I compare the achieved output values with the power amplifier of WA2EBY (QST April 1999) it looks comparable.

If you click on a picture it will be expanded. You can go back to the web page with the back arrow of your web browser.

.

.

.

Technical Data

• Output Power: 3.5 Mhz-30 Mhz, up to 70 W.
• Power Supply: 13.8 V - 26 V DC
• Version: RF_AMP_530_V306, 20160621
• Bias current per MOSFET: 30 mA
• Size: 60 x 100 x 25 mm

When you click on a picture or diagram it will be shown in the original size for better reading.

Connections

• J1: RF output @ 50 Ohm

• J2: RF input @ 50 Ohm

• VCC_PA 13.8 V, 0 V

• Unfortunately no PTT relay function is supplied.

• A PTT relay was added, with the following modifications:

  • Connector J2 (Input) was soldered horizontal to the PCB.

  • A new SMA RF output connector was soldered 25.4 mm beside J2.

  • For the connectors placing see the picture on the right.

The picture on the right shows the complete assembled Power Amplifier with Power Supply and Dummy Load.

The shown PSU current of 81 mA is the quiescence current of the MOSFET's plus 20 mA for other supplies.

The thermostat shows a temperature of 25.6 °C at the heat sink.

The little green spot on the lower right position of the PA shows that the filter band 40 m is selected.

Modifications

In the origin layout the power amplifier has three stages. Because I have more than 5 W for input, I connected just to the last stage, two MOSFET IRF530 in push-pull arrangement.

See the schematic on the right side. For better printing download PA-45W_mod.pdf. If you want to make modifications to the schematic download PA-45W_mod.sch (Eagle version 7.5).

In order to improve the input impedance, a 2.6 dB Pi attenuator (330R, 15R, 330R, each 1 W) was inserted after the input connector.

The input connector was soldered in 90 degree turned, so it can be used if the PA board is mounted in a housing.

A second output connector (SMA) was soldered parallel to the input connector, and wired to the PTT relay. See the picture in the chapter #Low_pass_filter.

Because the MOSFET bias adjustment is very coarse by RV1 abd RV2 (4k7), I inserted in a first step a 4K7 SMD 0806 resistor R93, R94 in the RV1/2 pin to 0V. But because the bias tuning is still difficult to tune, I replaced RV1/2 with a standing 5K 25 turns trimmer (10 mm wide). The original trimmer is difficult to unsolder, it is better to use the 25-turn trimmer as a replacement when building the circuit.

In order to stabilize the MOSFET bias current while heating up the MOSFETS's, a circuit from DK1HE was used, with parts U2, R95, C93, C97, D1 and D2. The diodes D1 and D2 must be mounted on the heat sink, close to the MOSFET's. The other parts (SMD housing) are mounted on the fly beside the 12 V regulator.

Capacitor C4 = 270 pF, and C29 = 100 pF are not used, just C30 = 100 pF.

Transformer T3 has original 2 X 1 to 4 windings (VCC = 12 V). If you use for supply 24 V (recommended) reduce the secondary winding to 3 turns.

The 9 V regulator was exchanged by a 12 V regulator 7812, in order to feed a 12 V PTT relay.

As a PPT relay a Panasonic DS2YS12 12 V is used. At https://www.conrad.de you can buy it for about 2.20 EUR.

For testing a 30 V supply the capacitor C3 1000 uF 25 V was exchanged by a capacitor with 35 V. Unfortunately one MOSFET blows at 30 V, so I stood with 25 V.

Thermal runaway

At the beginning of the use, I had massive problems with thermal runaway, means that at a high supply current the current still increases in a steady state situation. The reason is the rising ON-resistance with rising temperature. That sometimes killed the MOSFET's.

The cure was to use 2 silicon diodes (e.g 1N4148) to control a 5 V regulator as a circuit to lower the bias voltage in case of a temperature rise. To make it work, the diodes must be mounted (e.g. with hot glue) beside the MOSFET's on the heat sink.

• The cathode end is connected to ground.

• Usually the bias voltage of the MOSFET's is about 3.6 V.

The temperature coefficient of a silicon diode is about -1.7 mV / K. In a practical case of a 30 °C temperature rise, the voltage at the 2 diodes drops by about 102 mV.

