RFI problems on diy headphone amp.

[Amperage]

Hello. I've been working on a couple headphone amp designs for weeks now. One is a design using a maxim max9722 headphone amp IC and the other is a slightly more complex build using NJM4556AD dual op amps. Both amplifiers share the same power delivery and charging system composed of a texas instruments BQ24072 li-po battery charger and STM6601 smart push button controller to handle soft on/off and low voltage protection. The 4556 based headphone amp also has a LTC3122 based DC/DC boost converter circuit configured to deliver +/- 12v to the 4556 op amp.

Both amplifiers are nearly silent almost no hiss or other odd noises I can attribute to the amplifiers themselves and I'm quite happy with the audio quality but I've been having persistent issues with RFI from my 4g cell phone when it's transmitting or receiving data within approximately 1 foot of either headphone amp.

I have read numerous guides on dealing with RFI and have implemented a number of suggestions. low pass filters on input and feedback circuits, ferrite beads, obsessive ground fill on both board layers. No matter what I try it doesn't seem to affect the signal bleeding in from the cell phone. I've held the amps right next to a number of other RF sources including routers, microwaves, AM/FM transmitters etc. and the only component that seems to introduce any noise at all is cell phones.


I've been a forum lurker for quite a while and I apologize for my first post being one asking for assistance but I'm kind of stumped currently on how to go about this problem and thought perhaps somebody with experience could give some suggestions.

 

[s3tup]

Pics of the device, both sides of pcb (stuffed/empty) will help.
I'd guess you've overlooked some sensitive loop...

 

[Amperage]

Sorry took so long. I've tried a couple other things including a test board that contains only the audio section to try and eliminate noise coming in from the rest of the circuit. it's currently running from 2 3.7v li-po batteries for testing to produce +/- 3.7v. Nothing has changed with the RF noise. Below is a current schematic for the test board.

[IMGDEAD]http://207.255.124.163/Forum/Output.jpg[/IMGDEAD]

 

[Amperage]

Oops. Actually I missed 2 input load resistors on the schematic that are in the actual circuit. The fixed schematic below.

[IMGDEAD]http://207.255.124.163/Forum/Output_Updated.jpg[/IMGDEAD]

 

[s3tup]

Schematics appears to be fine. The RFI thing is mostly layout/connections thing.

Please show the pics of actual devices from both sides of PCB, connected to your test equipment (this one matters too).
Have you tried the amp with batteries, headphones and portable source - do you still get that buzz?

 

[Amperage]

Here is the board layout top and bottom. The input is connected to a HTC Rezound 4g phone using a 12 inch shielded (braided copper + foil wrap) 3.5mm male to male TRS cable. The output is only driving various types of headphones. It's not used as a preamp or feeding anything above a 40ohm load so far.

[IMGDEAD]http://207.255.124.163/Forum/4556TestTop.jpg[/IMGDEAD][IMGDEAD]http://207.255.124.163/Forum/4556TestBottom.jpg[/IMGDEAD]

Thanks for the assistance by the way . I've been fighting with this for more than 2 weeks now. It's no wonder so many retail products have RFI problems.

 

[Fredfred2004]

Just a couple of suggestions

What you have there is basically a cmoy with a very high gain of 11, try reducing the gain to 2 or 3,

your virtual ground would benefit from a resistor divider, just the two caps will not work as well to divide the voltage

The 2.2K resistors should be much larger say 100K

good luck with the build

cheers

 

[qusp]

yes, gain of 11 plus driving only low impedance headphones, a poorly decoupled power supply and highish impedance ground, are all things that can lead to RF interference. I would also suggest a common mode filter/choke part to replace the discrete inductors

 

[Amperage]

Sorry for the confusion. I only posted the output section of the schematic but if you look in the OP you can see I'm not using a virtual ground system. The ground is 0v ref common ground all the way through.

For testing purposes to try and isolate the RFI problem I built a board with only the output stage replacing the power end of the design with 2 li-po cells center tapped but in the full amp I'm generating a true + and - 12v using a LTC3122 DC/DC converter with output regulation. Not a fan of virtual grounds so I added the LTC3122 operating at 2mhz to the mix.

Again I've temporarily eliminated all the power supply circuit and went with a test board with only the amplifier section driven by 2 li-po cells in series. Sort of a divide and conquer approach. Since the RFI still exists with only 2 batteries powering it I know that it's not being introduced by the battery charging system or DC/DC converter.

