moria.de Radio astronomy: Satellite TV LNB power supply

---

Satellite TV LNB power supply

A satellite TV LNB gets the power through the coax cable by using a bias tee that splits/combines the RF signal and the power. Bias tee modules are available in high quality for little money, but a power supply is still needed.

What's required?

The LNB polarization is controlled through the power supply voltage:

13-14V vertical
15V decision level
17-18V horizontal

I have been unable to find a normative reference on the exact levels. LNBs usually use a LDO and more voltage equals more heat, hence I want to stay at the lower end and avoid an excessively long coax cable.

A modulated 22 kHz signal with 0.65 Vpp ± 0.25 V switches from the low (10.70–11.70 GHz mixed with 9.75 GHz to 950–1950 MHz) to the high band (11.70–12.75 GHz mixed with 10.6 GHz to 1100–2150 MHz).

For that reason, the power supply needs a digital control for the voltage and the 22 kHz signal.

The design

Every satellite TV receiver contains such a power supply, so of course there are special ICs. They can be obtained at Ebay at affordable prices, but they are hard to solder due to their big thermal pad at the bottom and most of them still require an external 22 kHz generator or a high supply voltage, so they don't save quite as many parts as it may look like at first. In addition, life cycles of such specialised parts are often shorter than hobbyist projects.

Converting +5 V, a popular supply voltage for many components, to the needed levels with a cheap LM2577 (or the chinese clone XL6009) module is easy. Lacking experience with DCDC converters, it is tempting to switch the feedback voltage divider of the step up converter to control the voltage, but trying that with a XL6009 without load caused a temporary output voltage > 50V. Usually a sharp decrease in the feedback voltage would be a sudden load and the response does make sense, as there is no load. Plugging in a LNB in that state would briefly expose it to overvoltage. Such DCDC-converters are not made for fast voltage switching, like needed for a Dicke-switch.

The datasheets of some LNB power supply ICs show they use a step up converter to feed a LDO, which is very fast, improves noise and avoids the overvoltage situation. That suggests a discrete circuit with a step up converter running at fixed voltage and an adjustable linear voltage regulator.

That's exactly what George Smart built for his LNB power supply with integrated bias tee. I mostly copied his excellent circuit.

Built at home

The circuit contains a NE555 for the 22 kHz signal, which is coupled to the feedback voltage divider for the linear regulator. A transistor changes the voltage divider by optionally bridging one resistor:

The LM317 may have to drop 20 V − 13 V = 7 V at 200 mA (modern LNBs don't use more) and I prefer not too much heat:

$$\text"7 V" · \text"200 mA" = \text"1.4 W"$$ $$\text"20 K over ambient" / \text"1.4 W" = \text"14.3 K/W"$$

A heatsink around 14 K/W should be fine, which even most tiny heatsinks satisfy. In the real world, I measured 18 K over ambient at 110 mA for a 15 K/W heatsink.

This configuration of the NE555 delivers an almost symmetrical signal of 22 kHz, no need to use diodes:

$$t_1 = ln(2) · ( \text"330 Ohm" + \text"6.8 kOhm" ) · \text"4.7 nF" = \text"23.2 µs"$$ $$t_2 = ln(2) · \text"6.8 kOhm" · \text"4.7 nF" = \text"22.2 µs"$$ $$f = 1 / { ln(2) · ( \text"330 Ohm" + 2 · \text"6.8 kOhm" ) · \text"4.7 nF"} = \text"22 kHz"$$

My NE555 generates 3.8 V at the output, which is only lightly loaded. The datasheet specifies typically 3.3 V for 100 mA load, but does not specify lower loads. The voltage divider used by George Smart delivers 3.8 V · 100 Ohm/1100 Ohm = 0.38 V, but measurement shows 0.33 V for my circuit, which is outside the spec. I use a 3.8 V · 390 Ohm/2200 Ohm = 0.65 V, but measurement shows 0.5 V for my circuit. That is theoretically and practically within the spec. I wonder if the lower voltage is a result of the divider load.

The datasheet for the XL6009 does not mention undervoltage protection, which the LM2577 has. If the power supply drops below 3.6 V, the indeed missing undervoltage protection causes it to generate 58 V. Using a 3.9 V Zenerdiode with a resistor shuts down the XL6009 at 3.8 V or below. Warning: There are many fake LM2577, which are really XL6009.

Without L1 you get HF noise like this (13 V AC coupled):

First I tried a 0.6 mH inductor with a ring core. It suppressed the noise, but caused trouble with the load regulation. A single wire through a ferrite toroid works better:

I tried two different ferrites, the smaller one being 10x20 mm, and both work equally well. With 22 kHz there is only minimal ringing (13 V AC coupled):

The voltage switching speed at 1 kHz looks fine: