Re: Experimentonomen quits class d
Steve Conner, Tue May 28 2013, 02:13PM

Looks cool! What chip are you using for the comparator?

Re: Experimentonomen quits class d
Experimentonomen, Tue May 28 2013, 04:15PM

MAX913, but LT1016 and AD8561 will drop right in with no changes, downside is that LT1016 is ridicolously expensive, so when i build more of these i will have to go with AD8561 as it is the cheapest as well as the only one i can get in PDIP at a reasonable price(around 4 euros).

Re: Experimentonomen quits class d
Experimentonomen, Sat Jun 22 2013, 10:17AM

I got a registered letter from seeedstudio:


Schematic:

Re: Experimentonomen quits class d
Experimentonomen, Sun Jun 23 2013, 11:08AM

All videos and pictures of board layouts deleted after discovering my class d make as good am transmitters as amplifiers.

All my class d work is now terminated until i have learned more about proper layouting above 20kHz.

Re: Experimentonomen quits class d
Steve Conner, Sun Jun 23 2013, 07:11PM

Boo :(, but on the bright side you are learning more about EMC every day! Did you get complaints from the neighbours?

Re: Experimentonomen quits class d
Dr. Dark Current, Sun Jun 23 2013, 08:44PM

Just put it in a shielded case with common mode chokes and dont care

Re: Experimentonomen quits class d
Experimentonomen, Sun Jun 23 2013, 11:42PM

UNACCEPTABLE!!! Building in shielded cans is just sweeping the problem under the carpet.

If class d modules from Hypex and Anaview/Abletec can be run without shielded cans in the same box as a AM/FM tuner with no interference what so ever, i want my diy to have the same ability.

Therefore i put class d on the shelf until i come up with a board layout that fullfills this requirement. Until then, no more class d will be produced.

Re: Experimentonomen quits class d
Steve Conner, Mon Jun 24 2013, 07:28AM

The Coldamp module I had used a few tricks: A second stage of filtering on the output, a RC snubber on the switching node to suppress ringing, and it was shipped with some small ferrite chokes to be inserted in the power supply wires if you had any trouble. Using it as a bass guitar amp, I mostly had trouble with switching noise getting back into the preamp input, causing gross-sounding distortion. I had to limit the bandwidth of the preamp quite a bit.

RC or LC filters will only be effective if the filter capacitor is grounded to a low-impedance ground. Putting the whole circuit in a metal box and grounding the EMI filter caps directly to the box where the wires pass through is a brute force technique, pretty much guaranteed to work. The box forms a Faraday cage and all the return currents from the caps are confined to the inside.

I would be very surprised if any Class-D amp module could be used in the same box as an AM tuner without some pretty epic levels of shielding and filtering. Or nowadays a clever noise shaping algorithm connected to the tuner, that moves the noise energy away from whatever station you're listening to

Re: Experimentonomen quits class d
Experimentonomen, Mon Jun 24 2013, 08:59AM

My boards did use a ground plane.

I took these pics for a troubleshooting thread at diyaudio:


And i stayed up late last night and drew this, but im gonna assume its just as terrible as the previous one, if not worse.
Worse or better ? you decide:

Re: Experimentonomen quits class d
Steve Conner, Mon Jun 24 2013, 10:43AM

Hard to say if it's worse or better than the previous one, since you deleted the previous one and I can't remember what it looked like.

Putting all the terminals together on one side is good, it reduces the common-mode ground impedance.

To get the full benefit of this low ground impedance, add some filters to the power rails and speaker output, placed right next to the terminals, with their grounds close together. I like the little 3-terminal disc capacitors with built-in ferrite beads, for example the DSS6 series from Murata. Surface mount versions are available with even lower stray inductance.

If you don't like the thought of a ceramic disc cap on the speaker output, you could make a discrete LC filter with a film capacitor.

Re: Experimentonomen quits class d
Experimentonomen, Mon Jun 24 2013, 11:15AM

Something like this ?

Re: Experimentonomen quits class d
Experimentonomen, Mon Jun 24 2013, 12:54PM

Added more caps everywhere:

Re: Experimentonomen quits class d
Steve Conner, Mon Jun 24 2013, 01:02PM

Try to get the ground terminals of your little SMT capacitors closer to the ground spade terminals (X1 and X5). They are on a "spur" of groundplane that will have a fair amount of inductance. Maybe add some groundplane to the top layer and stitch it down to the caps with vias.

Also consider adding a second inductor to the speaker output line, to help the filter cap do its job. A ferrite bead like L2 and L3 would do. If you're worried about distortion from the ferrite, maybe even use an air cored inductor of a few turns. You may have to relocate the feedback tap before this extra inductor, or it may work fine since the corner frequency of the second filter will be so much higher than the first one, it probably won't add much excess phase to the loop.

Finally I suggest making room for a series RC snubber from the switching node to ground. This will help suppress the VHF ringing you usually see here, which is probably what clobbers FM reception. You don't have to populate it. Use a SMT capacitor and 1210 size SMT resistor to minimise inductance.

Re: Experimentonomen quits class d
Experimentonomen, Mon Jun 24 2013, 01:32PM

I had no problems what so ever on the FM band , all stations i can get here normally showed up unchanged with the amp modules running right next to the tuner.

My problems are on the AM band at second and third harmonic of the switching frequency(around 600-700kHz and 1-1.2MHz) The tuner does not see the fundamental at 300-400kHz, only the first few harmonics which show up as full strenght signal AM stations filled with loud buzz and noise and slight very distorted hint of the music played.

