Wow thanks! Lots of nice tools in that. Will save me tons of time later scouring the web for online calcs. Where's the ohm's law calc though?
should have been the very first tab I posted the layout mainly to try and drum up some ideas on why my previously posted test board layout for the other amp and this one would be so different. They aren't extremely different in trace lengths or ground plane fill dimensions and I actually have some filtering in place on the test board because that amp has RFI problems. This smaller amp is pretty quiet now with this layout without any extra suppression.
I would lean towards the maxim headphone amp IC having built in RF suppression but they don't advertise it on this chip and previous layouts using the same chip have resulted in the same type of interference. It was never quite as strong as the signals picked up by the larger NJM4556AD based amp but they were still very noticeable even with music playing. With the layout I just posed I can barely hear the noise with no music playing. Electrically identical to the previous layout I just moved the components around to fit the new connector arrangement.
Just a small update for anyone curious how I'm progressing with this issue.
I'm still messing around with layouts and components. I've managed to get the noise down to something much more manageable but far from ideal by shrinking the entire test PCB to 0.9" x 1.2" (posted below).
Nothing I've tried, Ferrite beads, X2Y caps on all signal paths, power supply decoupling etc. have made any meaningful reduction in the noise except a sharp reduction in ground plane surface area.
That no amount of low pass filtering I've tried made a dent in the noise combined with the fact that the designs have been very well behaved in the presence of other strong RF signals brings me back to a earlier suspicion that the noise I'm hearing from the LTE data communications in close proximity to the board may just be lower frequency magnetic pulses being induced and I'm just not going to get rid of it without excessive shielding along the lines of steel casing or MuMetal etc. The finished amp is fully cased in aluminum but it's a non-ferrous metal and does nothing to shield against this.
I may just have to live with placing the cell radio in 3G data mode when the amplifier is right next to it. LTE is quiet enough now to be pretty much inaudible while music is playing but I still don't like that it's lurking there.
The original amp design used a pcb size of roughly 0.9" x 3" with a mostly uniform ground plane on one side but this plane was shared and uninterrupted for all the circuit blocks (power supply, DC converter etc). I think I may keep the board size the same but slot the ground foil between segments to try and keep it from appearing as a long continuous area. Maybe even as far as totally severing the sections and joining them electrically using ferrite beads.
I'm still messing around with layouts and components. I've managed to get the noise down to something much more manageable but far from ideal by shrinking the entire test PCB to 0.9" x 1.2" (posted below).
Nothing I've tried, Ferrite beads, X2Y caps on all signal paths, power supply decoupling etc. have made any meaningful reduction in the noise except a sharp reduction in ground plane surface area.
That no amount of low pass filtering I've tried made a dent in the noise combined with the fact that the designs have been very well behaved in the presence of other strong RF signals brings me back to a earlier suspicion that the noise I'm hearing from the LTE data communications in close proximity to the board may just be lower frequency magnetic pulses being induced and I'm just not going to get rid of it without excessive shielding along the lines of steel casing or MuMetal etc. The finished amp is fully cased in aluminum but it's a non-ferrous metal and does nothing to shield against this.
An externally hosted image should be here but it was not working when we last tested it.
I may just have to live with placing the cell radio in 3G data mode when the amplifier is right next to it. LTE is quiet enough now to be pretty much inaudible while music is playing but I still don't like that it's lurking there
.The original amp design used a pcb size of roughly 0.9" x 3" with a mostly uniform ground plane on one side but this plane was shared and uninterrupted for all the circuit blocks (power supply, DC converter etc). I think I may keep the board size the same but slot the ground foil between segments to try and keep it from appearing as a long continuous area. Maybe even as far as totally severing the sections and joining them electrically using ferrite beads.
Last edited:
Here is the current schematic minus the ferrite beads on the + and - power supply inputs.
The X2Y parts are special filtering caps with 4 connections. 2 in the middle of the capacitor body go to ground plane using vias.
An externally hosted image should be here but it was not working when we last tested it.
