50V Pump

Mundaka LNA 50V Voltage multiplier

I need max. 50V for the photodiode bias. I could use a DC/DC stepup converter, but the chips are expensive and difficult to obtain worldwide (impression from the Distrelec website). I could also use a ferrite coil or transformer but that requires manual winding or reduces the availability or both. Therefore I took a diode-capacitor charge pump.

To require less stages I decided for a bridge configuration. I need something that can oscillate and produce voltage which should reach as close to the rails as possible and not have too much internal resistance. I checked several opamps - ua741, 324, TL062, TL072, TL082 and lot of others (Distrelec). CA3410 would be suitable but it's relatively expensive and doesn't come in pairs. Therefore the winner is TL082 because it has the least transistors between the output and the rails and also doesn't have any extra bells and whistles like lownoise or low power consumption.

I also thought about 555/556 but that has an horribly weak output.

I started with the Greinacher multiplier I knew from a TV. The disadvantage here is the capacitors are in series which reduces their capacity, increases ripple and makes the analysis more difficult. Conceptually I understand each stage as two capacitors, one works as a piston and the other for storage of voltage. The piston sucks in electricity through the first diode and spits through the second. The voltage increase on the stage is equal to peak to peak voltage of the source minus twice diode drop.

The opamp oscillator is taken from Opamp Oscillator Circuits, answer 4a. The virtual ground network was optimized to save one resistor.

I tried an opamp oscillator with 3x27k in the resistor feedback network and 120k in the capacitor feedback and 10nF capacitor running at about 600 Hz with ICL7621. Unfortunately this chip is extremely slow and has a weak output. 6-diode stage with SB160 and 6 ceramic 220nF capacitors gave 70V without load, 44mA into short but with 72uA draw it dropped already to 48V - unsatisfactory!

Therefore I made an oscillator with 74HCT14 and two push-pull emitter followers preceded by switches converting 5V pulse into 12V running at 100kHz. That works fine, giving enough power to generate stabilized voltage on a 47V zener diode and still giving 60V. The resistor values were selected based on what was available, rough estimates of currents involved, and adjustment on running circuit.

Pump interference

The pump generates about 200mVpp RF noise on the 5V line. It shows as 100MHz ringing with the capacitive probe (PAE Radiotechnika). It was discovered when testing the frontend. It shows as 100MHz rings (about 5-10 waves) repeating with 260kHz frequency.

The cause is the inductance (3nH) of the 100nF capacitor. I will replace this capacitor with SMD and put a 47nH coil in series. The resonance of these two is at 2.3MHz. Critical series damping is 1.35Ohm, the 47nH coil has less than 0.1 Ohm (according to ohmmeter). Therefore I add 1.5Ohm in series to prevent resonance between the L and C. The current draw of the HCT is 820uA, that makes a laughable drop of 1.23mV. However adding this 47uH/1.5Ohm in series didn't help at all. There was still 200mV of noise on the regulator.

Actually the same noise is visible if the probe tip is put on the ground next to the regulator, suggesting the noise travels through air and induces into the probe ground wire.

The noise appears even when the probe is not connected to the circuit at all and the ground wire is shorted with the tip. It is weak from the empty side of the tin or on the part of the box where the SMPS is not. When moving above the circuit it's the strongest. If the ground short is disconnected the rings are not visible, however when the tip is approached to the circuit, the signals from the circuit are clearly visible on the scope screen! (20mV/div on the probe tip).

Disconnecting between the HC and the bases drastically reduces the interference: tip to metal, clip to metal: 30mVpp, air propagation on closed loop also drastically.

Bases back, cascade disconnected from the push pull emitter followers: 40mV on the tin, 160mV on air loop. Air loop oscillations: 150MHz. Seems the sharp edges spread persistently into the whole converter circuit.

I wil try a transistor multivibrator that should be slower and doesn't contain any totem pole which creates spikes. The collector resistors will be 6.8k and saturation 1:10. Minimum gain is 70, therefore the base resistors should be 6k8*7=47k

The output level is only 9.8V with 12V supply. The low portion of the waveform is flat, the upper portion is an exponencial. The period is 160us.

Changing the resistors from 47k to 270k improved the waveform to practically square the upper voltage swing stays still about only 11V and the period decreased to 760us. For 10us period the C would have to drop to 29pF which is too small considering the be parasitic.

I decided to keep the repeater transistors. Decrease the base resistors to 39k. That gives 59V. Decrease of capacitors to 330pF gives 60V (loaded with zener diode).

Now it runs at 50kHz and the output of the push-pull followers has a notch after the leading edge.