With that modification in place I have not seen any more a thermal runaway, and the supply current changed from 66 mA at 21.3 °C (heat sink) to 70 mA at 28.7 °C only.

Cooling

A very important point is cooling. In the beginning I mounted the MOSFET's IRF530 with the supplied silicon pads on a heat sink (150 x 100 x 25 mm). At a power dissipation of about 50 W it happened that a MOSFET died.

The heat sink is a Fischer SK508 with the Thermal Resistance of 1.6 K/W.

It looks like, that the MOSFET isolation (TO-220 housing) to the heat sink has a too high thermal resistance.

So, I started searching Application Notes about that topic. The thermal resistance °C/W for a TO-220 housing to a heat sink with thermal compound (also named grease) varies between 0.5 to 1 °C/W, see the following list of publications (sequence of findings):

• http://www.richardsonrfpd.com/resources/RellDocuments/SYS_10/accessories_64-68.pdf

  • TO-220 mounting torque 0.9 Nm 0.50 °C/W, torque 1.7 Nm 0.16 °C/W
• ON_AN1040-D.PDF Mounting Considerations For Power Semiconductors 2001
  • page 4 directly to heat sink
  • TO-220 mounting torque 8 Lb/0.9 Nm lubed 1.0 °C/W
• Vishay_TO-220_torque_72674.pdf 2003
  • page 3 directly to heat sink
  • TO-220 mounting torque 7 Lb/0.8 Nm with grease 0.97 °C/W, torque 15 Lb/1.4 Nm with grease 0.90 °C/W
• Philips_9701.pdf 1996
  • page 1 directly to heat sink
  • TO-220 mounting torque 0.6 Nm with grease, screw 0.5 °C/W, clip 0.3 °C/W
• Fairchild_ON_AN-4166.pdf 2014
  • page 6 directly to heat sink
  • TO-220 mounting torque 0.6 Nm with grease 0.9 °C/W
• Infineon-IGBT_MOSFET_package_super220_mounting_guidelines-AN-v01_01-EN.pdf IRF AN-1000
  • page 5 directly to heat sink
  • TO-220 contact force 20 N with grease 0.5 °C/W

Concerning the electric isolation of the TO-220 housing to the anodised aluminum heat sink I tried as a trial a direct mounting with thermal compound. It had worked once, but any little metallic particle or scratch in the heat sink ruined the isolation. So I gave up in favour of an isolation pad. But what to use? I found a very detailed description: Design of heat sinks. There is listed:

• Mica (Glimmer) with compound: ~ 0.75 - 1.0 °C/W

• Aluminium Oxide with compound: ~ 0.4 °C/W

• Sil-Pads no compound: ~ 1.0 - 1.5 °C/W

The easiest to use are the Sil-Pads, but the best are the Aluminium Oxide Pads. I bought some of them (housing: TO-220, hole: 4 mm, smoothened) at Conrad.de. More details on quick-ohm.de

The thermal characteristics conductivity was give with 25 W/m * K. So, how to come to the thermal resistance for a TO-220 housing?

• Rth = length [m] / thermal conductivity [W/m*K] / area [m2]

for housing TO-220, 
length (thickness) = 1.5 mm = 0.0015 m
thermal conductivity = 25 W/m * K
area = 1.8 * 1.2 cm = 2.2 cm2 = 0.00022 m2
Rth = 0.0015 / 25 / 0.00022 = 0.3 °C/W

This is just for the pad. Added are 2 layers of thermal compound.

A quick calculation, what power could be dissipated, first the conditions:

• IRF530: 1.7 °C/W
• Heat sink Fischer SK508: 1.6 °C/W

• Aluminium Oxide Pad: 0.4 °C/W

• Maximum MOSFET die temperature, with some safety: 150 °C
• Environment temperature: 25 °C

Then the calculation:

• Total thermal resistance: 1.7 + 1.6 + 0.4 = 3,7 °C/W

• 150 °C - 25 °C = 125 °C / 3.7 = 33.8 W, dissipation max. x 2 = 67.6 W

• At maximum power at 3.5 MHz: Out 70 W, supply 120 W, dissipation 50 W

In the best case of SWR below 1:3 (25% loss, about 18 W, see here, or calculator) that will be OK (50 + 18 = 68 W), but at a SWR higher 1:4 (50 + 25 = 75 W) or other mismatch conditions (open, short, wrong LPF) the transistors will over heat.

