HT-LT Inverters for Valve Portables

[kalee20]

There have been a few threads on this topic, the most relevant being https://www.vintage-radio.net/forum/...light=inverter and https://www.vintage-radio.net/forum/...light=inverter. I too have been experimenting with inverters to provide power from batteries for portable valve radios, and it’s definitely not the easiest project to tackle! Here are my experiences, after building several single-ended and push-pull converters.

The radio which I chose as a test bed is an Ever Ready Model K. This is a LW/MW all-dry 4 valve set, DK91-DF91-DAF91-DL92. Power consumption according to the service sheet is 90V @ 12.5mA plus 1.4V @ 250mA – the valve filaments being wired in parallel, with one terminal connected to chassis. Grid bias (10V) for the output valve is provided by a resistor in the HT –ve lead, which is a common arrangement. The power is switched on and off by two poles on the waveband switch, which separately switch LT and HT. Reception is via a self-contained frame aerial.

I gave the radio a good overhaul and realignment. Then, I set about a few small modifications. First, I connected a 0.1uF multilayer ceramic capacitor from each valve LT+ to chassis. The LT+ lead, which is daisy-chained from valve to valve, I disconnected and put a ferrite bead in each loop. The HT decoupling capacitor (8uF electrolytic) I replaced with a 10uF 100V polyester.

It seemed really inelegant to waste 10V at 12.5mA of hard-earned HT across the bias resistor, so I thought I would generate an auxiliary negative bias rail from my inverter. Then the inverter would only have to provide 80V – or, since the valves are characterised at 90V anyway, I could generate 90V and apply all of this to run the radio, thus getting a slightly higher performance. So, I removed the bias resistor. The DL92 originally had a 2.2megohm stopper resistor directly in series with the grid pin, but as it would soon be operating with fixed bias, I reduced this to 1 megohm (the 4.7megohm grid leak I left at the same value). Finally, I rewired the on/off switch so that HT and LT –ve go direct to chassis – the two spare poles on the switch now being wired in parallel so they could switch the battery supply to the inverter.

I wanted to design the inverter to operate from 6V, which would allow use of a 6.3V lead-acid accumulator, or five Nicad cells. At this voltage, it seemed grossly inefficient to supply the radio LT via a dropping resistor, so I decided on a 1.4V LT output. I chose a flyback converter operating with complete energy transfer, frequency about 20kHz. This is inherently short-circuit proof. The output voltage for this type of circuit is highly dependent on load, and as I have not included any means to regulate the output voltage, it would not be suitable for a radio with a Class B output. However, the Model K is Class A and thus puts a nearly constant load on its HT supply – the only variation being the small decrease in current when a strong station is tuned in due to AVC action. I have included a 100V Zener diode across the HT output, to absorb power if the load becomes disconnected.

The circuit is as shown in the attached pdf file. I designed it with two goals in mind: minimum RFI and maximum efficiency. The flyback circuit helps here, because the current in the HT rectifier diode reduces smoothly to zero each cycle and there is no big reverse recovery current spike, as could occur with other converter types. The switching transistor is a BD131 which has a very low saturation voltage (30mV) yet switches well at 20kHz. Base drive is provided, not via a wasteful current-limiting resistor, but from a current transformer in the collector lead, thus giving a base current proportional to collector current with minimum power loss. Transistor ‘on’ time is determined, not by saturation in the power transformer (which would give a nasty current spike), but by a programmable unijunction transistor timing circuit. If anybody is interested, I will put up a circuit description – but this post is long enough for now!

I added LC filtering in all input and output rails (except the negative bias output, which supplies zero current and thus has RC filtering). I built the circuit on a PCB and laid out the tracking to minimise any common impedances. The filter chokes are all toroidal, single layer wound, to minimise end-end capacitance. All this paid off, as even with a 20MHz ‘scope, there is less than 0.25mV of ripple and spikes on any input or output lead, measured with a 20MHz ‘scope.

To reduce radiated RFI, I originally intended to mount the PCB in a diecast box, but having done some EMC compliance testing in the ‘day job’, I found a diecast box not quite as good as I had hoped. Accordingly, I made a copper box out of 0.25mm copper foil, folded around a wooden former, and soldered at the corner seams. The lid is another piece of 0.25mm sheet, with small lips to form a shallow tray. This fits fairly tightly, but I further held it to the box with self-adhesive copper tape (which has conductive adhesive). I debated whether to use a toroidal power transformer (using low-permeability MPP material), in the interests of lower external field, but for ease of winding, plus lower core losses, settled on a ferrite RM core. I added a copper shield surrounding the transformer, to reduce external field.

The results of this are: Efficiency is fractionally under 80%. Of course, as the radio originally wasted 11% of its HT power, I was fairly satisfied with this. I will some day do an analysis of where power is lost. The radio works well, with no hint of interference… until the copper box is put in position! The Model K cabinet layout is such that the batteries originally sat within the frame aerial, and even with the copper screening box, there is enough external field to give slight heterodyne squeals when tuning across the band. The fact that it is quiet when the inverter is even just 6” away indicates that the interference is not entering the radio on the HT and LT rails, so filtering is adequate. LW is worse, and I put this down to the copper box being less effective at these frequencies (it’s worth noting that the penetration depth at 200kHz is 0.16mm and at 1MHz is 0.075mm). So, maybe the box should be made from 0.5mm sheet, despite being harder to bend. As it is, performance would no doubt be OK for an attache case radio with frame aerial in the lid.

Lessons learned are that the RFI is definitely the hardest nut to crack. A much higher frequency would help, because the harmonics will be further apart so giving fewer whistles when tuning across the dial. The next version will be regulated, probably by sensing the LT rail and leaving the HT to take care of itself. And, I will add a shut-down circuit to protect against over-discharge of the 6V accumulator.

[igranic]

An excellent and well planned project.

The positioning of the inverter within the frame aerial will certainly tend to encourage inductive coupling of any stray 21KHz from the inverter transformers. Could some of the harmonics arise from non-linearity in the mixer/oscillator stage. If so, adding a 21Khz trap to the control grid of the mixer/oscillator may help reduce the heterodyne whistles.

Edward

[Robert Darwent]

Originally Posted by igranic

An excellent and well planned project.

I have to agree with Edward here, a really first class effort!

Personally I'm less into the theoretical aspects of circuit design and action and more into the practical side of building and constructing them. But I found your circuit description fairly easy to follow and most interesting reading. Thanks for sharing with us your efforts thus far!

Regards

[kalee20]

Thanks, guys, for kind comments!

Edward, you've postulated a mechanism for interference that I'd not thought of. Although the set-up is useable (I've got it providing background music right now), the slight background whistles are detectable and I want to tame them! I am at the moment, fairly sure that the problem is the magnetic field from the inverter transformer coupling into the frame aerial, and as the field is anything but a sinewave, it will have harmonics and I reckon these are extending well into the broadcast bands and upsetting things. But, your theory is that the 21kHz magnetic field itself is being picked up by the frame aerial, and the harmonics generated within the DK91.

The field at 21kHz will be stronger than any of its harmonics, and as it's of lower frequency it will be attenuated less by the copper box (I'm thinking in print here). So, your theory is plausible. I'll have a look later this week with a 'scope on the DK91 signal grid, for any signal at 21kHz. And, I'll try adding a resonant trap (probably series resonant between grid and chassis). It's a simple job, and will at least prove or disprove the theory.

Thanks for this! I'll put up the findings!