There is a lot of contradictory material out there.
In my experience star grounding is of more relevance to audio signal integrity than EMI hardening. Planes are an essential part of RF proofing circuits in the absence of other conductive structures which can serve as Faraday cages. It's best if they don't have to serve dual-duty as a star ground, but engineering is often about making compromises, so if a ground plane is thought desirable and the number of layers is limited, then components should be laid out to permit line-of-sight passage of return currents, avoiding crossing paths. The best way of achieving this is to explicitly place the star ground before floodfilling the whole area. Sometimes other mechanical considerations, such as the cosmetics of a package and the positioning of controls or connectors make this problematic, in this case the judicious use of slots in the ground plane can help to keep return paths apart.
Slots have dangers of their own, as already noted, they have a propensity to act as antennas, it's sometimes only after more than one experimental layout that even experienced engineers achieve the desired result.
In my experience star grounding is of more relevance to audio signal integrity than EMI hardening. Planes are an essential part of RF proofing circuits in the absence of other conductive structures which can serve as Faraday cages. It's best if they don't have to serve dual-duty as a star ground, but engineering is often about making compromises, so if a ground plane is thought desirable and the number of layers is limited, then components should be laid out to permit line-of-sight passage of return currents, avoiding crossing paths. The best way of achieving this is to explicitly place the star ground before floodfilling the whole area. Sometimes other mechanical considerations, such as the cosmetics of a package and the positioning of controls or connectors make this problematic, in this case the judicious use of slots in the ground plane can help to keep return paths apart.
Slots have dangers of their own, as already noted, they have a propensity to act as antennas, it's sometimes only after more than one experimental layout that even experienced engineers achieve the desired result.
Audio Systems Group, Inc. Publications
in particular: http://www.audiosystemsgroup.com/AESPaperFerritesASGWeb.pdf
suggests right material, dimension tubular beads over I/O cables at your amp can be an important part of a solution
in particular: http://www.audiosystemsgroup.com/AESPaperFerritesASGWeb.pdf
suggests right material, dimension tubular beads over I/O cables at your amp can be an important part of a solution
of corse, 1)sensitive circuits inside phone have soldered cages aroundHey, don't forget the cellphone already has DAC and tiny headphone amp INSIDE, which isn't prone at all to the EMI agressor located SO CLOSE.
It's just the layot bug somewhere.
Get ferrite clamps/rings and put 'em on all the wires...
2)gsm antenna radiates in narrow field-not affecting audio circuits
Thanks everyone. Lots of good reading material being posted. I've got quite a bit of reading to do but I love it. Most of the time I'm pretty bored laying out traces and etching boards. While it's annoying when I'm trying to listen to music and get a call at least I'm not bored with this project.
Now if I only had Mouser and Digi-Key next door I place weekly orders and it takes 2-3 days to get components. I'm like a little kid opening Christmas presents mid week. Next order I suspect will consist of quite a bit of filtering components.
I have some new Linear chips to play with (LT3579). Beefing up the DC/DC converter on the amp a bit. My current supply based on LTC3122 can manage 300mA +/- 12v sustained which is fine for only driving headphones but I want to change It out for a new design that can do 800mA. They draw comparable idle current and take up roughly the same board space so why not It will give me some headroom later to beef up the audio output for multiple headphones or maybe even small bookshelf type speakers.
I also have some AD8397ARDZ making their way here. Going to replace the NJM4556AD I'm currently using.
Now if I only had Mouser and Digi-Key next door
I place weekly orders and it takes 2-3 days to get components. I'm like a little kid opening Christmas presents mid week. Next order I suspect will consist of quite a bit of filtering components.I have some new Linear chips to play with (LT3579). Beefing up the DC/DC converter on the amp a bit. My current supply based on LTC3122 can manage 300mA +/- 12v sustained which is fine for only driving headphones but I want to change It out for a new design that can do 800mA. They draw comparable idle current and take up roughly the same board space so why not
It will give me some headroom later to beef up the audio output for multiple headphones or maybe even small bookshelf type speakers.I also have some AD8397ARDZ making their way here. Going to replace the NJM4556AD I'm currently using.
Regarding antennas in your circuit, don't just think length. Think loops that are not tightly closed up.
