Electronics > Projects, Designs, and Technical Stuff
EMF pickup from amplifier in I2C line causing glitches. (Now with scope trace!)
dmills:
--- Quote from: Starlord on July 04, 2016, 03:28:52 am ---And just how strong do you think a signal from a device drawing only 2A at 12V is going to be? I admit, I'm no expert on analog, but I'm pretty sure electromagnetic radiation is electromagnetic radiation, so I think it's safe to say that one can visualize how powerful a transmitter is by envisioning how bright a lightbulb would be when driven with that same current.
--- End quote ---
Well I can routinely send radio messages around the world at the sort of frequencies that thing will be radiating on with less then 10W of input power, but the real problem is that you might be stomping on stuff within a few miles, not that you might be stomping on weak signal comms halfway around the planet, if you cause interference to an FCC licensed service, folks will come looking, and they will find your customer, what does that do for your reputation?
Light bulbs are typically < 10% efficient, and light does not really bounce off the ionosphere, radio kind of does.
Class D power amps are a wonderful thing, but you do by god have to be careful about radiated emissions, especially when a speaker cable is in play (as opposed to a speaker mounted within inches of the board), it is unlikely that ferrite beads alone will get the job done, you probably need an L/C network on each output (Maybe even a couple of poles).
Yes, you can do filterless class D, if the amp is right on the back of the speaker and if the whole dogs dinner is in a metallic enclosure, and if your layout is sufficiently tight, otherwise, just don't go there (Your problem is orders of magnitude worse because of the external cable).
I would seriously look at a few hundred ohm pullups on that I2C link, your edges are horrible at the moment, and I would probably look at SPI instead even if I could not go for something intended for this kind of thing (485/CAN/Whatever), it has the virtue of active drive in both states, so is somewhat better behaved.
A 7 foot cable will be a a quarter wave resonator at 35MHz, a halfwave in the 4M band, and a full wavelength somewhere around the 2M ham band, if it radiates they will hunt your users down.
You might want to look at what the fines the FCC levied a few years ago on musical equipment makers who failed to play by the rules were like, tens of thousands of dollars and up, and they can come after US companies more easily then Chinese ones....
Regards, Dan.
Starlord:
--- Quote from: dmills on July 04, 2016, 12:42:16 pm ---Class D power amps are a wonderful thing, but you do by god have to be careful about radiated emissions, especially when a speaker cable is in play (as opposed to a speaker mounted within inches of the board), it is unlikely that ferrite beads alone will get the job done, you probably need an L/C network on each output (Maybe even a couple of poles).
Yes, you can do filterless class D, if the amp is right on the back of the speaker and if the whole dogs dinner is in a metallic enclosure, and if your layout is sufficiently tight, otherwise, just don't go there (Your problem is orders of magnitude worse because of the external cable).
--- End quote ---
http://[url=http://www.ti.com/lit/ds/symlink/tpa3131d2.pdf
--- Quote ---7.3.14 When to Use an Output Filter for EMI Suppression
The TPA313xD2 has been tested with a simple ferrite bead filter for a variety of applications including long
speaker wires up to 125 cm and high power. The TPA313xD2 EVM passes FCC class-B specifications under
these conditions using twisted speaker wires. The size and type of ferrite bead can be selected to meet
application requirements. Also, the filter capacitor can be increased if necessary with some impact on efficiency.
There may be a few circuit instances where it is necessary to add a complete LC reconstruction filter. These
circumstances might occur if there are nearby circuits which are sensitive to noise. In these cases a classic
second order Butterworth filter similar to those shown in the figures below can be used.
--- End quote ---
--- Quote ---10.1 Layout Guidelines
The TPA313xD2 can be used with a small, inexpensive ferrite bead output filter for most applications.
--- End quote ---
I don't know how one could interpret that to mean anything other than that this amplifier is designed not to need an LC filter in most applications involving speaker wires which are less than 4' long. Mine are less than 2' long.
Btw, without the ferrite bead filter, issue goes away with a 10' speaker cable that has been coiled up so it's 2' long. I haven't checked with the scope to see how the signal is affected, but I assume this is because the emissions are reduced somewhat. I'm not sure if it's the length or the coiling or both which is reducing the emissions though.
--- Quote ---I would seriously look at a few hundred ohm pullups on that I2C link, your edges are horrible at the moment, and I would probably look at SPI instead even if I could not go for something intended for this kind of thing (485/CAN/Whatever), it has the virtue of active drive in both states, so is somewhat better behaved.
--- End quote ---
I would seriously look at it as well, except I can't tell from the MCP3021 ADC datasheet if it can sink 20mA:
http://ww1.microchip.com/downloads/en/DeviceDoc/21805B.pdf
As for using another kind of bus, I don't know why people keep suggesting CAN busses and the like. Is there any way making that switch would not result in having to almost completely redesign the circuit and rewrite all the code? There are several I2C chips at the end of this cable. I don't know if they even make LED drivers that communicate over a CAN bus, and I don't know if they make a CAN to I2C converter. Even if the latter is true, learning about how the CAN bus works and implementing such a solution would likely take weeks, and there's a high probability I'll screw something up which would then add another 3-4 weeks to get a new prototype made.
