Interesting digital radio challenge - find the design fault

[JimB]

I also wired the radio as per V2.0 on a breadboard and this gave as poor results as the actual
populated circuit board. There is really not much difference in the layout.
I am wondering whether the breadboard is also a poor RF design

That all depends what you mean by "Breadboard".
If you mean one of those matrix boards where you plug the components into strips - it is a wonder that it works at all.
If you mean a sheet of copper clad board with the components soldered directly to the groundplane and all signal connections made short and direct, then I am surprised that you had poor results.

Have a look here:
https://www.electro-tech-online.com/articles/breadboards-how-bad-are-they-at-rf.278/
for something which I wrote here some time ago.

JimB

 

[tunedwolf]

I have to agree, "in the air" should produce decent results, so I can only assume he did use breadboard to test with. Of couse he could have 3' hook-up wires going everywhere too

 

[Jack Thonet]

Hi Jim, Hi tunedwolf,

I actually did use a breadboard. The first time I had put V1.0 on a breadboard I had decent results.
My view was this: if I got significantly better results on the breadboard than on V2.0, then I could easily conclude there was sth wrong with the board design...

Results are just as bad on the BB.

Here below you'll find a picture of the BB and a video of V1.0 installed as a car radio.

Thank you for your feedback.


Jack

 

[Jack Thonet]

Hi Jim, Hi tunedwolf,

Here below you'll find a picture of the BB and a video of V1.0 installed as a car radio.

Jack

Here is the video.
To see the video, upload it from my Dropbox:
https://www.dropbox.com/s/xz75ygbwm7ocuse/DSP radio v1.0 demo.3gp?dl=0

Thank you Jim for the link. Your thread showcases really well what is happening with my circuit.
I certainly agree with all the comments. I think most of the interference caused in the breadboard is happening "under the hood",
as the metallic connection strips under are close to each other side by side this introduces undesirable capacitance and inductance.
Maybe I wasn't expecting it to be as detrimental as it actually is.

I had a lot of different things to sort out for this project. One of them was the microcontroller program so early on I built it on a breadboard to make sure
I had a good grasp on the schematic, components used and I could develop the interface at ease.

In the light of this separate thread, I think the test I did on the breadboard doesn't bring to much to the debate.
I am still trying to figure out what is inherently wrong with the second version..
The RF section design seems to be the cause...
If one of you spots any bad design aspects in the RF section of V2.0 please let me know.

Jack

 

[mvs sarma]

you need to surely follow rf practices with ground plane and also proper track width etc.
generic pcb layout like we use for slow speed digital circuits would deteriorate rf sensitivity..You better try to follow the elektor component placement. Of late i am trying non pth double sided pcbs a home.
should not be troublesome if you can use uv process. even eagle (under cam ) has provision for inverting the artwork to create negative.. please do study.

i really appreciate your keenness trying out si47** chip

 

[Jack Thonet]

Hi MVS sarma,

Thank you for your post.

I certainly agree, I need to follow RF practices. The thing I need to know is just that:
what are these RF practices?
What did I miss in V2.0 although it is so similar to V1.0?
Is there any reference you know, online or on paper where I can learn about those design rules?

Again I agree that Elektor certainly work to some extent on RF aspects of the board. It's probably why V1.0 works fine. I would like to understand what is causing the poor performance.

If you know about the RF design rules, it would be great if you could have a look at the RF section (last attachment from my previous comments).

Not sure why your are using non plated through holes. I only use these for mounting holes. PTH make for a better connection.

For my boards I use board houses. The price has come down so much that don't find it worthwhile to DIY.

Thanks again for your post and to the other contributors!


Cheerios,
Jack

 

[mvs sarma]

did you have ground plane?
can we see the assembled photograph , (a crisp close up photo !)

 

[Jack Thonet]

Hi mvs sarma,

Yes, the top layer is a ground plane.
I would agree that allowing for a good large ground plane is probably the number one rule for a good RF design.

Here below, you will find the files where that can be seen. I wish I had a better camera as the quality is very average.
Yet, on the upper face where the display is, there are only a few traces, which all appear as two parallel lines as they
are routed on the ground plane. The PCB layout from DesignSpark is also. I also included the original Elektor article showing
the circuit board with a clearer picture. The layout of V1.0 is very similar to that.

Thank you for your comments!


Jack

 

[Jack Thonet]

Hi all,

Thank you in advance for what I'm going to ask!

Does anyone know of specific RF design rules with the Silicon Labs DSP FM radio chips ?

I have found a SMD design for the Si4731, very similar to the Si4735,
This design uses the QFN package.

The RF section schematic and PCB layout is attached.

Apart from:
- the ground pad under the chip
- large traces for ground on both sides
- capacitors being close to the chip

I can see any specific rule being observed.
All components seem to be located at a location convenient for the PCB layout...

Maybe I am missing something...

Thank you all!



for now,
Jack

 

[Jack Thonet]

Hi all,
I was discussing this thread with a peer. He's an electrical engineer and we attended electronics lessons together.
Looking at the RF section of the board, which is I believe the source of the discrepancy in the board performance between V1 and V2,
I believe, something might be adversely affect reception in V2.0

When you look at the trace between C17 and (D4 / D5), this trace is very short in V1.0 (<1cm) and long with 2 vias along the way in V2.0.
I think this part of the schematic is used as a filter to prevent damage from EMC. Stray currents due to electrostatic discharge would be brought to ground
via D4 and D5.