My Windom antenna (not ideal mounted) has matched on the 80 m band a SWR of 1:2.4 to 1:2.7, so it just fits. On the higher bands the SWR is matched below 1:1.7.

That were best case calculations. But if the torque at one MOSFET is too low, or some dirt under a MOSFET could lower quickly the maximum heat dissipation.

Lets calculate for a popular Sil-Pad:

• Total thermal resistance: 1.7 + 1.6 + 1.5 = 4,8 °C/W

• 150 °C - 25 °C = 125 °C / 4.8 = 26.1 W, dissipation max. x 2 = 52.2 W

Here the SWR must be lower than 1:2.1 (50 + 9 = 59 W) on the 3.5 MHz band.

Thermal measurements

I measures with an IR thermometer the temperatures, heat sink vertical:

Calculation:
7 MHz Pout: 70 W, 24 V, 4.3 A, 103 W -> heat 33 W
heat sink: 1.6 °C/W * 33 W = 53 °C
Alu-Pad:   0.4 °C/W * 33 W = 13 °C
Room:                        26 °C
Sum:                         92 °C

max. temperature measured:   93 °C after 5 minutes settled
IRF530:    1.7 °C/W * 33 W = 56 °C
Sum:                        149 °C
max. junction temperature:  175 °C
Transformer T2: 72 °C
Transformer T3: 62 °C

So, it looks like, that theory and reality does match closely.

The supply current dropped down to 4.2 A after 5 minutes, but the output power looked constant. It looks like, that the thermal compensation with the 2 diodes work.

Harmonics

Law in USA: For amateur use, FCC §97.307(d) requires that harmonics be suppressed at least 43 dB for operation below 30 MHz.

Law in Germany: Erforderliche Dämpfung unerwünschte Aussendungen gegenüber der maximalen PEP des Senders:

• 1,7 - 35 MHz: 40 dB

In a well designed solid-state push-pull short wave amplifier, the second harmonic is 30 to 40 dB below the fundamental (first harmonic) if the balance is good, but the third harmonic is only down by about 13 dB.

It can be seen, that the most dominant third harmonic is in all cases suppressed more than 40 dB (including Low Pass Filter).

The Harmonics of the power amplifier was measured with a Hantek DSO with FFT function, at a supply voltage of 13.8 V and 24 V.

As you can see, the harmonic level is usually lower with a supply voltage of 24 V.

When you add to the harmonic dB values below carrier the dB values attenuation of the Low Pass Filter, in the case of the 24 V supply voltage all harmonics after the filter are all more than -40 dB below carrier level.

The most critical harmonic is the 2nd harmonic in the 15 m band. Unfortunately the 2nd harmonic is very strong in this case, and the 10 m filter damping is just 5 db at this frequency. With some tuning at the bias current (10 - 15 mA) it was possible (25 V supply) to just archive 40 dB damping. Another possibility is to use instead the 20 m band filter, but the drawback is to loose about 2.5 dB of output power (42 -> 24 W).

In case of the 13.8 V supply voltage, only one harmonic (20 m, f2) is at a level of -37 dB (-25 dB + -12 dB) below carrier level.

• 80 m band:

.

• 40 m band:

.

• 20 m band:

.

• 15 m band:

.

• 10 m band:

.

Input impedance matching

In order to achieve a good RF input matching the resistors R8 and R14 are provided in the schematic.

Additional I inserted a Pi type attenuator with 2.6 dB (330R, 15R, 330 R, all 2 W).

• Transformer T2: 4 : 1 turns

The Input SWR was measured from 1 to 30 MHz with my Vector Analyzer Yana. The original values for R8 and R14 are 20 Ohm each. For optimizing the values are changed to 15 Ohm and 10 Ohm.

While the measurement, the PA was powered with 24 VDC, and the output connected to a 50 Ohm dummy load.

Yes, the SWR is better below 12 MHz , the lower the resistance was, but the Output Power in the upper 2 bands decreased. The SWR over the short wave band is shown in the following screen shot's.

Therefore I stood with the 20 Ohm value, as a good compromise.