What's wrong with a chip like this?
http://www.nxp.com/documents/data_sheet/PCA9615.pdf
Or one of these?
http://www.ti.com/lit/ds/symlink/p82b96.pdf
http://www.ti.com/lit/ds/symlink/p82b715.pdf
That last one seems like the best option, as you can just stick it inline with the existing cable and use the same number of data lines as before. In my case, if I had to give up two additional lines on the Cat5 cable, I'd have to choose between reduced functionality, or getting rid of one pair of 5V/GND lines, which I added in the first place to reduce voltage sag and I'd have to go over everything again to figure out where too much sag might cause issues.
--- Quote ---A 7 foot cable will be a a quarter wave resonator at 35MHz, a halfwave in the 4M band, and a full wavelength somewhere around the 2M ham band, if it radiates they will hunt your users down.
--- End quote ---
Well the I2C communications are 400KHz, and the speaker cable isn't 7' long. I don't know if the cable can re-radiate the noise from the speaker, but I already mentioned I would be dealing with that.
--- Quote ---You might want to look at what the fines the FCC levied a few years ago on musical equipment makers who failed to play by the rules were like, tens of thousands of dollars and up, and they can come after US companies more easily then Chinese ones....
--- End quote ---
I'm sure those were big companies selling tens of thousands of units. And they may not be able to go after Chinese companies as easily, but they can't tell the thing emitting is made in China unless they examine it. And when have you ever heard of them knocking on a random citizen's door and questioning them about their hoverboard?
Also, there are 318 million people in the US, and my devices will be in the hands of a couple hundred of them, if I'm lucky. I have a better chance of winning the lottery than having the FCC show up on my doorstep over a device that will be used 3 or 4 times a year for a few hours by the end user.
Anyway, I said I'll be adding the ferrite bead filter to the amp. And there's not much I can do about the I2C if you think that 400KHz data at extremely low power is an issue. Do you think that's an issue? It wasn't clear. I was considering making a switch to shielded cable at some point, but only if I was still getting glitches. Surely 3mA a 5V is not going to make a powerful enough transmitter to catch anyone's attention.
tautech:
--- Quote from: Starlord on July 04, 2016, 09:29:13 pm ---
--- Quote from: dmills on July 04, 2016, 12:42:16 pm ---Class D power amps are a wonderful thing, but you do by god have to be careful about radiated emissions, especially when a speaker cable is in play (as opposed to a speaker mounted within inches of the board), it is unlikely that ferrite beads alone will get the job done, you probably need an L/C network on each output (Maybe even a couple of poles).
Yes, you can do filterless class D, if the amp is right on the back of the speaker and if the whole dogs dinner is in a metallic enclosure, and if your layout is sufficiently tight, otherwise, just don't go there (Your problem is orders of magnitude worse because of the external cable).
--- End quote ---
[url]http://www.ti.com/lit/ds/symlink/tpa3131d2.pdf]http://www.ti.com/lit/ds/symlink/tpa3131d2.pdf]
[url]http://www.ti.com/lit/ds/symlink/tpa3131d2.pdf[/url]
--- Quote ---7.3.14 When to Use an Output Filter for EMI Suppression
The TPA313xD2 has been tested with a simple ferrite bead filter for a variety of applications including long
speaker wires up to 125 cm and high power. The TPA313xD2 EVM passes FCC class-B specifications under
these conditions using twisted speaker wires. The size and type of ferrite bead can be selected to meet
application requirements. Also, the filter capacitor can be increased if necessary with some impact on efficiency.
There may be a few circuit instances where it is necessary to add a complete LC reconstruction filter. These
circumstances might occur if there are nearby circuits which are sensitive to noise. In these cases a classic
second order Butterworth filter similar to those shown in the figures below can be used.
--- End quote ---
--- Quote ---10.1 Layout Guidelines
The TPA313xD2 can be used with a small, inexpensive ferrite bead output filter for most applications.
--- End quote ---
I don't know how one could interpret that to mean anything other than that this amplifier is designed not to need an LC filter in most applications involving speaker wires which are less than 4' long. Mine are less than 2' long.
Btw, without the ferrite bead filter, issue goes away with a 10' speaker cable that has been coiled up so it's 2' long. I haven't checked with the scope to see how the signal is affected, but I assume this is because the emissions are reduced somewhat. I'm not sure if it's the length or the coiling or both which is reducing the emissions though.
--- Quote ---I would seriously look at a few hundred ohm pullups on that I2C link, your edges are horrible at the moment, and I would probably look at SPI instead even if I could not go for something intended for this kind of thing (485/CAN/Whatever), it has the virtue of active drive in both states, so is somewhat better behaved.