Also, you are in or almost in "the near field" of the phone's antenna, which means a large magnetic field component, still. And most metal doesn't shield against magnetic fields. Aluminum, especially, will basically not shield them at all. It would need to be a ferrous (iron-based) metal, such as steel. (Also, be aware that many or most modern cell phones emit 100 to 1000 times more power on the side that would be AWAY from your head, if your were using the phone. i.e. much more power will radiate from the rear of the phone than from the front.)
A time-varying magnetic field will induce a corresponding current in any conductive loop. With all else remaining equal, the amplitude of the induced current will be proportional to the geometric area enclosed by the loop. (This is one of the most fundamental electromagnetic phenomena. It is basically one of the four "Maxwell's Equations", which describe all things electromagnetic. It is also known as Faraday's Law.)
Any "natural pair" of conductors probably forms such a loop. e.g. signal/ground, power/ground, output/ground, +/-, etc, etc.
The fact that pulling the opamp out of the circuit stops it might mean that: a) There is a loop that gets a current induced, then the current flows through some impedance and forms a voltage across it, and then that voltage is present at one of the opamp's pins, perhaps a high-gain input, or, b) the loop itself runs through the opamp and pulling the opamp breaks the loop.
One thing you could try would be to compare the noise with a) the inputs both shorted, to b) when they're both open (not connected to anything). Actually, in addition to that, try them with c) 50 Ohms across each one, too. if the noise is worse with them shorted or with 50 Ohms across them than it is when they're open, then they are probably acting as magnetic loop antennas.
At some point, think mainly about where the current goes. What loops does it follow? Do they enclose any area? For example, the current comes from the power source, through conductors, across and through capacitors, into the chip. Some may go back to the source through the ground, from the chip or through capacitors. Those are loops. Much of it might go to the load, then back to some ground and then back to the power source. That, too, completes a loop.
Imagining for a moment that the current in those loops only goes in one direction, ask yourself, is the outgoing current always physically right next to the returning current?? If not, that's a problem. The same type of analysis must also be applied to the input signal and input signal ground reference conductors for each channel. And those are often especially important, since they do feed a high-gain amplifier, so even relatively-small induced signals might be significant.
However, with RF, especially at these relatively-high frequencies, "everything is an input". So you should have any actual wire pairs tightly twisted together, or twisted and shielded. Use at least 4 turns per inch, but more if you can.
I am not sure if anyone mentioned it, but there is some basic material about this in Chapter 7 of the free online Walt Jung book at Analog Devices, "Op Amp Applications Handbook", which is at: ADI - Analog Dialogue | Op Amp Applications Handbook .
RF can be quite insidious. The simple case of demodulating AM signals with a diode (or a PN junction in a chip) and some shunt capacitance is just the tip of the iceberg. Most RF is not amplitude modulated, these days. And every PN junction could rectify RF. Very worrisome, to me, is the possibility that the rectified RF will simply become DC-like and will alter some DC Operating Point(s), deep inside of an IC somewhere, in a time-varying manner, and in such a way that it's not extremely obvious.
And since the RF environment varies tremendoulsy depending on location and time and type of RF sources present, users in some locations could experience problems the designer has never imagined. So, unless a device is to be tested the way companies like RCA did it back in the good old days (very thoroughly), then we should be putting in good RF protection, everywhere we know it might be needed, when possible.
Also, you are in or almost in "the near field" of the phone's antenna, which means a large magnetic field component, still. And most metal doesn't shield against magnetic fields. Aluminum, especially, will basically not shield them at all. It would need to be a ferrous (iron-based) metal, such as steel. (Also, be aware that many or most modern cell phones emit 100 to 1000 times more power on the side that would be AWAY from your head, if your were using the phone. i.e. much more power will radiate from the rear of the phone than from the front.)
A time-varying magnetic field will induce a corresponding current in any conductive loop. With all else remaining equal, the amplitude of the induced current will be proportional to the geometric area enclosed by the loop. (This is one of the most fundamental electromagnetic phenomena. It is basically one of the four "Maxwell's Equations", which describe all things electromagnetic. It is also known as Faraday's Law.)
