As you all saw, my class d modules do not intefer with a FM tuner, but they do seem to interfer with each other.
each module on its own has little noise but as soon as i bring the second module into operation the noise level increases quite a bit, and it also seem to go up and down, sometimes with a little bit of crackle and hum in it, and sometimes barely even audible from a meter away.
This is why i put on the ferrite thingies, but they had little to no effect, separating the gate drive to a isolated supply for each module reduced the hiss alittle.
It turns out these modules are a perfect full signal AM station!! PCB layout FAIL!
each module on its own has little noise but as soon as i bring the second module into operation the noise level increases quite a bit, and it also seem to go up and down, sometimes with a little bit of crackle and hum in it, and sometimes barely even audible from a meter away.
This is why i put on the ferrite thingies, but they had little to no effect, separating the gate drive to a isolated supply for each module reduced the hiss alittle.
It turns out these modules are a perfect full signal AM station!! PCB layout FAIL!
Last edited:
This is it, this is the end. I am leaving class d... for now.
After discovering my class d modules make for perfect full signal strenght AM transmitters, i am pulling the plug on my class d and related work until i have learned more abounf proper layouting above 20kHz.
I have already gone through my computer and image hosting and deleted all pictures of board layouts, leaving only the spice simulations and pics of the built modules.
The remaining bare boards from this project will be gathered up and destroyed, the modules already built i will keep for now and will eventually be stripped for usable parts.
I urge moderators on this forum to delete all my uploaded picturs of my board layouts to further prevent lesser knowing ppl from copying my work and also end up with unintentional AM transmitters.
After discovering my class d modules make for perfect full signal strenght AM transmitters, i am pulling the plug on my class d and related work until i have learned more abounf proper layouting above 20kHz.
I have already gone through my computer and image hosting and deleted all pictures of board layouts, leaving only the spice simulations and pics of the built modules.
The remaining bare boards from this project will be gathered up and destroyed, the modules already built i will keep for now and will eventually be stripped for usable parts.
I urge moderators on this forum to delete all my uploaded picturs of my board layouts to further prevent lesser knowing ppl from copying my work and also end up with unintentional AM transmitters.
Last edited:
The problem you're describing sounds exactly like switch frequency locking between channels and is both very common and very hard to control. Before you bash your head into the piano keys and destroy everything I'd recommend to fully, individually shield the boards you already have and lead the grounds and supplies all to central points and see what happens. If you don't do that first you'll lose a lot of work and experimental knowledge. There is probably no way to build such a circuit that creates zero radiation, so shielding is the only option. Use output chokes with pot type cores or evenly wound toroids, and keep those under the shields as well, maybe even make sub-shields for them. Interference by conduction then is a matter of proper lead wire layout and dressing. If your boards operate correctly individually, I'd worry less at first about the per-channel circuit layout. Even manufacturer demo boards come with a warning about this trouble.
Last edited:
Well the UcD modules from hypex can and do operate properly right next to each other with none of the issues i have with no shielding what so ever when ran from the same psu just like my setup.
This confirms that my issue is board layout alone, hence my decision.
I still wanna thank everyone for the interest in my adventures in class d, but this is the end... until i have learned more about proper layouting to prevent this issue i discovered.
This confirms that my issue is board layout alone, hence my decision.
I still wanna thank everyone for the interest in my adventures in class d, but this is the end... until i have learned more about proper layouting to prevent this issue i discovered.
Where oh where will you start fixing if you don't find out exactly where your "transmit and receive" nodes are in the boards you already have?
Also, "none" is an impossibly small amount. If I were serious about housing multiple switching amps in one box for audiophile purposes I'd shield them all no matter what.
Also, "none" is an impossibly small amount. If I were serious about housing multiple switching amps in one box for audiophile purposes I'd shield them all no matter what.
Last edited:
Thats the thing, i dont know anywhere near enough enough about layouting to know where to look for the node acting transmitter, but i wouldent be the least suprised if its actually the speaker wires, but then i wonder how the hell do the Hypex UcD, Anaview and all others pass EMC testing if just 5 meters of speaker wire is all it takes to make a class d into a AM transmitter ? Because having a AM receiver pick up the class d all around 600kHz -2MHz(switching fundamental and its 10 harmonics) i'd say would be unacceptable in EMC testing.
