PDA

View Full Version : Recommended layout for better RFI/EMI results?



6581
11-19-2010, 10:07 AM
Hi,

I'm designing a propeller-based board and have seen that there is a lot
of high-frequency noise coming from the propeller.
It might be that my setup is not the best, but here's my current setup:

- 100F electrolytic capacitor at DC input (e.g. power supply - stabilized)
- 100nF and 10nF at LM2937-5.0 for 5.0V and LM 2937-3.3 places as near as possible
- Power supply traces are wider than the signal traces and almost connected with 0 Ohm to any VCC/VDD
- Star-alignment used for the grounding
- 100nF and 10nF's at each VCC pin of the P8X32A (Propeller)
- 100nF and 10nF's at the supply of the EEPROM.

Maybe I can make some pictures using a oscilloscope.
The reason I need to reduce the noise is that I'm going to use it for
sound synthesis, where noise can be quite annoying.

Any more suggestions?

Zap-o
11-19-2010, 02:13 PM
Do you make a ground plane? In other words I make the ground on my PCB the entire space left over on the circuit board. Fact is, I typically auto route my traces except the ground then once all the traces are routed I make the remaining space ground.

I have not noticed a high frequency noise on my boards that use a prop and I have some of them go through a rigorous CE testing process.

Maybe post a picture of the board?

Leon
11-19-2010, 03:03 PM
It's impossible to comment without a board layout. For instance, decoupling capacitors can be connected incorrectly making them ineffective.

6581
11-19-2010, 04:48 PM
Do you make a ground plane? In other words I make the ground on my PCB the entire space left over on the circuit board. Fact is, I typically auto route my traces except the ground then once all the traces are routed I make the remaining space ground.

I have not noticed a high frequency noise on my boards that use a prop and I have some of them go through a rigorous CE testing process.

Maybe post a picture of the board?

Sure. This is one of my boards with these effects:

http://loom.no-route.org/pmb-rendering.png

http://syso.name/wiki/images/thumb/5/57/Pmb-pcbv1-resized.jpg/700px-Pmb-pcbv1-resized.jpg

Note that the SMD 100nF and 10nF Capacitors are directly on the bottom side of the PCB connected using vias.

6581
11-19-2010, 04:51 PM
It's impossible to comment without a board layout. For instance, decoupling capacitors can be connected incorrectly making them ineffective.

I could see if I can convert my gerbers to some suitable graphic
e.g. png to post here.

The decoupling capacitors (ceramic) are located as near as possible to
each of the components and connected symmetrically to each of the pins.
The maximum trace length from a part to a decoupling capacitor is about
0.0787 inches (2mm).

Top and Bottom polygons are both GND planes.
The PCB itself is a 2-layer board with FR-4 material and gold-plated contacts
(since the PCB fab has done gold-plating as a gift on this order).

Leon
11-19-2010, 05:05 PM
You don't need to convert the Gerbers, just post the image from the PCB layout editor. The PCB software I use can save the screen (or a portion of it) to a file, or plot to Windows files. You can use a separate screen capture program if necessary.

Zap-o
11-19-2010, 05:41 PM
Looks like some nice PCB software what type are you using? Check that the xtal for the propeller has the correct capacitance. Maybe your switching power supply is causing the problem.

a suggestion:

Take a board and only populate the propeller and eeprom. Run it to see if it is in fact noisy. If not keep adding parts till the noise appears. If it is check input power, if that is okay then PCB may be truly causing the problem.

Cluso99
11-19-2010, 09:29 PM
I published my prop QFP layout (try the RamBlade thread). The prop should also have a 10uF tantalum or similar as the power spikes it requires can be large.

Ariba
11-20-2010, 06:50 AM
It is very important that the 3.3V regulator generates a stable, clean 3.3V voltage. Otherwise you can optimize the layout as much as you want, you will always have this digital noise.
I made good results with a LF33, you seem to use the LM2937 which I never tried. But from the datasheet of the LM2937:
"The minimum output capacitance required to maintain stability is 10 F (this value may be increased without limit).
Larger values of output capacitance will give improved transient response."

If you really have only a 100nF + 10nF at the output of the regulator, then try to replace the 100nF with a 10uF. I think you will get much lower noise.
Also the 5V regulator needs an output capacitor of several uF.

Andy

Sapieha
11-20-2010, 11:19 AM
Hi Ariba.

You are NEXT correct --- BUT never omit 100nF --- BUT if possible replace 10uF with 33uF and bigger.

100nF are for cutting high frequencies. Some times it is needed one 10nF one 100nF and one BIG one in parallel near REG.

10nf and 100nF on both sides of REG and at last 10uF on Input side and bigger on On output side



It is very important that the 3.3V regulator generates a stable, clean 3.3V voltage. Otherwise you can optimize the layout as much as you want, you will always have this digital noise.
I made good results with a LF33, you seem to use the LM2937 which I never tried. But from the datasheet of the LM2937:
"The minimum output capacitance required to maintain stability is 10 F (this value may be increased without limit).
Larger values of output capacitance will give improved transient response."

If you really have only a 100nF + 10nF at the output of the regulator, then try to replace the 100nF with a 10uF. I think you will get much lower noise.
Also the 5V regulator needs an output capacitor of several uF.