The peaks in the curves are RFI (Radio Frequency Interference) from the transceiver. I had to push the PTT switch in order to get a curve.

Now I have a one port Vector Analyzer FA-VA5 together with the very good software VNWA (from DG8SAQ, version 36.7.6) and measured the PA Input Impedance Z, see the screenshot on the right.

The software VNWA is master calibrated, you can see it in the lower left corner of the diagram with letters MC.

You can see that the Z curve (blue) is about the mirror curve to the SWR value in the above YANA screenshots.

The smith chart or Smith-Diagramm (red curve) shows that below about 12 MHz the impedance is inductive, above capacitive.

.

.

.

As already shown, the input impedance of the PA is not perfect 50 Ohm. Therefore I measured with a tuned 5 W display of the power monitor (second column) of my tranceiver what it means in case of a 50 Ohm dummy load (third column), instead of the Power Amplifier.

The last column shows the voltage (Vpp) measured at the input of the Power Amplifier, in order to have a reference value. Unfortunately my DSO has a non linearity of about +1.5 dB @30 MHz (+19%) in this frequency range (68 Vpp -> 57 Vpp).

Output power

The following modification or adjustments were made to optimize the output power, but with all harmonics are 40 dB below carrier.

• Supply voltage set to 25 V in order to stay below 5 A.

  • It improves the linearity, less harmonic level.

  • It allows to use a lab type power supply with current limiting (30 V, 5 A, about 40 - 42 EUR).

• T3 secondary winding reduced from 4 winding to 3 windings.

• The quiescence current for each MOSFET adjusted to about 30 mA.

• The Input Power was adjusted with the Power Meter from the tranceiver.

The Input Voltage right from the slash is a corrected value because of DSO overshoot (0.8 dB=1.1, 1.5 dB=1.19).

I tried also 30 VDC for the supply voltage. I started measuring at the upper bands - OK. But at 3.6 MHz one MOSFET broke. To find out which one broke, you can measure Gate to Drain resistance. If you measure a low resistance, in my case 25 Ohm, the MOSFET is broken. Normally this resistance is greater 1 MOhm.

If the MOSFET dies because of overheating, you can see at the back side deformations in the tinned surface.

Ferrite

I needed a long time to figure out the Ferrite material type. In the archive SKU303197 RF-AMP-2078_45WV306.rar (download "manual" from Banggood.com, Product ID: 1018834) I found in the file _RF-AMP-2078器件清单V306.xls the material type NXO-100. In the file 20081223015952.pdf from company EECTech.biz I found the dimensions for the Multi-Hole Cores (type III).

Finally in file www.indo-ware.com_toroid_nikel-seng.pdf I found the material data for NXO-100:

• Initial permeability μi: 100 ± 20
• Relative loss factor tgδ/μ (* 10**6): 63(1MHz), 200(15MHz)
• Temperature coefficient αμ * 10**6/°C: 400(+20 °C ~ +60 °C)
• Resistivity ρ (Ωcm): 1 * 10**5
• Saturation flux density Bs(mT): 330
• Remanent flux density Br(mT): 220
• Coercivity Hc(A/m): 320
• Curie temperature Tc (°C): 350
• Applied frequency (MHz): 15
• A-L value: about 130 [nH/t**2] @ 15 MHz (measured)

Low pass filter

The 5-Band Low pass filter uses 4 relays to switch the filters, see my description.

It is connected to the PA via a 12 cm long coax cable RG316 with SMA plug.

Fortunately my old transceiver F850 from Sugiyama Electric does have a connector J7 at the back side, which delivers all needed control signals:

                 PA plug (9 pin)
                    cable color
J3 Cinch ALC In  9  wht  Automatic Level Control, not used
J5 Cinch Key          for CW keyer
J7 ACC -              for Power Amplifier
  Pin  Signal    PA
  1    1.8 MHz
  2    3.5 MHz   2  red
  3    7 MHz     3  pnk
  4    14 MHz    4  yel
  5    21 MHz    5  grn
  6    28 MHz    6  blu
  7    50 MHz    
  8    144 MHz   
  9    13.5 V    7  vio
 10    NC
 11    TX-GND    1  brn  PTT OUT
 12    0 V       8  gry
Pin's 1-6 have a diode 1N4148 in series, cathode to TX, 100 nF at cathode to 0 V.