 

[Amperage]

Also I started out with much higher standard "cmoy" resistor values but decided to lower them to tighten up op amp dc offset and lower hiss etc.

For the intended purpose of the amplifier a gain of 2 to 3 would end up being too low for the output voltage swing on the source device and quite a number of other cell phones and mp3 players. The amp is driving mostly inefficient larger low ohm headphones like the sony XB series or stuff like the Beats by Dre line . A gain of 3 wouldn't provide enough output with the max clean input level my rezound can provide.

Something like an iPhone has more clean voltage swing at the headphone jack and can get away with lower gain levels and still push the power output of the amp near it's limit.

Gains of 10 seem to be fairly common for the "high" setting on tons of existing amps. So I've been aiming on achieving that.

Maybe I should switch to a newer generation of op amp with better noise rejection?
The output section of my amp idles at around 23mA. A bit higher than I was shooting for anyway. Would like to get it down to under 10mA to extend battery runtime.

 

[JMFahey]

I've been having persistent issues with RFI from my 4g cell phone when it's transmitting or receiving data within approximately 1 foot of either headphone amp.

Short answer: keep your cell phone further away.

Monster RF interference makers, which by the way is to be expected, they are 2 way powerful RF transceivers, what else do you expect?

Mine gets into the PC, the phone, the TV set, *everywhere* ... and it's fine.

If it's communicating to some 100 to 200 ft high RF tower which may be up to a couple Miles away, is it strange that some of that energy causes interference 1 ft away?

 

[Amperage]

Short answer: keep your cell phone further away.

While I can appreciate the ease of this fix I'm not sure it's possible since this is a portable headphone amp likely to be in a pocket living either alongside the cell phone or very nearby.

I would like to write this off as "not possible" and I know cell phones are monster RF emission machines by design but there are some retail amps that are apparently immune or at least very well behaved strapped right onto a cell phone. FiiO E6 comes to mind.

I have a number of devices that experience RF bleed from my cell phone and living in a low signal area doesn't help one bit (the phone has to ramp up power to connect) but that doesn't mean I shouldn't try to do it better and thought perhaps I was missing something obvious or at least something more seasoned forum members may have battled with and won.

 

[s3tup]

Okay, PS decoupling. throw some as-large-as-you-can-find X7R MLLC caps, something in 1206-1210 size ans 4.7-22uF capacity, right at opamp's supply pins looking towards the middle of the board at the GND layer. this way we do as good decoupling as possible.

Then, given the ground fill and tiny circuit which should work against EMI and they dont - i'd bet on the improper ferrites use (btw, which are these? there are lots of these each of which are kinda different...)

The improer way is... EMI thing says - put filtering on _every_ wire going in/out of the circuit. You've missed the GND wire and PS section. Given the largish length of the connected wires and leading towards another EMI mistake ("put all the connectors and wires on same side, same spot").. I'd put ferrite beads on every connected cable and see whether it fix something...

A bit of workaround would be use of FET-input opamp just to check whether it helps... Their input stage isn't prone to RF demodulation thing.


It's great that youv'e put up this issue, as it seems it is both really common AND really ignored especially at DIYers community.

Oh, and it's a nice puzzle to solve

 

[Ruwe]

Hi,
I assume that your kit is in some sort of metl or aluminum box, which is grounded. If that's the case the RFI can enter most likely from the inputs and the outputs. The input seems high impedenace to ground, but your output is not.
The first thing I would try is to short the output coils or to replace them with some small resistor (may be 1-2 ohms for headphones output is OK). The coils form a sort of tank circuit with your feedback capacitors. I don't see the values of your coils, so can not calculate the frequency, but it's possible that your output cable works like a nice antenna at high frequencies.
Try that first, if you still have problems short the input coils too.

 

[Amperage]

Okay, PS decoupling. throw some as-large-as-you-can-find X7R MLLC caps, something in 1206-1210 size ans 4.7-22uF capacity, right at opamp's supply pins looking towards the middle of the board at the GND layer. this way we do as good decoupling as possible.

I didn't put larger decoupling caps at the op amp because in the actual amp (not the test circuit) literally 1/2 inch away are 47uf X7R caps coming from the DC/DC converter. They aren't the cheapest of components and since they are really close in the full amp layout and still EFI problems I just didn't place them on the test board.