Atleast its not ringing based interference as this would be in the 20+MHz region = it wouldent affect AM, but be all over the FM broandcast band, so that part of the layout i seem to have gotten right, only one FM station was affected by the class d, but that station was very weak and near unlistenable to begin with regardless of the class d beeing on or off.
The transmitting note is on the board itself, removing the speaker wires and shorting the audio input near the modules did nothing other than shift the AM reception up from 670kHz to 800kHz.
Atleast its not ringing based interference as this would be in the 20+MHz region = it wouldent affect AM, but be all over the FM broandcast band, so that part of the layout i seem to have gotten right, only one FM station was affected by the class d, but that station was very weak and near unlistenable to begin with regardless of the class d beeing on or off.
The transmitting note is on the board itself, removing the speaker wires and shorting the audio input near the modules did nothing other than shift the AM reception up from 670kHz to 800kHz.
Last edited:
Class D is critical about layout and decoupling.
It took me 3 attempts to get a pcb that didnt reset all the time.
The class d IC needs decoupling very closely.
The output mosfets need decoupling closely too , from b+/b- b+/gnd and b-/gnd.
PCB tracks should be as short and as thick as possible.
Components should be kept as close together as possible.
I sold a couple of my amplifiers. I had one customer who said the amp kept resetting all the time.
Turned out he had removed the output transistors and mounted them on an external heatsink on 4 inches or wire !!
Of course putting them back on the pcb sorted out his problems.
I have had very little problem with hiss and hum on my pcbs.
Star grounding is also important (as with any amp class) but more important in class d.
It took me 3 attempts to get a pcb that didnt reset all the time.
The class d IC needs decoupling very closely.
The output mosfets need decoupling closely too , from b+/b- b+/gnd and b-/gnd.
PCB tracks should be as short and as thick as possible.
Components should be kept as close together as possible.
I sold a couple of my amplifiers. I had one customer who said the amp kept resetting all the time.
Turned out he had removed the output transistors and mounted them on an external heatsink on 4 inches or wire !!
Of course putting them back on the pcb sorted out his problems.
I have had very little problem with hiss and hum on my pcbs.
Star grounding is also important (as with any amp class) but more important in class d.
Last edited:
It sounds like you know just about everything you need to know to complete your work, although it might take a couple weeks. 
The only thing I know for sure is that scrapping what you have and just starting over would be a bum plan. Step by step troubleshooting technique will get you where you need to go.
The only thing I know for sure is that scrapping what you have and just starting over would be a bum plan. Step by step troubleshooting technique will get you where you need to go.
You need to experiment with things to see what improves things or makes them worse.
You could try a ground sheet of aluminium between the pcb's to see if it is a radiated or conducted emissions problem.
I use to try extra decoupling.
Also the separate pcbs need star grounding from each other if using the same supply.
I have just finished a high power valve amplifier and I found that good earthing was vital to low noise.
You could try a ground sheet of aluminium between the pcb's to see if it is a radiated or conducted emissions problem.
I use to try extra decoupling.
Also the separate pcbs need star grounding from each other if using the same supply.
I have just finished a high power valve amplifier and I found that good earthing was vital to low noise.
I added a 470nF cap directly between the rails as close to the mosfets as practically possible, did nothing to the AM transmission, however i do notice that noise level reduces alot if there is only the power ground wires joining the two modules by leaving the audio grounds separate to each module, perhaps i shouldent be taking audio ground from each board but rather from the main psu caps. Or separated audio ground from the main ground plane by a 10-20 ohm resistor.
However as there is still the AM transmission issue, i no longer care any about the audible noise, it could be nigeria waterfall loud for all i care. What i care about from now on is stopping the modules from transmitting all over the AM band.
However as there is still the AM transmission issue, i no longer care any about the audible noise, it could be nigeria waterfall loud for all i care. What i care about from now on is stopping the modules from transmitting all over the AM band.
The AM transmission is from the output carrier signal.
You need to keep any tracks with this on as short as possible.
If your output is not filtered well then the speaker lead also becomes an aerial !
I would revisit the output filter and see if the component values make sense.
