Interesting digital radio challenge - find the design fault

[Jack Thonet]

G'day guys,

Following the steps of the Elektor DSP radio using the famous Si4735 SW/LW/FM radio chip,
I have designed the circuit board of the radio.

The radio version works fine. I even built a few models with the first design and
the signal is good. AM is harder to get working than FM but both work fine.
The signal strength measured by the chip itself is generally between 30 to 40 (dBum I believe).
This provides a good signal. With signals of that strength, the SNR is generally between 12 and 25 dB which is plenty.
It also allows for the radio to pick up the RDS information (incl. station name and time) fairly quickly.


Now I altered the design a little bit for version V2.0.
Really not much has changed. But with the first two models I built, I have a very annoying
design problem. I really believe it is the second design which is for some simple reason flawed.
Both models work but have the same symptoms of poor signal.

The reception is now around 10dBum only and generally the SNR gets to a maximum of 8 to 10dB.
This gives poor sound quality and most of the time the RDS info cannot be picked up by the chip.
Also, at the bottom and high end of the FM spectrum, reception is slightly better.
For most stations the signal is 20 to 30 dBum weaker than it should be.

Could you have a go at comparing both designs and point out what the matter is with V2.0?
Thank you very much in advance.

Jack

PS: I don't know the answer hence I need your help. Any comments are welcome.
I just would like to get version 2.0 working as smoothly as version 1.0. Thank you.

 

[blueroomelectronics]

Post your schematics in a readable resolution & png format.

Elaborate on what you changed and why.

And yes, that's some bad soldering.

 

[Jack Thonet]

Here are the two schematics of V1.0 and V2.0 in a more readable format.

Changes from V1.0 to V2.0 are minimal:
- new jumper JP6 to switch off battery pack (no change noticed if closed or open)
- new switch SW6 in order to swap between sets of speakers (K9 under or K5 on top)
- 4 extra mounting holes (MH5 to MH8)
- push buttons layout changed (smaller footprint)
the elements in the PCB have been brought closer in order to make some space for
the extra components
- routing of the traces in the Digital RF Processing section of the board is slightly different
- a few ceramic capacitors and the crystal X2 around the radio chip Y1 have been shuffled around
to accomodate for new components and traces

When you commented on the bad soldering, are you especially referring to the top side of V1.0 or
to all the pictures in general?

Thanks in advance for the help!

Cheers,

Jack

 

[JimB]

Are you sure that JP1 is on the correct pins?
I think it should be on pins 2 and 3, as best I can tell it is on pins 1 and 2 which will kill the AM radio sensitivity.
Is your problem with the AM or FM sensitivity, it is difficult to tell from your description.

Do you have the correct values for the inductors?
I notice that on your schematic, L3 L4 and L5 have different values form the Elector schematic.

JimB

 

[tunedwolf]

In addition to what Jim mentioned, you might want to consider scrubbing both boards with some good flux remover, it will be doing nothing but adding to your stray capacitance

 

[Jack Thonet]

Are you sure that JP1 is on the correct pins?
I think it should be on pins 2 and 3, as best I can tell it is on pins 1 and 2 which will kill the AM radio sensitivity.
Is your problem with the AM or FM sensitivity, it is difficult to tell from your description.

Do you have the correct values for the inductors?
I notice that on your schematic, L3 L4 and L5 have different values form the Elector schematic.

JimB

Hi Jim,

Thank you for your question.

The problems I have are with FM sensitivity.
The jumper is normally set between pins 1 and 2.
It normally allows for sufficient sensitivity with both modes, AM and FM.

Shorting pins 2 and 3 helps with sensitivity in the AM mode.

I uploaded the second article from Elektor showing the different antenna configurations.

I have built a few units of version V1.0 and managed to have it work quite smoothly in AM,
I got then about 20 stations in the AM band.
But my main worry is the FM mode, I'd like to get it to work.
Hence I need to use the jumper between pins 1 and 2.

I noticed that my schematic has a small issue.
The inductors have all the same footprint. When designing the PCB with DesignSpark, I created only one component for the inductors,
using one schematic symbol and one PCB footprint for all five inductors.
I should have corrected the values in the schematic but it didn't happen.

