OL interference

[DCuadrado]

Hello everybody!

I have recently built a module which up-converts a baseband signal between 90 and 190 to an IF signal between 2263,5 and 2363,5 MHz. This module needs an OL signal of 2453,5 MHz @ 0dBm which passes through a buffer amplifier consisting of a band pass filter, an amplifier and an attenuator. The OL reaches the mixer with a level of +12 dBm.

Now that I have exposed the situation here is the problem: the OL tone is interfering every part of the PCB. When I connect the spectrum analyzer to either the output or the input of my circuit, there is a tone of 2453,5 MHz and a considerable level. In fact, if I desolder the AC coupling capacitors, so that there is no physical conexion in the microstrip circuit between the three ports, the interference is still in both the input and the output.

It seems that the OL is interfering all the circuit through the air.

Please, do you have any solution to avoid this interference? The box is already in an aluminium enclosure box.

Thank you very much in advance.

 

[VEBill]

Debugging EMI/EMC issues at the PCB level is such a complex topic that it's going to be pretty difficult to help over the Internet.

Is the PCB laid out in accordance with best practices, e.g. strip lines with carefully controlled return paths to avoid creating current loop areas?

 

[DCuadrado]

Thank you for your answer.

Yes, I have avoided abrupt turns in the strip lines. I think the problem has more to do with the SMA connectors...

I have built a microstrip circuit just with the two SMA connectors and a line between them. I have cut the line and when I connect the spectrum analyzer to the other port, there is still a tone in the generated frequency.

I don't know how to eliminate the interference... Please, give me more ideas!

 

[Higgler]

True, tough to figure it out.
Power Supply and Ground connections are the first thought.

A picture will help.

 

[DCuadrado]

Higgler I am not allowed to upload any pictures... sorry. What should I take into consideration about the power supply and the ground connections?

 

[BrianG]

Quote: "I have built a microstrip circuit just with the two SMA connectors and a line between them. I have cut the line and when I connect the spectrum analyzer to the other port, there is still a tone in the generated frequency."

Some questions:

SMA Test circuit:
How big was the break in the test circuit? Does the SMA outer connect to a ground plane? Is the input to your spectrum analyzer terminated in 50 ohms? What is the level in dB that you see on your spectrum analyzer with the track cut?

PCB layout:
Is the input to your spectrum analyzer terminated in 50 ohms? Does the pcb have a ground plane? Is the band-pass filter screened; have you checked it's not oscillating at 2453,5 MHz?

 

[Higgler]

I'd say, make a stick figure in Powerpoint to show the components and ballpark sizes. I understand how pic's uploaded can get you fired, or worse.

 

[DCuadrado]

Thank you very much for your answers.

BrianG, yes my network analyzer is 50 ohms matched and there is a ground plane on both layers. The filter is not oscillating.

Please, I'd appreciate if somebody answered a couple of questions.
- What distance (clearance) should there be between the RF line and the ground plane (both on the top layer)?.
- Since I don't have a machine for metallizing the vias, does it matter if the vias in the ground plane are soldered with metal (tin)? I mean, there won't be air connection between the two planes.

 

[VEBill]

You're starting with +12 dBm.

What is the relative amplitude of the troublesome local oscillator leakage? A "considerable level" is too vague.

On a good spectrum analyzer, -50 dBm can appear to be huge, but it's better than 60dB down from the input (not great, but not bad either).


(This is a long shot, but worth asking I think.)

 

[zappedagain]

Quote - "BrianG, yes my network analyzer is 50 ohms matched and there is a ground plane on both layers."

We had a problem once with SMA connectors due to a poor connection between the top side and bottom side GND planes. This was at about 20 GHz though. Just because you have a large area of metal on the top and the bottom doesn't mean they are well connected.

Quote: "I have built a microstrip circuit just with the two SMA connectors and a line between them. I have cut the line and when I connect the spectrum analyzer to the other port, there is still a tone in the generated frequency."

How wide is your cut? A typical knife blade cut will make a really nice AC coupling capacitor at 2 GHz and you won't even notice the difference. You haven't mentioned the length of the trace or the size of the PCB; Is your length a significant fraction of a wavelength so you are seeing stub effects? Can you strip off more of the input/output traces to see if things change?

"What distance (clearance) should there be between the RF line and the ground plane (both on the top layer)?."

Are you saying you have a coplanar waveguide instead of a microstrip? Or are you asking what gap you need to not turn your microstip into a coplanar waveguide? You might try the calculators at microwaves101.com to see if your ground plane is too close.

Have you tried making S-parameters measurements on this PCB? That might explain a lot.

Z

 

[DCuadrado]

Both the top and the bottom ground planes should be well connected since there are vias every 2 mm and the connectors are soldered on both planes.

In the test circuit that I mentioned, the gap has a lenght of approximately 6 mm.