This within a one meter radius, after that the signal meter on the tuner drops to zero and goes quiet within a few centimeters.

Re: Experimentonomen quits class d
Experimentonomen, Mon Jun 24 2013, 02:03PM

Redid the caps at the power inputs alittle:

Re: Experimentonomen quits class d
Steve Conner, Mon Jun 24 2013, 02:58PM

Looks good

Re: Experimentonomen quits class d
Experimentonomen, Mon Jun 24 2013, 03:19PM

Trying alittle more:

But im fairly sure this will not be any better than the previous version that interfered with AM reception.

TBH im not sure im actually willing to sink any more money into further boards and components only to find out this board produces as much AM garbage as the previous version, however i have a hunch that maybe its the fact that i ran the power traces on eather side of the board that could maybe be a reason for the EMI that that could possibly maybe create some form of a antenna that having the power traces run alongside each other on the same side of the board doesent.

I wish i had any kinda clue what a low impedance ground plane would look like.

Re: Experimentonomen quits class d
Steve Conner, Mon Jun 24 2013, 03:33PM

A low impedance ground plane looks like an unbroken sea of copper with as few gaps in it for tracking as possible.

AM radio really is a hard test to pass. Even if your PCB layout was perfect, the inductor could be radiating enough leakage flux for an AM radio to pick up at 1m. Mounting the module inside a metal box would probably help even without changing the PCB.

Running both power traces round the same side would certainly help.

Re: Experimentonomen quits class d
Experimentonomen, Mon Jun 24 2013, 03:45PM

I think im gonna take all of the yet unpopulated boards, pack them in a cardboard box with some crumpled up news paper, fo out somewhere where theres no ppl around and burn the crap.

The boards for the all discrete version arrived today but as they have the power traces running on either side of the board, they will surely be the same transmitters as the ic based version i have the EMI problems with, so im not even gonna bother wasting components to try one out.

Re: Experimentonomen quits class d
Experimentonomen, Mon Jun 24 2013, 05:43PM

I found a highly interesting video about EMC:

And after watching said video, this is my conclusion of the problems i have:

Re: Experimentonomen quits class d
Avalanche, Mon Jun 24 2013, 10:09PM


don't burn the boards, use them! The whole point of development work is to end up with a load of heavily modded and chopped up boards! In the process, you'll learn loads about good layout technique, and get the chance to experiment with different components and configurations hands on - rather than trying to figure out everything as you're laying out the board.

If I were you, grab a board and populate it then cut up the grounding and re-arrange it with copper tape, bits of wire, whatever it takes, to minimise your loop areas. Take the chance to investigate what's radiating and why, try different inductors/filtering etc and make gradual improvements. Once you have a bodged to hell board that works, you can think about putting those changes onto a new PCB layout.

If you want some inspiration, I can post some amusing pics of boards from work - of some real commercial products that have gone this way. First, second, boards always bring up new problems - my first one had 5v shorted to ground

Re: Experimentonomen quits class d
Experimentonomen, Mon Jun 24 2013, 10:35PM

I guess i could take one of the modules already assambled, cut up one of the rail traces and bodgewire it up similar to the latest layout and see if theres any change what so ever.

Latest layout:

Re: Experimentonomen quits class d
Steve Conner, Tue Jun 25 2013, 11:31AM

Yeah do that, you can easily reroute traces with pieces of wire, tack on ceramic caps and so on.

Re: Experimentonomen quits class d
Experimentonomen, Tue Jun 25 2013, 04:46PM

I did an Avalanche:

Re: Experimentonomen quits class d
Experimentonomen, Tue Jun 25 2013, 05:52PM

And just as expected, it did NOTHING, module is still aAM transmitter!! :(

Re: Experimentonomen quits class d
Avalanche, Tue Jun 25 2013, 10:14PM

Is there any improvement, or is it worse? and by how much?

This sort of thing won't be a case of suddenly 'solved it', I think before getting too stuck in you need to do two things - first establish how you are going to 'measure' the unwanted emissions as you're making changes on the board (could just be, say, a radio 2 meters away for making comparisons). Secondly, set yourself a realistic goal for what you're going to accept as final.

You won't get the emissions down to zero when you're only a couple of harmonics away from fundamental, on something that is switching, and especially without any kind of EMC enclosure. You can only minimise it and know where to stop!

I'm definitely no expert on EMC, but I've been lurking somewhere in the background on a couple of totally separate occasions now, where there was a lot of last minute panic to try and avoid paying for yet another day of EMC consultancy! I'll try and dig out a document someone at work gave me with some good board layout tips for EMC.

Stick at it

Re: Experimentonomen quits class d
Conundrum, Wed Jun 26 2013, 05:21AM

Put the boards in a picture frame and sell on Ebay as "Electronics Art" with a certificate of authenticity.

Re: Experimentonomen quits class d
Avalanche, Wed Jun 26 2013, 07:26AM

here's the document on layout tips
]pcb_design_guidelines_for_reduced_emi.pdf[/file]

Re: Experimentonomen quits class d
Experimentonomen, Wed Jun 26 2013, 05:45PM

Modded the layout slightly moore:

Re: Experimentonomen quits class d
Experimentonomen, Sun Jun 30 2013, 10:02AM

Doing this im sure would have no effect at all on my emi prollem as bringing the rail traces together did nothing.