BPSK Modulator details]

[Andre' Kesteloot]

James Moritz wrote:

> Dear Mike, LF group,
>
> I am in the process of putting together some info about my BPSK modulator for the Decca class D PA - it will take a little time to collate all the details from my meticulously cross-referenced notebook (ahem..), but here is some general information about the method I am using.
>
> The BPSK signal can be generated by exclusive-ORing an AF or RF carrier with the binary keying data. this produces an output whose phase is unchanged when the data value is zero, but shifted by 180 degrees (ie. inverted) when the data is 1. This signal can be amplified by any linear or non-linear amplifier, and is a perfectly OK signal as far as demodulating goes, but it also generates sidebands analogous to key-clicks from a CW TX. In fact, for the same transmit PEP, the clicks generated by BPSK will be 6dB stronger than an on-off CW signal. So obviously, if you are trying to generate 1W ERP it isn't really on, although you can get away with it for a QRP transmitter.
>
> The remedy is similar to that used for CW: the output is gently reduced to zero before the phase change occurs, then gently increased to it's full value again after, in other words amplitude modulation is applied to the BPSK signal, as well as phase modulation. It turns out the ideal amplitude envelope is like a full- wave rectified sine wave - see the description of PSK31 in the LF handbook, which is basically the same.
>
> There are two main ways of generating this signal in practice - the most popular is to synthesise the waveform directly using a DSP algorithm and a digital to analogue converter, eg. soundcard plus software. However, most DSP DACs are limited to audio frequency outputs, so the signal must be mixed up to RF, and also amplified in a linear PA so the modulation envelope does not get distorted. However, If you have an HF rig with LF output, and a QRO linear PA, this is the easiest route to a BPSK capability.
>
> I didn't have, so...
>
> The other method is to apply the phase modulated signal direct to the PA, and simultaneously amplitude modulate the PA to get the required envelope. A class D PA is well suited to this due to the linear relation between supply voltage and RF output amplitude - what is required is to replace the normal DC supply with the full- wave rectified sine waveform which follows the phase transitions.
>
> The advantages of this are:
> -Efficiency can be relatively high compared to a linear
> -No frequency conversion means few spurious outputs, and minimal frequency errors.
> -Simplified digital system required.
>
> In my present system, the 136kHz carrier is generated by a synthesiser, and phase keyed using a simple ex-or gate. The envelope modulating waveform is generated from the phase keying signal. The envelope-shaping part is built using 4000 series logic and some analogue bits. It is a prime candidate for implementation using a PIC with a D/A, but I haven't got a programmer, hence the hard wired circuit. It is synchronously clocked at 256 x the bit rate by a 555 timer. The incoming phase signal is appied to a transition detector. When a transition occurs, an 8-bit counter is reset, which clocks through all it's states and then stops. The 4 MSBs are used as the addresses for a 16 input analogue MUX. The MUX inputs are connected to taps on a sine weighted potential divider fed with a DC reference voltage, producing a stepped approximation to the modulation envelope, which is low pass filtered to smooth it out. This then goes to the power stage of the modulator. The "middle" count of the
> counter is decoded and used to clock the phase data through a D flip-flop to get a delayed phase output, which is what is actually fed into the ex-or phase modulator, to keep it in step with the modulation envelope. The reference voltage can also be keyed through another LPF to get a nice textbook CW keying waveform.
>
> The high power modulator part is physically the biggest bit.
> I started off wanting to build a PWM modulator, but it ended up
> being linear - this sounds like a thoroughly bad idea, massive heatsinks and so on, but the power dissipation did not turn out as bad as you might think. It evolved like this:
>
> With a PWM, the output tends to be quite spiky due to non - ideal
> filter components, and the amount of filtering you can apply is limited - very big filter capacitors limit the slew rate of the output in the troughs of the modulation, multi-section filters introduce other problems, especially if you want to use feedback to improve regulation, reduce mains ripple etc. The spikes will produce IM products in the PA output, which Murphy's law dictates will be at the right frequency to cause maximum trouble. One way round this would be to clock the PWM at the carrier frequency (or better, 2x), but then you end up with something quite big and complex due to the high frequency, with higher switching losses to boot, which defeats the object.
>
> So I looked again at using a linear circuit. If you keep the input-
> output voltage differential to a minimum while transmitting a continuous carrier, the power dissipation is minimised. Since you don't really need a perfectly constant output provided it is reasonably ripple free and so on, the modulator output can be made to track variations of the input voltage, and maintain a roughly constant, small, differential. I did this by using a filtered sample of the raw input DC as the reference voltage for the envelope waveform circuit described above.
>
> When sending CW at full power, the output to the Decca PA is 60V max, 21.4A, and the differential 5V. Power dissipation is 107W. During the phase transitions of BPSK, power dissipation is higher, but in fact with a sinusoidal envelope, average dissipation works out to 243W for phase transitions on every bit, or 175W for a phase transition every two bits, which I reckon is about average. The power dissipated is therefore similar to a 300 - 400W linear, which is quite managable. I think the overall TX efficiency DC input - > TX out is near 75 - 80% still, so not too bad. With a PWM modulator, there would still be some dissipation and you would be lucky to get 90% overall efficiency, so overall there is not a huge difference between the two approaches - less than 1dB in signal terms. The pass element uses 8 x STW34NB20 Mosfets, which is a bit oversized, but was done in case I felt like increasing output in the future. 5 or 6 would do. The linear design makes it very easy to include
> foldback current limiting and so on. Each MOSFET is driven by a seperate op-amp to ensure even current sharing, and individual current limiting for each device. Overall, the circuit maintains a fraction of the output voltage equal to the instantaneous amplitude of modulation waveform - in other words, it is a big feedback amplifier. This also serves the function of removing mains ripple from the PA supply
>
> The phase keying signal can be any logic level signal - for "Coherent", it uses the signal on one pin on the PC RS232 port provided for the purpose. For WOLF, I built a dedicated EPROM keyer, very similar to the Lowfer designs. This enables me to use the reference output of the synthesiser, divided down to 10Hz, to provide very accurate timing for the BPSK signal. The synthesiser uses a Racal 9442 OCXO reference, stable to within 1 part in 10^7, which it seems to manage easily. So you just dial in the frequency, and there is no messing around with frequency calibration for soundcards, HF rigs etc., thank goodness!
>
> The complete system is fully working with CW, QRSS and BPSK, but not yet finished. It is compatible with existing software for these modes. It should be flexible enough to use with other modulation methods, should these look promising. I think the results so far show that superior results can be acheived compared to a soundcard/SSB exciter/linear PA TX, althought there is obviously a lot more work involved. The same techniques could be applied to any class D TX. One thing I will do when I get the chance and the bits is to replace the current PSU with a variac/transformer/rectifier arrangement. I will get some legible circuit diagrams together when I get time; I am a bit pressed at the moment.
>
> Cheers, Jim Moritz
> 73 de M0BMU