The values will be different for 4 ohm and 8 ohm speakers.
Running 8 ohm speakers on a 4 ohm system (and vice versa) will ruin the filter response.
You need to keep any tracks with this on as short as possible.
If your output is not filtered well then the speaker lead also becomes an aerial !
I would revisit the output filter and see if the component values make sense.
The values will be different for 4 ohm and 8 ohm speakers.
Running 8 ohm speakers on a 4 ohm system (and vice versa) will ruin the filter response.
That is only true for a pre filter feedback class D, not for a post filter feedback topology (UcD)
Thats the whole point of UcD, to make the frequency response load invariant, the trace between the output filter inductor and the fets is just 1 centimeter, give or take a millimeter.
Between output filter inductor and cap its about 2 centimeters.
The output filter inductor is 20µH and the cap is 470nF.
And like i said in earlier posts, the AM reception on a tuner in a one meter radius remains regardless of any speakers beeing hooked up or not, only diff is a slight diff in switching frequency. This to me suggests it has nothing to do with do with speaker wires acting antennas or faulty output filters.
Thats the whole point of UcD, to make the frequency response load invariant, the trace between the output filter inductor and the fets is just 1 centimeter, give or take a millimeter.
Between output filter inductor and cap its about 2 centimeters.
The output filter inductor is 20µH and the cap is 470nF.
And like i said in earlier posts, the AM reception on a tuner in a one meter radius remains regardless of any speakers beeing hooked up or not, only diff is a slight diff in switching frequency. This to me suggests it has nothing to do with do with speaker wires acting antennas or faulty output filters.
Can you make new links in this thread to drawings and pictures of your work that could allow others to help you troubleshoot? Try clipping a scope probe ground lead to the probe tip, turn down the attenuator to max sensitivity or wherever you can get a good relative scale, and wave it around the circuit. See if you can find a hot spot. Also the waveform you see can help you figure out what you're dealing with.
The AM transmitter and the severe frequency locking are probably closely related at this point, so I probably wouldn't just ignore one or the other because it's least irritating or something.
The AM transmitter and the severe frequency locking are probably closely related at this point, so I probably wouldn't just ignore one or the other because it's least irritating or something.
Last edited:
Dear Tekko, you are lot more experienced than me and more technically sound too. but I am just sharing my basic experience. my first design was too noisy and made all kind of transmission. ofcourse the culprit turned out to be layout and the output coil.
I got an application note on IR site. This note gave clear indication which path are noisy and how to control them. sure you must have already read it. this helped me for a noiseless layout. I also realized that if the coil saturates it generates lot of noise. ( wont claim this is this is the exact one.) but when I used EE core with proper calculated for its saturation. The design went perfect.
Shared with you, In case if it helps.
I got an application note on IR site. This note gave clear indication which path are noisy and how to control them. sure you must have already read it. this helped me for a noiseless layout. I also realized that if the coil saturates it generates lot of noise. ( wont claim this is this is the exact one.) but when I used EE core with proper calculated for its saturation. The design went perfect.
Shared with you, In case if it helps.
Also i forgot to mention that the AM receiver is not picking up the fundamental which is in the 300-400kHz range, AM reception starts at around 600kHz and is stronges around 1MHz which suggest second or third harmonic, this implies just like i said earlier that the culprit is not the output filter, but the mosfets themselves.
I know my scope is only 20MHz bandwidth, but the square wave before the outoput filter is near perfect, no sign of ringing. However clipping thwe ground lead to the tip of the probe and touching part of the ground lead to the power traces between the mosfets and the onboard lytics (1000µF 50V) i get about 100mV spikes, this with probe set to 1x and scope set to 50mV/div.
These spikes must be shoot through spikes, but considering how small the heatsink is and that it never gets too hot to touch suggest i have more than enough deadtime,
considering that a UcD180 module with just that blue ally heatspreader gets burning hot in less than a minute, i know this as i had a set of UcD180 modules a few years ago.
These spikes must be shoot through spikes, but considering how small the heatsink is and that it never gets too hot to touch suggest i have more than enough deadtime,
considering that a UcD180 module with just that blue ally heatspreader gets burning hot in less than a minute, i know this as i had a set of UcD180 modules a few years ago.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.