Then, given the ground fill and tiny circuit which should work against EMI and they dont - i'd bet on the improper ferrites use (btw, which are these? there are lots of these each of which are kinda different...)

The improer way is... EMI thing says - put filtering on _every_ wire going in/out of the circuit. You've missed the GND wire and PS section. Given the largish length of the connected wires and leading towards another EMI mistake ("put all the connectors and wires on same side, same spot").. I'd put ferrite beads on every connected cable and see whether it fix something...

The ferrite beads I'm using are these
LI1206H151R-10 Laird-Signal Integrity Products | 240-2402-1-ND | DigiKey
I've only just started learning about EFI/EMI suppression (crash course really). I checked around for material on calculating the proper type of ferrites and didn't come away with anything solid so I just got those to try last order. If you think they are usable I'll go ahead and add more to the other lines.

A bit of workaround would be use of FET-input opamp just to check whether it helps... Their input stage isn't prone to RF demodulation thing.

Are there any you can suggest with comparable output current to the NJM4556AD I'm currently using? If need be I can probably squeeze output buffers in alongside a lighter opamp. I have some AD8397's on the way currently but they are standard bipolar I believe.

Hi,
I assume that your kit is in some sort of metl or aluminum box, which is grounded. If that's the case the RFI can enter most likely from the inputs and the outputs. The input seems high impedenace to ground, but your output is not.

The test circuit isn't but the actual amp that I am troubleshooting is in a full aluminum casing grounded very well to the board but not earth ground (portable battery powered).

The first thing I would try is to short the output coils or to replace them with some small resistor (may be 1-2 ohms for headphones output is OK). The coils form a sort of tank circuit with your feedback capacitors. I don't see the values of your coils, so can not calculate the frequency, but it's possible that your output cable works like a nice antenna at high frequencies.

Those are just SMD chip ferrite beads. The software I'm using (diptrace) doesn't have ferrite bead symbol in the library so I just used the standard inductor symbol to represent them.
I have already tried removing them in pairs, outputs, input and both installing jumpers across the pads. Nothing seemed to change the amount of RF signal bleed. In fact I can remove all filtering including the DC blocking input caps and the RFI doesn't seem to get either better or worse.. It's just being stubborn

 

[agdr]

Are there any you can suggest with comparable output current to the NJM4556AD I'm currently using?

Try the OPA2132:

http://www.ti.com/lit/gpn/opa2132 (opens PDF)

which is FET input rather than bipolar. From the chart on page 6 of the datasheet it current limits at around 30-40mA with a 10V swing, depending on temperature, a bit less than half the NJM4556A's 70mA. The 2132 is pin compatible with the NJM4556A and has nearly the same operating voltage range.

Any non-linearities in the circuit or chip, like the diode (base-emitter) junctions in the bipolar inputs and outputs, can de-modulate the RF.

To help track down where the RFI is getting in, temporarily tack solder a short on each NJM4556A input under the board. Just a tiny wire from the socket pin to the nearest ground point. If that doesn't stop the problem then your input circuit is OK.

A 2500Mhz cell phone's wavelength is around 12cm, way too long for the tiny bonding wires and die inside the chip to be an effective antenna. Has to be something outside the chip conducting the RFI to the chip and/or demodulating the RF.

If shorting the inputs doesn't stop it, try making a "T" network out of two 10 ohm resistors soldered in series with a 150pF-or-so ceramic cap (with leads on the resistors and cap cut as short as possible) soldered at the junction of the two resistors. One resistor free end goes to an output under the board on the NJM4556A, the other free resistor end to one headphone channel, and the free end of the cap to a ground point near the output pin. If this filter reduces the problem then the RFI is conducting in through the output lead and getting demodulated in the chip.

If neither the input or output test nails it, that leaves the feedback loop and power supply feed.

BTW, welcome to the forum! That is a great user name. Amazing that on a forum like this nobody had thought to use "Amperage" before.

 

[qusp]

Amperage, I can highly recommend both the parts and reference materials at the Murata site regarding common mode and differential filters. you will likely need a bit of both. Noise countermeasures.

they have some very good compound filter parts with common mode and differential filters (C || L || CRC type stuff) SMD filters. the selection guide on the site is quite good also. common mode noise is not the sort of thing you can really do a broad solution for, its more a build and measure thing, or at least model, you need to know the frequencies of interest, load and source impedance etc. not easy to know beforehand.