--- End quote ---
I would seriously look at it as well, except I can't tell from the MCP3021 ADC datasheet if it can sink 20mA:
http://ww1.microchip.com/downloads/en/DeviceDoc/21805B.pdf
As for using another kind of bus, I don't know why people keep suggesting CAN busses and the like. Is there any way making that switch would not result in having to almost completely redesign the circuit and rewrite all the code? There are several I2C chips at the end of this cable. I don't know if they even make LED drivers that communicate over a CAN bus, and I don't know if they make a CAN to I2C converter. Even if the latter is true, learning about how the CAN bus works and implementing such a solution would likely take weeks, and there's a high probability I'll screw something up which would then add another 3-4 weeks to get a new prototype made.
What's wrong with a chip like this?
http://www.nxp.com/documents/data_sheet/PCA9615.pdf
Or one of these?
http://www.ti.com/lit/ds/symlink/p82b96.pdf
http://www.ti.com/lit/ds/symlink/p82b715.pdf
That last one seems like the best option, as you can just stick it inline with the existing cable and use the same number of data lines as before. In my case, if I had to give up two additional lines on the Cat5 cable, I'd have to choose between reduced functionality, or getting rid of one pair of 5V/GND lines, which I added in the first place to reduce voltage sag and I'd have to go over everything again to figure out where too much sag might cause issues.
--- Quote ---A 7 foot cable will be a a quarter wave resonator at 35MHz, a halfwave in the 4M band, and a full wavelength somewhere around the 2M ham band, if it radiates they will hunt your users down.
--- End quote ---
Well the I2C communications are 400KHz, and the speaker cable isn't 7' long. I don't know if the cable can re-radiate the noise from the speaker, but I already mentioned I would be dealing with that.
--- Quote ---You might want to look at what the fines the FCC levied a few years ago on musical equipment makers who failed to play by the rules were like, tens of thousands of dollars and up, and they can come after US companies more easily then Chinese ones....
--- End quote ---
I'm sure those were big companies selling tens of thousands of units. And they may not be able to go after Chinese companies as easily, but they can't tell the thing emitting is made in China unless they examine it. And when have you ever heard of them knocking on a random citizen's door and questioning them about their hoverboard?
Also, there are 318 million people in the US, and my devices will be in the hands of a couple hundred of them, if I'm lucky. I have a better chance of winning the lottery than having the FCC show up on my doorstep over a device that will be used 3 or 4 times a year for a few hours by the end user.
Anyway, I said I'll be adding the ferrite bead filter to the amp. And there's not much I can do about the I2C if you think that 400KHz data at extremely low power is an issue. Do you think that's an issue? It wasn't clear. I was considering making a switch to shielded cable at some point, but only if I was still getting glitches. Surely 3mA a 5V is not going to make a powerful enough transmitter to catch anyone's attention.
--- End quote ---
Starlord, remove the [ quote ] on the third line. (I think)
Starlord:
Did some more testing. Here's what I found:
With no speaker or speaker cable connected my 5V supply is fairly clean:
[link removed]
With just the speaker cable connected and hanging in the breeze I see noise on the 5V supply:
[link removed]
And if I touch just one lead that is attached to the speaker to the amplifier output, the noise is much worse:
[link removed]
I calculated from those peaks (because I couldn't trust the triggering to report the right frequency) that they're at 1MHz. So what I think I'm seeing here is wideband noise of many frequencies all greater than 1MHz. So I'm thinking the ferrite is indeed the right way to go, but I need help selecting one.
As I mentioned before, this is the ferrite TI used in their evaluation module:
http://www.digikey.com/product-search/en?keywords=EXC-ELDR35C
10 mOhm DC resistance, 90 Ohms @ 100Mhz, 7A current rating.
I narrowed down the selection of SMT ferrite beads to these based on size, current handling, DC resistance, and impedance:
http://www.digikey.com/short/348zc8
I calculated the current handling by assuming two 4 ohm speakers, a 12V supply, and 90% efficiency. That gave me 3A for the worst case, a square wave at max volume. Average case with real world audio is probably more like 1.5A or less.
Not sure where to go from here. Should I select a ferrite with as high an impedance at 100MHz as I can get because I need all the impedance I can get down at 1MHz? (While examining the graphs to see which are really better at those lower frequencies of course.) Will a DC resistance of 2-3x as high as the TI part be okay, but very slightly reduce the speaker volume and increase the heat dissipation in the bead? And how can I tell if it will have a self resonant frequency below 10MHz as specified in the datasheet?
I've also looked at inductors as an option, but they're really too large. At least the 10uH ones specified in the datasheet and I'm not sure how to properly select a different size.
tautech:
You can't explicitly trust what any scope displays unless the trigger level is set within the amplitude of the displayed waveform.
Channel AC input coupling is also useful for viewing ripple/noise on a DC rail at low V/div settings without needing to adjust the offset/vertical position.
Edit
This needs fixing too in reply #36:
http://http://www.ti.com/lit/ds/symlink/tpa3131d2.pdf
Navigation
[0] Message Index
[#] Next page
[*] Previous page
Go to full version