The correct values are the one from the Elektor article (refer list of components).
Inductors
L1 = 10uH (Fine Inductors Fastron type SMCC-100X-YY)
L2 = 0.1uH (Fine Inductors Fastron type SMCC-R10X-YY)
L3 = 4.7uH (Fine Inductors Fastron type SMCC-4R7X-YY)
L4 = 220uH (Fine Inductors Fastron type SMCC-221X-YY)
L5 = 2.2mH (Fine Inductors Fastron type SMCC-222X-YY)
I used the original Fastron ones that have a good Q factor.

Thank you Jim for looking into this!

Kind regards,
Jack

 

[Jack Thonet]

In addition to what Jim mentioned, you might want to consider scrubbing both boards with some good flux remover, it will be doing nothing but adding to your stray capacitance

G'day tunedwolf,

I'd agree with you, a good scrubbing wouldn't hurt.
I didn't have it handy at the time and the first model worked fine, so I just left it.
As the second unit (V2.0) is soldered in the same manual way, I think both units have some
level of stray capacitance, which still doesn't explain why the first one works fine and the second
one not quite.

Looking at the symptoms, I realise that the symptom clearly points at a weak antenna.
Most stations are way below their normal signal strength. The stronger stations are just acceptable
and the weaker ones get just to weak to listen to. 90% of the available RDS data is not retrieved.

Thank you for contributing to the thread!

Cheerios,
Jack

 

[ronsimpson]

famous Si4735 SW/LW/FM radio chip

I worked with these "famous" Si parts when they were new. I did a demo project, working for "Si", using these parts. I used the square package which is hell to solder down with the equipment I had then. I am a little surprised to see it working well on a adapter. I know what is inside the IC and it needs good PCB layout.
Look at the ground! See how short the "RF" traces are. The power supply capacitors are at the part.
If I had to use a adapter I would include the ground and the RF parts on the adapter along with power supply capacitors.

 

[Jack Thonet]

Hi Ron,

Thank you for your post.
Yes, I saw there was a QFN package that could be used. There were only 20 pins to handle instead of 24 but the SMD soldering
was as you said tricky to solder. Also, I think it is even more prone to impedance mismatch and more dependent on the PCB design.

Thank you for posting the pictures, very interesting to see a project by Silicon Labs themselves.
I bet nobody would know the chip better than Si!

Maybe you just hit the nail there. The version V1.0 design is very close to what the original Elektor design was.
Even though the PCB design using a breakout board looks very straightforward without much fine tuning,
maybe their team had arranged the circuit so as to get the right signal to the chip and minimise any interference between components
and traces...
In version V2.0 the RF traces are laid a bit differently. The components are the Silicon Labs chip have been shuffled around.
I didn't see much on design rules in the chip datasheet / design manual. I've attached it for anyone who needs more bed reading material.
So I assumed that the layout wouldn't influence the chip's performance that much. Maybe I was wrong...

Just a few questions on the Si47xx daughterboards you posted:
- what is the type of connector for the FM test and AM test. It looks gold plated.
- in the first picture, what are the pins of J2 used for?
I just trial and tested a few things. When adding a capacitance of around 50nF between pins 8 and 9 of the chip (Y1),
it improves the signal, but it is still not as per V1.0.

Has anybody an idea of any specific design rules for the chip Si4735 ?

Thank you all!

Jack

 

[JimB]

The problems I have are with FM sensitivity.

OK

The jumper is normally set between pins 1 and 2.
It normally allows for sufficient sensitivity with both modes, AM and FM.

The connection from JP1 1-2 is the configuration for separate antennas for AM and FM
When using the FM antenna connection (BNC) there is no direct connection to the AM input of the Si 4735.

Shorting pins 2 and 3 helps with sensitivity in the AM mode.

Yes, because that connects the AM input of the Si 4735 to the BNC connector.

Other than problems with RF layout of the circuit board as suggested by ronsimpson, I can only ask about the test environment.

Are the tests taking place on the work bench?
What antenna are you using?

Is it possible that your antenna is just a piece of wire laying on the bench?
Is it possible that your bench is a noisy RF environment from various computer equipment etc?

Do you have a signal generator to make real sensitivity measurements rather than rely on off air signals?

Other than that it is difficult to diagnose these sort of problems at a distance.

JimB

 

[ronsimpson]

- what is the type of connector for the FM test and AM test. It looks gold plated.

I have boxes of them but I can't remember the name. Probably what is on "wireless 2ghz antennas", and routers.

I just trial and tested a few things. When adding a capacitance of around 50nF between pins 8 and 9 of the chip (Y1),
it improves the signal, but it is still not as per V1.0.

That is interesting. I don't have time now.