Any "natural pair" of conductors probably forms such a loop. e.g. signal/ground, power/ground, output/ground, +/-, etc, etc.
The fact that pulling the opamp out of the circuit stops it might mean that: a) There is a loop that gets a current induced, then the current flows through some impedance and forms a voltage across it, and then that voltage is present at one of the opamp's pins, perhaps a high-gain input, or, b) the loop itself runs through the opamp and pulling the opamp breaks the loop.
One thing you could try would be to compare the noise with a) the inputs both shorted, to b) when they're both open (not connected to anything). Actually, in addition to that, try them with c) 50 Ohms across each one, too. if the noise is worse with them shorted or with 50 Ohms across them than it is when they're open, then they are probably acting as magnetic loop antennas.
At some point, think mainly about where the current goes. What loops does it follow? Do they enclose any area? For example, the current comes from the power source, through conductors, across and through capacitors, into the chip. Some may go back to the source through the ground, from the chip or through capacitors. Those are loops. Much of it might go to the load, then back to some ground and then back to the power source. That, too, completes a loop.
Imagining for a moment that the current in those loops only goes in one direction, ask yourself, is the outgoing current always physically right next to the returning current?? If not, that's a problem. The same type of analysis must also be applied to the input signal and input signal ground reference conductors for each channel. And those are often especially important, since they do feed a high-gain amplifier, so even relatively-small induced signals might be significant.
However, with RF, especially at these relatively-high frequencies, "everything is an input". So you should have any actual wire pairs tightly twisted together, or twisted and shielded. Use at least 4 turns per inch, but more if you can.
I am not sure if anyone mentioned it, but there is some basic material about this in Chapter 7 of the free online Walt Jung book at Analog Devices, "Op Amp Applications Handbook", which is at: ADI - Analog Dialogue | Op Amp Applications Handbook .
RF can be quite insidious. The simple case of demodulating AM signals with a diode (or a PN junction in a chip) and some shunt capacitance is just the tip of the iceberg. Most RF is not amplitude modulated, these days. And every PN junction could rectify RF. Very worrisome, to me, is the possibility that the rectified RF will simply become DC-like and will alter some DC Operating Point(s), deep inside of an IC somewhere, in a time-varying manner, and in such a way that it's not extremely obvious.
And since the RF environment varies tremendoulsy depending on location and time and type of RF sources present, users in some locations could experience problems the designer has never imagined. So, unless a device is to be tested the way companies like RCA did it back in the good old days (very thoroughly), then we should be putting in good RF protection, everywhere we know it might be needed, when possible.
Last edited:
Yeah I had considered that the noise may be magnetically induced vs RF but if I fold a thin sheet of aluminum foil around the phone the noise stops instantly.
Tried this already and the results were the same
I've gone through each pin individually bending them out and resocketing to disconnect from the circuit. The output pins + anything else (1 or both power pins, - and/or + input pins etc.) will pass the noise through to the output jack. Seems as long at the op amp is in the socket and connected to pretty much anything else on the circuit the noise gets passed through to the output jack. Obviously I can't do these tests with headphones since I'm not applying power to the chip so I've got it connected to the computer's mic input with gains maxed.
The cable connecting the phone to the headphone amp is also acting as an antenna.. everything I have plugged in really so I've started by moving the circuit and headphone cables away from the phone and leaving the input jack plugged in then filtering the input lines. I've had some pretty good luck cleaning up that loop so far.
I lifted the input ground and placed a ferrite bead there. I already had ferrites on the left and right signal inputs. I placed a 10pf cap to ground right after the signal ferrites and then another 150pf cap to ground right at the op amp + input pins and...... it's actually pretty good on the input now.
I can still hear it if no music is playing but it's barely audible where without any of the ferrites or HF filter caps it was overpowering the music. now I'm going to disconnect the input cable from the phone and amp and lay the headphone cable across the back of the phone and do the same till I have it cleaned up.
Once I have the in and out cleaned up I'll check the noise levels when holding the phone near the board again. At least at that point if it's still quite loud I'll know the majority of it is not originating at the in and out.
I'm also going to place an order for some of the X2Y parts tonight and see how they perform. According to the info I read from the links earlier a single one lowers noise DB by more than every component type I'm using now put together.
