Separation of RF board and microcontroller board caused problems

Welcome to the world of mixed signal design where GND is not GND, but a few mV above plus some noise on it depending on the spatial location. To start with, you first have to know what exactly you're measuring where and how.

Rene

--
Ing.Buero R.Tschaggelar - http://www.ibrtses.com
& commercial newsgroups - http://www.talkto.net

What thickness of copper was used on the flex board ? What's the power consumption of the radio-part ?

Possible solutions (one or all): Use a stranded connection for ground connection between both boards, better ground connection between both boards Use extra caps on the power lines on the boards (say 1nF for RF in parallel with say 47uF for power stability)

Ultimate solution Revise the layout, putting power-consuming parts closer to power supply (or power the boards from the side of the most power-consuming end)

Separation of RF board and microcontroller board has caused a problem which certain amount of distortion was measured from the transmitted DTMF tone. A check on the radio without RF and microcontroller separation showed that the distortion level was far lesser, 13% vs 5%.

The distortion disappeared when separate power supply was used for RF and controller board. Can somebody shed some light on this?

Probably the ground trace on the flex isn't wide enough and now the power trace acts as a ground connection as well. This can introduce noise if the power supply isn't very clean or becomes modulated by heavy load changes. Such load changes can happen when, for example, memory banks on the controller are read or written to.

Bottomline you may have to add ground connections, ideally a plane although that isn't easy on a flex. Also, look at your power supply with a scope set to AC. Check it down to the millivolt level.

Regards, Joerg

Is it FM? What is the transmitter output power, and where is the antenna? Where is the circuit generating the RF output? Is this a transceiver ... receiver? Please give more information. Maybe you can get more specific help.

You may be dealing with issues like injection locking. Changing layout may not be enough.

Regards,

Frank Raffaeli

I've checked the power supply with scope before. There were some noises on the line, with Vp-p ranged from 150mV to 200mV. These noises were found on 'normal radio' (without separation of RF and microcontroller) as well. But amazingly, there was no DTMF distortion seen in 'normal radio'.

A tweaking on audio PA to reduce its gain helped, but the audio volume was slightly low from rated audio. Distortion was mostly seen when audio knob was turned to max.

Audio PA resides in microcontroller board. When audio PA supply (7.5V) was connected with separate external supply, no problem was observed. By disabling the speaker from audio PA, there was almost no distortion seen at DTMF tone. The crux is, audio PA supply. Additional info is the audio PA circuit was similar with 'normal radio'. Can the problem occurs when RF and controller share similar supply? In 'normal radio', they are sharing same supply as well, but no problem was seen.

Hi,

150mV is a whole lot of noise. When you split the portions with a flex and there isn't too much ground in the flex then part of the ground return will inevitable be via the supply. That's when some of this 150mV noise can be modulated back onto RF signals, clocks, PLL loops and so on.

Can you try to regulate the voltage going to the DTMF circuitry, or to all the circuitry except for the PA? The idea would be to feed the PA raw power and then the rest of the unit gets clean regulated power. You can use a low drop-out regulator so you don't lose much in voltage. It looks like there is some noise caused by the PA stage that ends up on the supply rail and then enters the rest of the circuit. Alternatively, you could try to decouple the PA but that's harder. There you either need lots of capacitance and there may not be enough space for that, or some kind of nifty current regulator that evens out the spikes being sent from the PA.

One test I'd do for sure and it's easy: Lay a short and very wide copper path, for example 3M copper foil, from one board to the other and see whether the DTMF distortion drops to acceptable values.

Regards, Joerg

Finally, the root cause of the distortion had been found. There were four screws on the radios. The screws serves the purpose of holding radio PCB and chassis together. When one of the screws was removed, somehow, no distortion was seen at DTMF tone! That screw (contacted to ground) resided very near to 7.5V supply line in third layer of PCB layout. But I was still a bit bewildered on what happened actually. Could you please shed some lights?

I don't think that's the problem. When washer or insulator was mounted on the screw ground pad, no DTMF distortion was seen after the fourth screw been tightened.

The four mounting posts aren't coplanar, so when you tighten the fourth screw, it warps the board.

Cheers! Rich

Hi,

That sounds like either the screw may be touching some traces down in the PCB or, more likely, you have a serious case of a ground loop. Do both boards have a full ground plane?

Regards, Joerg

The screw only touches the ground pad at first layer. No full ground plane is used. Insulation of screw from ground pad or removal of screw can solve distortion problem. It's still a mystery to me.

I read in sci.electronics.design that Dummy wrote (in ) about 'Separation of RF board and microcontroller board caused problems', on Sun, 5 Sep

2004:

It's almost certainly a ground loop. But exactly what is happening depends on the details of the schematic and PCB layout, which is not practicable on the net.

--
Regards, John Woodgate, OOO - Own Opinions Only. 
The good news is that nothing is compulsory.
The bad news is that everything is prohibited.
http://www.jmwa.demon.co.uk Also see http://www.isce.org.uk

Hi,

I agree with John, this sure sounds like a ground loop. It can be very hard to make a RF/controller combination work reliably without a full ground plane. I would probably spring for a relayout with a full ground plane even if that means one or two more layers on the boards. The other issue with this unit might be EMC, especially with a plastic enclosure.

Regards, Joerg