Tearing my hair out with (what should be) a pretty straightforward transmitter

Ok, this is the problem:

I designed and built a single frequency conversion AM (ASK - Amplitude Shift Keying - to be accurate) transmitter.
In blocks it goes like this:

Oscillator (PLL) -> Double-balanced mixer -> Power amplifier -> antenna

The carrier frequency is 5.8GHz, the transmission rate is 500kbps with Manchester encoding.

Microstrip lines and components were used for 5.8GHz circuits. The thing was built in a mixed-dielectric printed circuit board as shown:

1) Top layer, microstrip lines essentially (copper)
2) PTFE/Teflon (dielectric)
3) Full ground plane (copper)
4) FR4 Prepreg (dielectric)
5) DC power supply tracks (copper)
6) FR4 (dielectric)
7) Low frequency tracks (copper)

Everything seems to work fine with CW (continuous wave, i.e., no modulation).

On the other hand, when the baseband signal is fed to the mixer, the thing goes wild!! The observed spectrum is heavily assymetric and far far away of whta should be (a nice, symmetric spectrum around the carrier).

Circuit is heavily influenced by approximating hands near the PCB, even 2 inches (5 cm) away! Sometimes, while moving my hands around, I can even get almost the perfect spectrum (of course, I can't ship the thing with my hands inside the box . Microstrip shouldn't be that sensitive!

What is going wrong? Any hints?

There are two important things you should correct:

As you wrote everything varies by your hands moving around, all your RF stuff must be installed in a metal case, with DC capacitor feedthroughs for DC lines, and coaxial connectors for other lines.
Manchester code has a DC component which may affect the balanced mixer and introduce the observed distortion. I would advise first to use a 0.1 uF coupling capacitor for the data line to your mixer, and a 1 kOhm potentiometer to adjust modulation level. Start with a rectangular pulse signal from a multivibrator, and see the received signal and spectrum. Adjust for no distortion, then adjust a similar amplitude for Manchester-code input.

OK, you are right, but I'm still confused....


Even if the circuit should (and it will be, after testing and debugging) enclosed in a metal case, my previous experience says circuit behaviour shouldn't be so wild even when I get my hands close. Some variation is to be expected, but at 1/2 inch distance the effect should negligible. As for power supply, it has a nice internal layer in the PCB with low impedance (wide) tracks and low impedance (large) vias.

Yes, I'm feeding some DC to the IF port of the mixer. The baseband signal has some DC. In fact, as far as I understand, AM (ASK if you prefer) modulationg signal never should go below zero, otherwise phase inversion will occur. While this is useful for PSK (Phase Shift Keying) or DSBSC (Dual SideBand Suppressed Carrier), it is undesirable for ASK.

So, you say I should avoid DC at IF port, my reasoning says I should keep it. What am I missing here? :?

I assume, that the intended modulation scheme is on-off keying. There won't be much use in applying less than 100 % modulation.

Regarding the observed problems, you should check as a first point, if it'a problem of instable PLL or self-oscillating PA. Without knowing the circuit details, it's pointless to guess further about possible issues. It's not unusual to build microwave transmitters/transceivers without closed metal case. Many WLAN devices don't have any.

It could be many things. I suspect that the changing load of the modulated mixer is causing your VCO to frequency modulate. so on the spectrum analyzer you are seeing at least two types of modulation superimposed.

First, put an oscilloscope probe on the mixer input (that connects to the VCO), and verify that you do not have any DC spikes coming out of it. If you do, HPF that port.

2nd thing i would do is put some isolation between the mixer and the oscillator. Even a 3 dB pad would be a good first step. Does adding isolation change anything? Don't forget that your modulation rate of 500 kbps is probably higher than your PLL bandwidth, so as far as the VCO is concerned, it is operating as if it were "unlocked".

If the isolation seems to help, try a 20 dB pad and a 20 dB amp jury rigged between your mixer input and the vco.

Did the problem go away? if so, report back and we can discuss how to fix it permanently.

BTW, you DO realize to get amplitude modulation from the mixer, you need an IF port that is DC coupled, and your signal must swing from 0V to perhaps 0.5 V. Not -V to +V.

FvM said:

I assume, that the intended modulation scheme is on-off keying. There won't be much use in applying less than 100 % modulation.

Click to expand...

Of course!

FvM said:

Regarding the observed problems, you should check as a first point, if it'a problem of instable PLL or self-oscillating PA. Without knowing the circuit details, it's pointless to guess further about possible issues. It's not unusual to build microwave transmitters/transceivers without closed metal case. Many WLAN devices don't have any.

Click to expand...

The PLL is an hybrid from Varil (now RFMD). Works pretty well (I'm using these hybrids for some years) and needs little more than an external frequency reference (I'm using a TCXO) and a serial interface to be program the dividers.

The PA doesn't seem to oscillate. With no modulation applied I get the fundamental (and it's harmonics, of ocurse). So, no spurious oscillation in a wild frequency is occuring.

---------- Post added at 11:34 ---------- Previous post was at 11:16 ----------

biff44 said:

It could be many things. I suspect that the changing load of the modulated mixer is causing your VCO to frequency modulate. so on the spectrum analyzer you are seeing at least two types of modulation superimposed.

First, put an oscilloscope probe on the mixer input (that connects to the VCO), and verify that you do not have any DC spikes coming out of it. If you do, HPF that port.

2nd thing i would do is put some isolation between the mixer and the oscillator. Even a 3 dB pad would be a good first step. Does adding isolation change anything? Don't forget that your modulation rate of 500 kbps is probably higher than your PLL bandwidth, so as far as the VCO is concerned, it is operating as if it were "unlocked".

If the isolation seems to help, try a 20 dB pad and a 20 dB amp jury rigged between your mixer input and the vco.

Did the problem go away? if so, report back and we can discuss how to fix it permanently.

Click to expand...

Hummm... Between the local oscillator (PLL) and the mixer I have two gain stages and a Wilkinson splitter. The Wilkinson splits LO signal for the transmitter and the (currently deactivated and well matched) receiver. The two amplifiers bring the LO power from 0dBm to comfortable +13dBm needed by the mixer. I expect those amplifiers to bring enough isolation between LO and mixer.

I'll try the attenuator and see what happens... Maybe some mismatch caused by the mixer operation is playing havoc with the amplification stage...!

biff44 said:

BTW, you DO realize to get amplitude modulation from the mixer, you need an IF port that is DC coupled, and your signal must swing from 0V to perhaps 0.5 V. Not -V to +V.

Click to expand...

Exactly what I was saying (but you put it in a simpler and more understandable way).