Has anybody an idea of any specific design rules for the chip Si4735 ?

When I worked for Si I had rules. I will look.

If you do make a layout for the square IC, please ask me first.

 

[ccurtis]

- what is the type of connector for the FM test and AM test. It looks gold plated.

They look like "SMA" jacks (female), but without seeing the end view, I can't be sure. The "SMA" plug (male) will mate with them. "SMA" connectors are 50 ohm, RF coaxial connectors good to about 18 GHz. Those are probably gold plated from the looks of them in the photos, but they are available in all stainless steel, and perhaps other metals and plating also.

 

[Jack Thonet]

OK


The connection from JP1 1-2 is the configuration for separate antennas for AM and FM
When using the FM antenna connection (BNC) there is no direct connection to the AM input of the Si 4735.


Yes, because that connects the AM input of the Si 4735 to the BNC connector.

Other than problems with RF layout of the circuit board as suggested by ronsimpson, I can only ask about the test environment.

Are the tests taking place on the work bench?
What antenna are you using?

Is it possible that your antenna is just a piece of wire laying on the bench?
Is it possible that your bench is a noisy RF environment from various computer equipment etc?

Do you have a signal generator to make real sensitivity measurements rather than rely on off air signals?

Other than that it is difficult to diagnose these sort of problems at a distance.

JimB

Hi Jim,

Thank you for your questions.

The general setting is using pins 1 and 2 shorted.
In this case AM radio is picked up before L2 and mainly uses the inductors. The long FM antenna that may be connected
would probably not affect the reception.
The FM radio signal is picked up after L2 and mainly uses the FM antenna connected through either the terminal block or the BNC connector.
The chokes are not affecting the FM range as I understand.

I have successfully used version V1.0 in that configuration with a good signal on all FM stations.
In the same conditions, the reception of V2.0 is a lot poorer. The difference is very significant.

The test environment:
on a wooden work bench
using a whip antenna (1.8m long) ass FM antenna
(using a 2m wire through the BNC connector has similar results)
no computers or RF equipment around.
The power supply (standard 5V power supply) is not noisy. I tried with batteries or with 6V from the car (using a resistance voltage divider)
and the result is the same.
I tried with a small chinese made 3$ voltage regulator (screw type pot) with 3 digit display and the result was awful,
the low power quality was affecting reception.

The enigma is really to find which difference in the design can affect the overall performance so much.

Thank you for looking at it !

Cheers,

Jack

 

[Jack Thonet]

I have boxes of them but I can't remember the name. Probably what is on "wireless 2ghz antennas", and routers.

That is interesting. I don't have time now.

When I worked for Si I had rules. I will look.

If you do make a layout for the square IC, please ask me first.

Thank you Ron.

If I look at making a layout with the QFN package, I'll make sure to run it by you.

I think that working with SMD components has quite a few advantages.

Thanks for your comments.

 

[Jack Thonet]

Thank you very much all for your comments on this thread.

I realise that the PCB layout V2.0 that I posted wasn't the very last version that I sent to the board house.
Sorry about that. I have attached the PDF file (PCB DSP V26 V2.0).

My conclusions so far are:

- as I have tested the radio chips, microcontrollers, buttons and power supply, they don't seem to be the cause of trouble

- the PCB layout of the RF part is quite more critical than I thought.
maybe the connection from the radio chip to the antenna is quite different from V1.0 to V2.0 and is detrimental to signal quality
( I have attached a picture showing both RF section, and a reference to the schematic, which remains the same).
I can see that on V2.0, the routing is longer and more convoluted. A trace carrying the RF signal is even crossing another one.
Even though they are well segregated (90 deg angle, one on top, the other on the bottom), this might cause interference.
What do you think?

I haven't found any major obvious flow with the new design. It works, the RF signal is just poor.

Many questions still remain:

- some PCB traces run under the breakout board for the radio chip, would that possibly influence the signal?
- the RF signal is weaker by a fair amount. How come?
- is there any issue with impedance matching on the RF input circuit? If yes, how to go about it when designing?

Thank you all again for your input.

Best regards,

Jack

 

[ronsimpson]

1)Your PCB traces have inductance and capacitance that are not helping and not on the schematic. At 100mhz a inch of wire adds to the schematic. You have wires going everywhere. Long wires to D4,5. The adapter board pins add only 2 to3pF and some inductance.
2)The grounding is not good. The bypass capacitor is a long way away form the IC.
3)I can not remember. Are you using the same inductors or did you change types?
4) On the other hand Elcktor has long traces BUT they have a ground plain! The entire top layer is ground.
5) Many of the inductors are used for (not FM or AM). There is much junk here for short wave. Down at AM the trace length is not important.
Look at this layout. There is ground every where.