Last edited:
Looking at the design one of the first things that caught my eye was the trace leading from the 10uF cap to the 4.7K resistor, it looks like an antenna! Some reference material with antenna design, for you to compare, just spotted somthing else so I will do some checking and be back.
http://resenv.media.mit.edu/rfrain/docs/AN_003_Antennas_1_1.pdf
Pictures Page 11, 12
A bit more RF based stuff.
http://www.eas.uccs.edu/wickert/ece5250/notes/RFMicrowave-PCB-Design-and-Layout.pdf
http://resenv.media.mit.edu/rfrain/docs/AN_003_Antennas_1_1.pdf
Pictures Page 11, 12
A bit more RF based stuff.
http://www.eas.uccs.edu/wickert/ece5250/notes/RFMicrowave-PCB-Design-and-Layout.pdf
Look at these, I dont think the X2Ys are correct for your implementation
http://www.tdk.co.jp/tefe02/e9611_ach.pdf
If you look at your schematic you are shorting out the devices, only one device should be between the In + and In -, they act as common mode devices, or between the in R and in L.
Just an initial thought on the input ones, but worth checking.
http://www.tdk.co.jp/tefe02/e9611_ach.pdf
If you look at your schematic you are shorting out the devices, only one device should be between the In + and In -, they act as common mode devices, or between the in R and in L.
Just an initial thought on the input ones, but worth checking.
I can short the input lines to ground right at the op amp and it makes no difference at all with the noise.
On single ended signal lines you can connect the X2Y caps like that. It effectively doubles capacitance. Basically like placing 2 very low ESR/Inductance 100pf caps in parallel. First figure on this page
X2Y for IC Power Bypass
I looked at that page and it didn't register, woops, I was looking at it in terms of power supply decoupling not line noise suppression, also when looking at the schematic I forget it is two seperate capacitors.
The TDK parts above were used for all inputs on a radio and vehicle communications system, again though talking to the engineer on that project, it is hard to totaly eliminate mobile phone interference.
Again looking through all we do and talking to a fepeople, at the end of the day thats why they dont want mobile phones in operating theatres etc, its almost impossible to keep that level of noise out without extrame engineering and shielding.
The TDK parts above were used for all inputs on a radio and vehicle communications system, again though talking to the engineer on that project, it is hard to totaly eliminate mobile phone interference.
Again looking through all we do and talking to a fepeople, at the end of the day thats why they dont want mobile phones in operating theatres etc, its almost impossible to keep that level of noise out without extrame engineering and shielding.
I only have a 15mhz scope to use currently and can easily see the audible noise but I can't find any substantial higher frequency harmonics. I'm not sure what I'm hearing is even originating from the 800-900mhz or 1.8-2.5ghz ranges. Obviously I can't directly observe those spectrums with the scope I have but you would think I could spot something in the higher frequencies before the scope runs outta steam if the audible noise is strong enough to send the waveform off the grid on the 0.02v/cm scale. it's all negative sweep btw very noisy and hard to get the scope to lock in on it but looks like square wave.
Later today I may make a small sheet steel casing to go around the board and see if that helps.
On a semi-related note. I plugged in the DC/DC converter to the test board and the X2Y cap I have across the power rails worked quite well to help clean up some of the stray 2mhz "fuzz" I had from the converter. It was a very small amount left over using standard decoupling caps to begin with but cleaner is better
I haven't been running the converter on the test board just to isolate the audio section for troubleshooting but I figured I would see how well the X2Y caps faired vs standard X7R and C0G caps I've been using. Not sure if it would be worth the additional cost in a retail consumer product which may be why you just don't see them in most stuff but they do work better than standard filters
I haven't been running the converter on the test board just to isolate the audio section for troubleshooting but I figured I would see how well the X2Y caps faired vs standard X7R and C0G caps I've been using. Not sure if it would be worth the additional cost in a retail consumer product which may be why you just don't see them in most stuff but they do work better than standard filters
Last edited:
It's been a little while since I updated the thread so I figured I'd give anybody who is still wondering a heads up on the issue I've been chasing a while. After working on things a bit I thought of a very simple way to find out if the noise I've been having trouble with is higher frequencies being demodulated or lower frequencies being magnetically induced.
The solution was to just hook up a tiny speaker to my oscilloscope and hold it near the phone when it's transmitting... and what do you know.
The same low frequency pulses I saw at the output of the headphone amp.. Looks like it's just magnetic fields after all.
Having said that I did find ways to significantly reduce the noise in my newer layouts. It may fly in the face of conventional wisdom but I broke up the ground foil into sections with no overlapping pieces. any connections from components to ground are done with vias as close to the component pads as I could manage to keep stray capacitance between the layers to a minimum.