Andy

BigFoot
11-20-2010, 05:44 PM
We have sold thousands of Propeller based systems that use the LM2937-3.3 and it is a good little regulator, we have never lost one.

Your boards look great but I didn't see any Tantalum capacitors on them. We use a 47uF 10v tantalum along with a .1uF ceramic on each side of the regulator. There are also a couple .1uF ceramics under the propeller chip. All of these need to be surface mount low DCR units.

Through hole Resistors and Ceramic disk caps have more inductance than there SMT counterparts. It is a good idea to stay away from Electrolytic caps, they are inexpensive but the good designs use all dry cap's.

For precision audio applications, I would also put an inductor with a small ceramic on each side (Pi Filter) in the input to the board and maybe even on the regulator output. Mixing digital and audio systems is never easy, most guys use separate supplies and some type
of isolation between each type of circuit.

Grounding is also extremely important and use lots of decoupling caps.

6581
11-22-2010, 04:23 PM
We have sold thousands of Propeller based systems that use the LM2937-3.3 and it is a good little regulator, we have never lost one.

Your boards look great but I didn't see any Tantalum capacitors on them. We use a 47uF 10v tantalum along with a .1uF ceramic on each side of the regulator. There are also a couple .1uF ceramics under the propeller chip. All of these need to be surface mount low DCR units.

Through hole Resistors and Ceramic disk caps have more inductance than there SMT counterparts. It is a good idea to stay away from Electrolytic caps, they are inexpensive but the good designs use all dry cap's.

For precision audio applications, I would also put an inductor with a small ceramic on each side (Pi Filter) in the input to the board and maybe even on the regulator output. Mixing digital and audio systems is never easy, most guys use separate supplies and some type
of isolation between each type of circuit.

Grounding is also extremely important and use lots of decoupling caps.

Would this be okay? (No PI-Filter in this circuit):

http://loom.no-route.org/powersup.png

PS: Sorry that it takes so long to answer all your answers. I'm quite
busy right now with various projects and work. I will answer asap.
Thanks for all your help!

6581
11-22-2010, 04:28 PM
Looks like some nice PCB software what type are you using? Check that the xtal for the propeller has the correct capacitance. Maybe your switching power supply is causing the problem.

a suggestion:

Take a board and only populate the propeller and eeprom. Run it to see if it is in fact noisy. If not keep adding parts till the noise appears. If it is check input power, if that is okay then PCB may be truly causing the problem.

Most of the time I'm using Eagle Freeware with Eagle3D scripts and a
highly-polished version of MegaPOV. For now I have no time to polish
up my scripts to release them to the public, but it's on my To-Do list
so anybody can make such fancy 3D renderings (or even better).

I'll follow your suggestions and tell you if there are any negative or
positive results. Thanks!

BigFoot
11-22-2010, 04:59 PM
Would this be okay? (No PI-Filter in this circuit):

http://loom.no-route.org/powersup.png

PS: Sorry that it takes so long to answer all your answers. I'm quite
busy right now with various projects and work. I will answer asap.
Thanks for all your help!

That looks good to me, almost identical to our system. The only difference is
we put SMT led's on the power supplies so our service people can see at a
glance if everything is working ok.

On our larger systems we also have a big Electrolytic on the input for bulk
storage.

BigFoot
11-22-2010, 05:20 PM
6581

What cad system do you use to design your boards ? The 3-D model
looks great.

6581
11-22-2010, 05:35 PM
6581

What cad system do you use to design your boards ? The 3-D model
looks great.

Thanks! I've posted some more information on that topic above (http://forums.parallax.com/showpost.php?p=956033&postcount=13).
BTW: What inductor values and capacitor value are you using in your designs?

Update: Here is v2

http://loom.no-route.org/powersup2.png

Sapieha
11-22-2010, 10:50 PM
Hi 6581.

As I think Reg's section is OK. Be only aware that C1 have Correct Voltage (at last 20% bigger as Input Voltage.)

I still need write on Propeller decoupling.
BUT as I not can write it as nice as INTEL on theirs PDF to Quad Processor I will simply copy theirs INFO (Propeller need same decoupling as it has same Voltage variations) In time it runs.

" 2.2.1 VCC Decoupling
VCC regulator solutions need to provide sufficient decoupling capacitance to satisfy the
processor voltage specifications. This includes bulk capacitance with low effective series
resistance (ESR) to keep the voltage rail within specifications during large swings in
load current. In addition, ceramic decoupling capacitors are required to filter high
frequency content generated by the front side bus and processor activity. Consult the
Voltage Regulator-Down (VRD) 11.0 Processor Power Delivery Design Guidelines For
Desktop LGA775 Socket."




Thanks! I've posted some more information on that topic above (http://forums.parallax.com/showpost.php?p=956033&postcount=13).
BTW: What inductor values and capacitor value are you using in your designs?

Update: Here is v2

http://loom.no-route.org/powersup2.png

Tubular
11-23-2010, 12:40 AM
6581,

Just to be clear, are you talking about noise getting into the audio channels (ie <20kHz), or radiated EMI emissions (100kHz and upwards)?

6581
11-23-2010, 01:17 AM
6581,

Just to be clear, are you talking about noise getting into the audio channels (ie <20kHz), or radiated EMI emissions (100kHz and upwards)?

EMI emissions :)