The 9 pin connector, the diodes and the capacitors (SMD 0805) are soldered on a piece of vero board.

Band fault detection

In order to protect the Power Amplifier against Overvoltage and Overheating if a wrong short wave band was selected, or the RF output is not terminated, or the SWR of the antenna is very poor, two protection circuits could be added. A big help in the case of protection is the current limiting of the power supply. If reality shows, that this is not enough, it would be added later.

Housing

By luck I got a Compaq 133518 PSU, size 16 x 11.5 x 20 cm. You can get it via Ebay for about 20 EUR. Once you remove the electronic boards, you have everything you need for housing this power amplifier.

1. The housing is made from metal
2. All components fit into the housing
3. The ventilation holes are already provided
4. At the back side a 80 - 100 mm fan can be installed
5. When you cover the side walls with paper, the air flow should be good
6. A thermostat module can be easily mounted

7. The 230 VAC plug and filter can be used for the 24 VDC supply. Take care that the wire color blue always is 0 V.

8. The DB9 plug with wires can be used for PTT and SW band switching

9. The heat sink can be mounted with 2 screws to the bottom

10. The heat sink temperature display and the SW band LED's can be seen through the front plate

I recycled the 3-pin 230 VAC connector for the 24 VDC supply.

How to do (see also the picture above):

1. Unsolder the white 3-pin plug from the original PSU board.
2. Cut one pin, bent the other two apart, so they fit on the PA VCC solder holes.
3. Solder the 3-pin plug to the PA board.

4. Rearrange the 2 wire pins blue and brown that blue is on 0 V and brown on PA VCC.

   • To pull out the female pins out of the plug body, use either a fitting pipe to bent back the lock tongue of the pin, or use a small screw driver to do so.

24 VDC Supply

For the 24 VDC supply I choosed an adjustable laboratory power supply 30 V @ 5 A with current limiting. For 42 EUR (including shipping) bought via Ebay it is cheaper than to build something home made. It has a switching regulator and a temperature controlled fan.

• The current display resolution is 1 mA.

• The voltage display resolution is 0.1 V.

Especially the current limiting is very important to protect the PA power MOSFET's.

The connecting cable to the PA housing is an old 230 VAC cable. I cut off the 230 VAC plug and mounted 2 banana plugs.

.

.

.

.

Wiring for SE850T

My old Sugiyama Transceiver SE850T supplies at the backside a 12 pin J7 (ACC) connector for the control of a power amplifier. There is the following connection:

# Signals from Transceiver to PA
SE850T J7 -> 9-wire cable DB9P -> DB9S wires 9 pin connector -> Band LPF relays, PTT

# Signal protection, pins 1 - 6
9 pin connector -+-|<|-- 1N4148 -- 
                 |
               100 nF ceramic
                 |
           0 V  -+-

# wire colors and routing
                      PA plug(9 pin), DB9S color 
                cable color
J3 Cinch ALC In wht   wht  9          blk9  Automatic Level Control, not used
J5 Cinch Key                                for CW keyer
J7 ACC -                                    for Power Amplifier control
  Pin  Signal                           
  1    1.8 MHz
  2    3.5 MHz  red   red  2          pnk
  3    7 MHz    pnk   pnk  3          red3
  4    14 MHz   yel   yel  4          wht
  5    21 MHz   grn   grn  5          blk5
  6    28 MHz   blu   blu  6          vio
  7    50 MHz    
  8    144 MHz   
  9    13.5 V   vio   vio  7          red7
 10    NC
 11    TX-GND   brn   blk  1          grn  PTT OUT
 12    0 V      gry   gry  8          blu

Thermostat

A thermostat with 3 digit LED temperature display was bought at https://banggood.com with ID number 1006449 and title: W1209-DC-12V-50-to-110-Temperature-Control-Switch-Thermostat, to control the fan.

The temperature sensor (diameter 4.9 mm) was seated in a 5 mm hole in the heat sink, close to the MOSFET's, see the pictures below. The hole got some heat-conductive paste for better conductivity.