The result in the latest amp is pretty good. I can only hear the noise if I wave the phone along the headphone cable. Sandwiching it against the amp enclosure produces nothing.
The latest portable amp is using that power pig of a TPA6120 paired with a ST MC33078DT input buffer stage. Still manage ~15h run time and pretty decent volumes but the amp is half battery.
The solution was to just hook up a tiny speaker to my oscilloscope and hold it near the phone when it's transmitting... and what do you know.
The same low frequency pulses I saw at the output of the headphone amp.. Looks like it's just magnetic fields after all.
Having said that I did find ways to significantly reduce the noise in my newer layouts. It may fly in the face of conventional wisdom but I broke up the ground foil into sections with no overlapping pieces. any connections from components to ground are done with vias as close to the component pads as I could manage to keep stray capacitance between the layers to a minimum.
The result in the latest amp is pretty good. I can only hear the noise if I wave the phone along the headphone cable. Sandwiching it against the amp enclosure produces nothing.
The latest portable amp is using that power pig of a TPA6120 paired with a ST MC33078DT input buffer stage. Still manage ~15h run time and pretty decent volumes but the amp is half battery.
This is the new amp. Sounds great and tons of grunt. The DC boost converter is outputting +/- 12v @ up to 0.8A
The size in the picture is hard to make out but it's about the same size as a iphone 4 but thicker 0.6"
The size in the picture is hard to make out but it's about the same size as a iphone 4 but thicker 0.6"
An externally hosted image should be here but it was not working when we last tested it.
Outstanding work on your troubleshooting! So inductive pickup by the ground plane. That explains why the (shielded) metal case wasn't solving the problem. B field lines would cut right through. It would take mu metal or some such. On conventional wisdom, I think back to that old Monty Python skit with a crowd saying "we are individuals" and then one guy pipes up "I'm not!".
Last edited:
Outstanding work on your troubleshooting! So inductive pickup by the ground plane. That explains why the (shielded) metal case wasn't solving the problem. B field lines would cut right through. It would take mu metal or some such. On conventional wisdom, I think back to that old Monty Python skit with a crowd saying "we are individuals" and then one guy pipes up "I'm not!".
Thanks.
It was kind of a facepalm moment when I realized that I could use a simple coil to detect magnetic fields around the phone so I first tried a little 10uH unshielded inductor and was seeing something but wasn't a big enough waveform on the scope to tell for sure then I found the small cell phone ringer speaker and that proved to be sensitive enough to get the scope to lock onto it and get a good look at the signal.
Because none of the filtering I was playing with had any effect I was suspecting inductive coupling a while ago but had no way to prove it.
I enjoyed the puzzle while it lasted and I learned quite a bit along the way so time well spent.
My new task with the current amp is to get idle power draw under control. With a 2200mah 3.7v li-po cell I'd prefer to have a couple days of "on" time and only significantly reduce that when really hammering the output load. I've been considering having the DC-DC boost circuit operate at +/- 5v while no music is playing then ramp up the output voltage to the full +/- 12v when it sees source input.
Complete on/off transients into the headphones from turning the converter on and off are fairly minor. a slight "thud" vs a harsh pop and the settled DC offset on the outputs are 0.8mv and -0.3mv which I understand isn't too shabby for the TPA6120 so I think ramping the output from the converter from 5v to 12v shouldn't be too bad in operation and will reduce the quiescent power consumption by more than half while idle.
It would be simpler to just power down the converter while no music is playing but I don't want to hear the thing come out of "power saving" at all. My phone does this already and it's pretty annoying. I can hear the slight pop generated on the phone's output when the phone sits for a few seconds with no music playing and again right as output starts again. Just feels low quality
Last edited:
Interesting looking chip. Looks like even at full bias it's below the TPA6120 in quiescent current draw but I'm somewhat confused as to the 2 tier switching function. From what I understand looking at the datasheet it has 2 input voltage levels and the current draw values are valid for Vcc(L) when the output voltage is below ~3.5v and only uses the Vcc(H) input when the output voltage level exceeds roughly 3.5v?
some explanation http://www.ti.com/lit/an/snaa128/snaa128.pdf
Class G power amp but OK explanation, example power calcs Class-G Amplifiers
Class G power amp but OK explanation, example power calcs Class-G Amplifiers
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.