 

[Jack Thonet]

1)Your PCB traces have inductance and capacitance that are not helping and not on the schematic. At 100mhz a inch of wire adds to the schematic. You have wires going everywhere. Long wires to D4,5. The adapter board pins add only 2 to3pF and some inductance.
2)The grounding is not good. The bypass capacitor is a long way away form the IC.
3)I can not remember. Are you using the same inductors or did you change types?
4) On the other hand Elcktor has long traces BUT they have a ground plain! The entire top layer is ground.
5) Many of the inductors are used for (not FM or AM). There is much junk here for short wave. Down at AM the trace length is not important.
Look at this layout. There is ground every where.
View attachment 91387

Hi Ron,

Thank you very much for your valuable comments.

1) yes, I think that by squeezing the board and especially by making the routing in the RF input part of the board, I have added undesirable inductance and capacitance.
The traces have to run somewhere but I can see that the V2.0 is not as good as V1.0.
I have no real idea how much inductance / capacitance two close or crossing traces add to the design.

2) agreed too. In the reshuffling of components to create some spaces, I overlooked the chip manufacturer's recommendations and also good practice.
I think the location and routes to C15, C16, C18, C17, D4, D5 and R6-R8 needs to be improved.

3) I reused the exact same inductors from the original design. Looking at the specs, the quality factor Q of most SMCC inductors used is above 80.
The original Elektor design, as well as my versions V1.0 and V2.0 use the same chokes.

4) My apologies here, I probably have been misleading. Both designs use a ground plane. I think that it is indeed very important for the overall RF performance.
In order to provide a good view of the routing, with as well top and bottom traces, I had to clear the copper pour (top layer ground plane). Otherwise bottom traces would stay hidden.
My phone camera is of poor quality so even if it can be seen on the pictures, it can be hard to tell.
I should have said it when providing the files, sorry about that.
Most traces run at the bottom as I wanted to have less traces on the ground plane and for the looks of it.

5) as far as I understand:
L1 is used to filter out the power supply
L2 is used for FM
L3 is used for AM (SW)
L4 is used for AM (MW)
L5 is used for AM (LW)
I am not too good with RF design. Do you the SW traces could be improved?
Is it the length or the routing that is troublesome?
The ground connections around the inductors are made to select from the software the desired antenna configuration I believe.
By pulling down pins 14 or 26 from the Atmega, the program shorts the part of the antenna that is not desirable. (not quite sure though..)
I am not too sure why D4 and D5 are required. If someone can shed some light on this...

Thank you for the comment Ron.
I am not sure where to start to improve the design..

Cheers,
Jack

 

[Jack Thonet]

Hi all,

I just wanted to provide a bit more background on the DSP radio.

I added a video showing the radio V1.0 in operation.
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 or if there is something wrong with the schematic
of version V2.0.

The RF design (of both versions) is unfortunately not optimised. Some mass-produced digital radio do slightly better.
I have optimised the position of the antenna to get good reception which makes up for that quite well.
But ideally, I would like to be able to optimise my design.

Also, I still don't grasp exactly why V2.0 gives such a poor result.
I think ronsimpson is absolutely right in the fact the cause is a poor board design around the RF reception.
Can anyone give advice or links on how to better design the RF part of the board?
What to look out for?
Which specific rules to follow?

Thank you very much in advance.
Regards,
Jack

 

[ronsimpson]

I don't have time to look at your grounds.
You have a very sensitive RF area mixed with a (DSP and 8051 in the IC) along with you PIC (can't remember) and a LCD display. All are very noise. There could be ground or supply current flowing from the PIC through the RF area.

 

[Jack Thonet]

Hi Ron,
Thank you.

You are quite right, there is definitely some improvement to do with the grounding. Even in the better version V1.0, communication between the microcontroller (AtMega168) and the DSP radio chip (Si4735) can be heard through the speakers (at low volume).
You can hear the bursts of data packets sent from the microcontroller, which creates some knocking / tapping / clicking in the loudspeakers.

If you have any ideas (even later) of how to improve the grounding of the RF part, please let me know.

Thank you in advance!

Cheers,
Jack