It is useful to secure the temperature sensor against falling out of the heat sink. I did it with a piece of foam 10 x 10 x 60 mm. The distance from the housing wall to the sensor (cable end) is about 50 mm. One end of the foam piece got a 5 mm long cutout for the sensor, so 5 mm are left for pressure to keep the foam in place. The wall side of the foam piece should also secured in place with some glue.

Specifications:

• Temperature Control Range: -50 ~ 110 C
• Resolution at -9.9 to 99.9: 0.1 C
• Resolution at all other temperatures: 1 C
• Measurement Accuracy: 0.1 C
• Control Accuracy: 0.1 C
• Refresh Rate: 0.5 Seconds
• Input Power (DC): 12V
• Measuring Inputs: NTC (10K 0.5%)
• Waterproof Sensor: 0.5M
• Output: 1 Channel Relay Output, Capacity: 10A
• Power Consumption

  • Static Current: <= 35mA

  • Current: <= 65 mA

• Environmental Requirements
  • Temperature: -10 ~ 60 C
  • Humidity: 20-85%
• Dimensions: 48mm x 40mm x 14mm
• Settings Chart
  • Long press the “SET” button to activate the menu.
• Code Description Range Default Value
  • P0 Heat C/H C
  • P1 Backlash Set 0.1-15 2
  • P2 Upper Limit 110 110
  • P3 Lower Limit -50 -50
  • P4 Correction -7.0 ~ 7.0 0
  • P5 Delay Start Time 0-10 mins 0
  • P6 High Temperature Alarm 0-110 OFF
• Long pressing +- will reset all values to their default

The manual can be found here.

• Temperature Set: A short press of the SET button will let the LED display flash. With the buttons + and - the new temperature to switch ON the fan can be adjusted. After some seconds of inactivity, the display will switch back to measure mode.

.

DSO Probe connection

The signals are measured with the probe tip and ground ring connected directely to the circuit.

See the picture on the right.

.

.

.

.

.

.

MOSFET Source Leg Length

Just by a trial I found out that the MOSFET source leg length is very important for the function (amplification, ringing) of the power amplifier. Just 10 mm length difference has a significant impact.

I first tried 1 - 3 windings of a coil 3 mm diameter in series with the source to ground, what WA2EBY suggested for damping the ringing. The amplification factor was very low. Then I reduced the coils to 2 and 1 winding, without much improvement. For an easier wiring I pulled a 6 mm long red shrink tube over the thin part of the MOSFET source leg. So I could isolate the source leg and connect it on the upper side of the PCB.

Next the source leg was connected to a 10 mm wire to ground, in total 10 + 13 = 23 mm from the MOSFET body to ground. That brought a satisfactory performance of the PA.

Last step was to connect the MOSFET source direct to ground (13 mm length), see the picture on the right. In the next chapter Waveforms you will see, that the ringing amplitude increases, but the harmonics amplitude (#Ouput_Spectrum) kept within the limit. So I stopped with further modifications.

Waveforms

Sometimes in the build process it is helpful to have some reference waveforms at hand in order to check if you are on the right track.

• Shown are the following signals:

  • Yellow trace: PA Output (Filter) connected to a 50 Ohm dummy load

  • Blue trace: MOSFET drain

• The signals are measured with the probe tip and ground ring connected directely to the circuit (No ground sleeve).

Watch the time resolution in the upper right corner of each diagram (4 - 20 ns/division).

.

.

.

.

Ouput Spectrum

The output spectrum was measured with a 50 Ohm Dummy Load (100 W).

Watch the time resolution in the upper right corner of each diagram (800 - 2000 ns/division).

.

.

.

.

Links

1. PA build instructions

2. PA windings instructions

3. PA documentation, original schematic

4. PA0FRI - PA schematic

5. PA schematic

6. Hackaday - 45W RF SSB CW AMP (EBAY) DIY build

7. Vpp - dBm Calculator

8. dBm to Volt table

9. dM - factor Calculator

10. circuit analysis program LTspice, overview

11. Elsie, Win32 program for electrical filter design and network analysis

12. AL value explanation

up to top

Contact Email, please enter your Email address

The entered Email address will not be published, or given away.

List of pages in this category:

-- RudolfReuter 2018-07-13 07:17:06

---

Go back to CategoryAmateurRadio or FrontPage ; KontaktEmail (ContactEmail)