A 120V 40mA Battery Operated Power Supply

A Spec Sheet for a Homebrew High Voltage Battery Power Supply

This PSU is intended as an anode power supply for noise tube measurements. It should have the following features:

• 120V, 40mA max.
• Noise in the ballpark of  1mVpp in a bandwidth of 0-10MHz.
• No measurable spikes or ripple
• Warning LEDs for low battery and overload. Short circuit protection.
• Low 1/f noise.
• Works with standard 9V block batteries. It takes 22 of them, however.

As you can see, the requirements are less stringent than with the 15V100mA unit.

My recipies to get things done were:

• All linear regulator components built with low noise parts. No AC allowed anywhere, in fact.
• Metal film or wirewound resistors for increased stability and less noise
• “Stalinistic” shielding, a “full metal jacket” enclosure and a coaxial output path. No banana jacks allowed.
• No meters (digital ones have clocked ICs, analog ones need large cutouts damaging shielding quality). Output is 1 BNC, with (internally) selectable polarity.

Below you can see a circuit diagram of the battery powered supply:

Click here for a higher resolution picture

At the top we can see the HV regulator using the usual TL431 based design, here with a bipolar pass transistor and an optocoupler-based overcurrent sensor. This part of the circuit is the only one run off the HV battery. The rest of the circuit contains two voltage watchdogs; one for the regulator voltage and one for the HV part. Both drive a duo LED to indicate the status; The overcurrent condition is also displayed by a LED.

The design itself could be used with higher voltages (and less current) if needed. The end of the line is probably ca. 400V@20mA, because of the pass transistor.

The regulator is short circuit proof, but dissipation in case of a short circuit could be 16W, so an appropriate cooler is a must. Under maximum load conditions and fresh batteries the pass transistor dissipation is (200-120)*0.04 = 3.2W, going down to less than a Watt when battery voltage approaches the minimum allowed (130V).

… and the PCB:

Battery Lifetime Considerations

My design uses two battery packs. One is for the anode voltage, and the other one is for the regulator and all the LEDs, …

The anode battery pack I plan to use here is two stacks of 11 x Alkaline 9V Block batteries. When fresh, the voltage of one stack is ca. 100V or even a little bit more, so we end up with 200V in total. If we discharge the cells down to 6V, we still have 133V before the regulator, so a 120V fixed output voltage should be OK and give a useful battery lifetime.

The regulator battery consist of 10 x AA cells and supplies 8 to 15V at 30-45mA max. That should suffice for several 10 hours.

Lifetime is mainly a function of load current. With full 40mA anode output current and a fresh pack of 22 batteries we can expect only a little more than 1 hour of operation. 40mA is a bit much for a noise tube anyway, so if we calculate with the usual anode currents like 5mA (2D3B, 7dB ENR) or 20mA (2D2S, 13dB ENR) the battery will last a few hours.

Click here to see a sample AA battery datasheet from Duracell …

Click here to see a sample 9V Block Battery Datasheet from Duracell …

The first Prototype

The final “product” looks like this:

 

… and the interior looks like this:

Click here for a Higher Resolution Picture …

 

On top sits the hefty battery pack (22 x 9V Block, 10 x AA). The high voltage sign is absolutely justified here, because the idle voltage with fresh batteries can be up to 210V, so beware ! One of the two battery “drawers” is shown below:

The drawer is a homemade design done by a 3D printing service using some salvaged parts from 9V battery holders:

Back to the other interiors. The regulator is located at the bottom, with the double voltage watchdog to the left and a classic “Changpuak” regulator to the right, this time with a bipolar high voltage transistor darlington combination of a BU208A (SOAR specified) and an 2N3439 as a driver.

Results

On the Tek one can see ca. 1.5mV RMS of something that looks like random noise (120V, 10kOhm load, 15uF decoupling cap, 50Ohm scope inputs).

The spectrum is shown below (internal preamp on, 100Hz BW, Log Average Detector )

There is a noise peak at -74dBm at 429kHz that I have to track down, the rest looks from 1MHz on is below 110dBm. Maybe a lowpass filter could eliminiate the peak ? Lets try !

I added a lowpass with a 330Ohms resistor in series and a 22uF/250V cap in parallel (fc = 137Hz), and it improved a lot:

The peak is now at -125dBm, the rest looks just beautiful. A cross-check on the Tek now reads ca. 250uV RMS total PARD. Work done !