 

[Ruwe]

I have already tried removing them in pairs, outputs, input and both installing jumpers across the pads. Nothing seemed to change the amount of RF signal bleed. In fact I can remove all filtering including the DC blocking input caps and the RFI doesn't seem to get either better or worse.. It's just being stubborn

Then, as agdr already suggested, probably the RF enters through the power supply or you are just exciting some feedback loop oscillations with your phone.
By the way, why are you using the 150pF capacitors in the feedback? The OpAmp should be stable at gain of 11 without them. Also those capacitors may be working together with the 150pF ones at the inputs and add to the problem. Have you tried to disconnect the 150pF in the feedback loop?

 

[gootee]

I didn't put larger decoupling caps at the op amp because in the actual amp (not the test circuit) literally 1/2 inch away are 47uf X7R caps coming from the DC/DC converter. They aren't the cheapest of components and since they are really close in the full amp layout and still EFI problems I just didn't place them on the test board.

You will want to put some small-size (not necessarily small value) caps RIGHT AGAINST the power pins. It's not optional. There is a positive feedback loop for high frequencies, through the power rails (for almost every transistor-based amplifier). You must short-circuit the HF to ground. It's even possible that the RFI is much worse than it would be, because it's exciting that feedback mode. A half-inch away is useless, for this. Even two mm is too far away.

Use physically-small caps, the smallest you can manage, but use the largest value that comes in that size.

That might not solve your immediate problem but it has to be done, anyway. You will also want to tighten up your layout. Every component that connects to an IC pin should be soldered as close as possible to the pin. As close as possible doesn't mean a few mm away. Any open lenth of trace connected to a pin will be presenting parasitics to the pin, which should be avoided.

Also, look for any loops that enclose geometric area. They will make antennas (See Faraday's Law.).

 

[Amperage]

Try the OPA2132:

http://www.ti.com/lit/gpn/opa2132 (opens PDF)

which is FET input rather than bipolar. From the chart on page 6 of the datasheet it current limits at around 30-40mA with a 10V swing, depending on temperature, a bit less than half the NJM4556A's 70mA. The 2132 is pin compatible with the NJM4556A and has nearly the same operating voltage range.

That's a pretty sharp cut in max power output but if nothing else I try helps that's something not too difficult or expensive to try. If the output is clear of interference I can add a 2nd stage unity gain op amp/buffer for increased power handling.

To help track down where the RFI is getting in, temporarily tack solder a short on each NJM4556A input under the board. Just a tiny wire from the socket pin to the nearest ground point. If that doesn't stop the problem then your input circuit is OK.

This is a great suggestion and what I'm going to try next. Easy, quick and free way to eliminate the inputs as the possible source.

If shorting the inputs doesn't stop it, try making a "T" network out of two 10 ohm resistors soldered in series with a 150pF-or-so ceramic cap (with leads on the resistors and cap cut as short as possible) soldered at the junction of the two resistors.

I had considered doing this before and it's easy enough to patch in quick but I avoided it because of all the negative information I've read about op amp stability problems driving capacitive loads.

Amperage, I can highly recommend both the parts and reference materials at the Murata site regarding common mode and differential filters. you will likely need a bit of both. Noise countermeasures.

Thanks for the link. I'll have to spend some time tonight checking it out

By the way, why are you using the 150pF capacitors in the feedback? The OpAmp should be stable at gain of 11 without them. Also those capacitors may be working together with the 150pF ones at the inputs and add to the problem. Have you tried to disconnect the 150pF in the feedback loop?

They are there simply cause it was in some of maxim's white papers for anti-RFI measures. the RFI existed without the feedback caps. Those were one of the later additions when I started messing around trying to clean this up. I added the input 150pf caps, then the feedback caps and finally the ferrite beads on the inputs and outputs.

You will want to put some small-size (not necessarily small value) caps RIGHT AGAINST the power pins.

I'll tack some on when I add above mentioned shorting jumpers across the input pins. I'll see if I can rework the layout a bit tonight. to speed things along I'm making all the prototype boards at home so I'm pretty much limited to double layer PCBs so I can't get too fancy with independent signal and ground planes. Sometimes I have no choice but to route traces a few mm around something else in the way.

Thanks again for the help everyone. I'll try some of these suggestions and come back later tonight with the results.