Divide and conquer.
Yeah I'll be doing a new layout and test board when I get them. I was thinking of just creating a straight 1/2" wide ground bus on the bottom of the board and arrange the component side layout to have all grounds connect there using vias. That way I'll have pretty open access to the ground bus In the area around the chip pins to ground the X2Y components. I was considering doing another area of copper fill and ground that to the chassis along the entire length of the board on both sides and have it connect to the same end of the ground bus the power ground goes to.
Example of how I will probably arrange the central ground bus. The fill on the component side only flows around the traces but doesn't electrically connect except at the bottom common ground point. This way I can easily modify it for the new X2Y parts without moving much.
An externally hosted image should be here but it was not working when we last tested it.
Last edited:
The latest board layout I just posted shouldn't have any current at all riding on the plane since no components carrying power connect to it.. I figure it will ground to the metal casing to provide a ground return for shielding. all of the current from the amplifier circuit would flow only through that red bus in the middle. The only connecting point for the ground bus and copper fill is at the very bottom middle via. that's where the power ground solders in. the connections to the left and right are -V and +V respectively.
A lot of people swear by the star ground setup and while I can see the reasoning behind it. specially for high current circuits where the power draw can create some pretty sizeable voltage drops along a length of copper foil. For something like a headphone amp and the massive amount of current that ground bus can handle relative to what will be flowing on it. I don't see the voltage difference from any point to the next to be anything significant.
This time around I'm aiming for consolidating my ground paths for the signal components to try and keep my cell phone from making an antenna out of my ground foil
Last edited:
they use ground planes in the LHC detectors... ATLAS, CMS. you can check the Atlas photo archive, they have some examples of the PCBs, in fact there is a remarkably open sharing of the design, perhaps that was part of their directive. you just use thick ground planes, star grounding doesnt protect you from voltage drop across the trace. the frequencies we are talking about dont really care too much about having continuous copper to travel on
Last edited:
Here is a picture showing the connection points.
RED = Single common ground point for power, ground bus for audio signals and the copper fill on the component layer. Maybe half star?
GREEN = contact area for aluminum enclosure. The board slides in to pcb rails.
YELLOW = +12v in the actual amp. +3.7v on the test board I'm troubleshooting my RF problems on.
PURPLE = -12v in actual amp. -3.7v on test board.
RED = Single common ground point for power, ground bus for audio signals and the copper fill on the component layer. Maybe half star?
GREEN = contact area for aluminum enclosure. The board slides in to pcb rails.
YELLOW = +12v in the actual amp. +3.7v on the test board I'm troubleshooting my RF problems on.
PURPLE = -12v in actual amp. -3.7v on test board.
An externally hosted image should be here but it was not working when we last tested it.
Any cable long enough to be an antenna will carry RF power in the presence of an RF aggressor (cellphone). This means the shields of audio ins/outs and the power cable.
I swear by star grounding for the SQ but IME it always makes cellphone pickup worse, not better. Because with star grounding the loop areas get bigger
<edit> What do the six (three pairs) of holes top centre of the board connect to?
Last edited:
Yeah I've had a harsh wake up call from this portable amp project. I fired it up. everything worked. sounds great... then "bzzt bzzt bzztt.. burr.. pop bzzzzt tick tick bzzzzz" pretty loud in my ears.
Before I etch the new layout I'm trying to find the right mix of filters to clean up the signals riding in on my input and headphone cables. Unless I take care of that I don't think the internal layout will matter much.
The entire power delivery system also gets closed in the aluminum case. 3.7v li-po cell. battery charger and power selector, soft on/off controller and low voltage protect for the li-po and the +/- 12v DC/DC converter that powers the Op amp. The only thing extending outside of the shell will be the tip of the volume pot shaft and the input/output cables.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
This is the beast that's attempting to deafen me every time the phone rings.
Everything has to fit in a 0.83" x 0.83" x 3.1" internal volume
Ah thanks - so long as the pot's not on flying leads that should be fine. So I must've missed something here - how is the audio getting on and off the board?
To the left and right of the volume pot are the pads for the headphone jacks. Foxconn hole through type
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.