Please tell me if my analysis below makes sense:
The impedance of this branch to ground is basically the internal capacitance of the diode.
The value of the capacitance is about, from the datasheet would be around 10 - 15 pF if the antenna provides a signal of around 500mV peak.
I calculated the value of 4.5cm of trace to be around 2pF in V1.0 while less than one cm would be around 0.3pF in V1.0.
That extra capacitance of 1.5pF might detune this EMC circuitry and actually allow noise into the RF section.

I'm rally not too sure about this as it obviously stretches my understanding of RF electronics.
Is anyone a bit more knowledgeable about this?

Thank you heaps!

Jack

 

[JimB]

When you look at the trace between C17 and (D4 / D5), this trace is very short in V1.0 (<1cm) and long with 2 vias along the way in V2.0.

Yes it does wander about a bit, but I don't think that it is excessive.

I think this part of the schematic is used as a filter to prevent damage from EMC. Stray currents due to electrostatic discharge would be brought to ground via D4 and D5.

Yes, I agree with that.

... if the antenna provides a signal of around 500mV...

Whoa! Hold on there!
500 MICRO volts would be a big signal, for 500 MILLI volts you would have to be standing very near to a big transmitter.

However, as a test it would be worth removing the diodes and shorting out C17 and C18 to see if any of them are faulty.

There is very little in the way of "tuning" and selectivity about that antenna circuit.

JimB

 

[Jack Thonet]

Hi Jim,

Thank you for your post.
From what I've read you seem quite right, 500mV is a lot more than what I will get at the antenna.
I had scribbled down a sort of voltage map on the schematic while I was troubleshooting.
It seems that I had 10mV DC and 50mV AC at the connection C17/C18/D4/D5.
So this would make the capacitance of the diodes around 20-25pF from the graph, which in turns makes insignificant
the added 2pF from the longer trace. It wouldn't impact the circuit really, which seems to be your conclusion as well.

I should test the capacitors to see if they are ok. The value of 47pf makes it quite difficult to test.
I have a good enough multimeter, also measuring capacitance, diodes and transistors.
It's a rugged model from a good brand, not really a basic model. But I found that capacitance below 2 nF come to the same value
of around 1.1 nF. So I cannot really measure a 4.7 pF, 47 pF or 100pF.
Also, I assembled two units and both display the same poor performance. The V1.0, using the same batch of components but the other board design
is working fine (I assembled two units as well).

I would like to test as you said C17 and C18 to see if they are faulty.
So if I understand correctly, I would desolder D4 / D5 to remove them from the circuit and
add two bridges over C17 and C18. This would actually connect the antenna directly to the inductor L2.

-> Is it what I need to do to test the circuit without C17 and C18?

Thank you very much for your input!

Jack

 

[JimB]

I would like to test as you said C17 and C18 to see if they are faulty.
So if I understand correctly, I would desolder D4 / D5 to remove them from the circuit and
add two bridges over C17 and C18. This would actually connect the antenna directly to the inductor L2.
-> Is it what I need to do to test the circuit without C17 and C18?

Yes, just unsolder one end of each diode, or remove them completely - same effect.

For the capacitors, you could just put a jumper wire across both capacitors.
Yes that would effectively put the antenna directly to the inductor. Just be carefull that your antenna does not short anything out or connect to any unwanted voltages.

JimB

 

[JimB]

If none of this improves things or finds a problem, I suggest that you investigate improving the grounding of the RF chip as hinted by RonSimpson in post #8.

To try an improve grounding as best you can for a test, connect short wires from the GND and RFGND pins on the Si chip directly to the ground plane on the circuit board.
Also, you could try connecting small 0.1uF capacitors between the supply pins and GND pins, again do this directly at the chip.

JimB

 

[Jack Thonet]

Thank you Jim for your advice.

I'll try these as soon as I get a chance.

Thanks again.

Jack

 

[Jack Thonet]

Hi Jim,

Thank you very much. It seems it worked.
I literally bypassed the EMC filtering part of the RF section and immediately had a better signal.

With the two capacitors bypassed, the faulty section acted as a low pass filter. I noted the signal strength on all stations. The first stations were still present and
increasingly attenuated as I tuned up within the FM range. Above 98MHz, I wouldn't get any signal at all.
Once I also took out D4 and D5, the signal improved dramatically. The overall performance is even better than V1.0 of the board.

I am not too sure why but the D4-D5-C17-C18 section was the culprit.

Now reception is great! Thanks a million !

I have attached a spreadsheet signal data with three different installations: V1.0, V2.0 (as per the original design, faulty), V2.0 modified (D4-D5 removed, C17-C18 replaced by jumpers).

The results speak for themselves. Thank you very much for your suggestion JimB, it saved my V2.0 design.

Now, I have a worry, but it may be only with this board and not actually with the design.

As it is kind of an new issue, I'll start a new thread but will refer to here, as it relates to the same project.

Regards,

Jack

 

[JimB]

Very interesting!

It feels very wrong to miss these components out altogether, so to find the fault, you could:

1 Rebuild with different components

2 Test the suspect components

3 Build the input circuit using "ugly" construction to determine if the PCB itself is the problem.

JimB