Concerning the clearance. I think what have is a coplanar waveguide since in the top layer there is the signal path, a clearance of 1 mm, and then the ground plane. Isn't it a coplanar waveguide?. Microstrip would be the same but without the ground plane, wouldn't it? The question is if that 1 mm is enough. I have read in some papers that the clearance should be of the same width than the RF path (1.5 mm calculated with the microstrip equations), but if I do this, the aspect of the PCB is not very good.

Thank you!

 

[biff44]

OL signal, OL tone...are these supposed to mean something to us?

http://www.MaguffinMicrowave.com

Maguffin Microwave wireless design consulting

 

[VEBill]

Based on the context, it seems to be "Local Oscillator".

 

[DCuadrado]

Yes, that's the local oscillator. I have built several small modules instead of a big one and now the interferences are smaller. They still exit, but are smaller...

 

[zappedagain]

"I have built several small modules instead of a big one and now the interferences are smaller."

So it sounds like your interference path changed. EMI has three variable you can play with to improve it - the source, the path, and the receptor. There can be many paths.

Can you try this - Hook up your LO (OL) to a PCBA and power it up. Place another PCBA with a 50 ohm termination on the LO connector near the first PCBA. Does the second PCBA show the same interference as the first PCBA? If the answer is yes then your LO has a radiated path; if not your noise is following a conductive path (ground bounce, poor termination, etc.). Does it change when you reorient the PCBAs? If yes that can mean your radiation source is on the PCB, not the LO/cabling.

How about the two PCBAs mentioned above powered off two separate power supplies? Is the path through the power supply?

I've recommended the following book many times on this site. The frequency range is getting a bit dated, but all the basics still apply:

http://www.emiguru.com/index.php?option=com_virtuemart&Itemid=35

Z

 

[biff44]

Well, reading this post with some hilarity, there are a number of potential issues you are facing.

1st, you need to do a sub-system level analysis of your block diagram. Sounds like you have some sort of mixer as an upconverter, and possibly amplifiers or filters. 1st u need to look at the mixer data sheet and see its LO port to IF or RF port isolation. When u hook up an LO signal....it IS GOING TO LEAK OUT of the IF or RF port at realatively high levels.

Your frequency plan looks suspect. You are sweeping 90 to 190 MHz, and upconverting to 2263-2363 MHz, but the LO leakage is at 2453. Lets say the LO to RF port isolation is 20 db, and you wanted 60 dB of LO suppression. You would need a bandpass filter with a passband of 2263-2363, and a rejection of 40 dB at 2453. My calculations say you would, theoretically need at least a 4 pole filter, and practically probably a 5 pole filter at the RF output port to achieve that.

Additionally, the mixer generates unwanted spurious outputs that are M x IF frequency +/- N x LO frequency. As the IF input changes, you will see these output spurs. For instance, when your input is 90 MHz, the mixer itself generates a strong 180 MHz 2nd harmonic inside of it. That 180 MHz upconverts with the LO to be a completely in-band signal. For this reason, often people design systems with MUCH HIGHER 1st IF frequencies. I know it is harder (faster DAC clock speeds, etc) but it solves a huge amount of spurious issues. You can not, for instance, transmit even 4-PSK or 16 QAM with an inband 2nd harmonic of an IF that is only -25 dBc down. Get a good mixer spur calculator program, and figure out the correct frequency plan for your application and specifications.

http://www.MaguffinMicrowave.com

Maguffin Microwave wireless design consulting

 

[biff44]

THEN, there are the board issues. Sounds like you have an open PCB--i.e. no shielding. While many systems can not afford the cost...it sounds like you are working more on a military type system where cost is less important. Try putting your PCB into a channelized metal housing. By channelized, I mean you make the LO path to the mixer follow a roughly straight line maybe .250" wide, and have a metal cover that has walls that are spaced .250" apart pressing down on the boards top ground plane. Underneath, of course, u want a solid ground plane. Do the same at the output...channelize all the way to the output connector.

You then need to design for other sneak paths. Active devices? They all get DC power from one voltage regulator. U need Lowpass Filters inserted into the DC bias lines. Digital lines on board? They can cause RF to jump around the board too.

Also, one often overlooked area is the connector itself. You need intimate contact of the connector body to the board ground plane. That board ground plane must be UN-CUT! No DC or Digital traces cutting it in half, as that would just generate a huge radiating standing wave that will cause isolation havock.

THen there is simple thru the air transmission of wireless signals--you are going to be hard pressed to isolate anything much higher than 35 dB just because the lines and the components all act like little antennas transmitting Rf into the air to be received somewhere else. Try putting some eccosorb on top of the board and see if your leakages go away. THIS is why systems requiring very high isolation either have channelized housings OR very clever frequency plans/filtering.

Good luck.

http://www.MaguffinMicrowave.com

Maguffin Microwave wireless design consulting

 

[DCuadrado]

Thanks for your answers. The best solution has been to divide the big PCB into smaller ones and put a filter at the input of each one.

 

[DCuadrado]

Of course each module must have its aluminium enclosure box with its SMA connectors.