Help to reduce crosstalk in headphone amp(also rate the design) :)

Status
Not open for further replies.
Jump to Latest
GTrenchev

GTrenchev

Member
Joined 2011
#1
Hello!
I've build a headphone amp with LM4562 Op-Amp and BUF634 buffer.Here's the schematic:
Screenshot_8_cr.jpg


THD is fairly low at 0,0015% with 32 ohm headphones.Noise is also good at -95 dB.However,crosstalk is at about -78 dB,which is not that great.Adding more filtering caps reduces it few dB down to about -83 dB.
Could you suggest how to reduce it further,as well as maxing out the schematic ?Thanks!
 
Last edited:
#2
Is it me,or there's something wrong with this forum?I've tried to edit my post,because the picture link cannot be open,in result the post just disappeared,then I wrote another one,which also did not appear,now the first post is here again,that's crazy.
Anyway,this link works(I think) :
[img=http://s2.postimage.org/99mp9cdg/Screenshot_8_cr.jpg]
Sorry if I've messed something up!
 

Attachments

  • Screenshot_8_cr.jpg
    Screenshot_8_cr.jpg
    82.2 KB · Views: 392
#3
GTrenchev said:
Could you suggest how to reduce it further,as well as maxing out the schematic ?Thanks!
Click to expand...

Well a couple of thoughts...

I'm assuming you are using one half of that LM4562 for the right channel and the other half for the left and that you are doing your own PC boards. If you are going after crosstalk, consider using a single LME49710HA (the metal can version, there is a plastic dip version too) for each channel. The LME4562 is the same as an LME49720. There is also a metal can version of the LME49720 for about 10x the money - some folks seem to like the metal cans better:

Digi-Key - LME49720HA-ND (Manufacturer - LME49720HA/NOPB)

For crosstalk layout is the other big thing, of course, making sure there is good separation between the input traces for the two channels and that you are using star grounding. The PSRR for the amp parts is good enough that you will probably not get crosstalk in via the power rails. Your schematic looks fine from what I can tell. I'm assuming those are 22uF tantalum for the adjustment pin bypass. One suggestion would be to go to something truly huge, like 1000uf aluminum electrolytic there for the best in ripple rejection and low output impedance. See appendix C in here

Amazon.com: Troubleshooting Analog Circuits (EDN Series for Design Engineers) (9780750694995): Robert Pease: Books

You might want to also take a look at the "wire" amp thread in this section by opc:

http://www.diyaudio.com/forums/head...igh-performance-headphone-amplifier-pcbs.html

opc is getting another run of boards together right now. That board uses an LME49600 for the output buffer, but there is a BUF634 with the same pinout and package style that should work in the board (the specs on the two parts are nearly identical). And the "wire" uses (surface mount) LME49990s which are newer and probably better than the LME49710. The "wire" is all surface mount parts. The original amp was balanced input to SE output, so take a look at post 106 which is the measurements for single ended, with one input grounded. You can compare those to your results. One of the new boards will be SE->SE.
 
Last edited:
#5
You are doing pretty well (measurements) for a perfboard. With a PC board (through-hole or surface mount) you could get a little more control over things like spacing between long(er) wire runs.

I also forgot to say that dual-gang potentiometers on the input and TRS jacks on the output are also known sources of crosstalk. Using an "audio" pot like

http://www.bourns.com/data/global/pdfs/ptd90.pdf

PTD902-10C5P-A203 Bourns Potentiometers

may help, or using an attenuator. For the jack get a good one like the neutrik jack opc used on the original "wire" amp. If you used a dual pot and/or jack when you did those measurements you might want to try them again, with just one channel at a time wired into the pot and output jack.

 
Last edited:
#6
You're right,I use crappy pot and jacks,did not expected to have such impact on crosstalk.
The sound is very high-quality indeed,very neutral and accurate.It just lets the music pass,without adding anything.Maybe that's "The Wire Lite" then😛
I want to ask something else.I have some minor HF oscillations on one channel.
I thought it was because of crappy layout,but that's the second board I made,very careful and neat.Yet the problem is still present.Here's a pic,take a look at the spikes at the treble area:
[img=http://s3.postimage.org/1kuo80byc/Screenshot_10_cr.jpg]
The other channel is just fine:
[img=http://s3.postimage.org/1ku4dk610/Screenshot_9_cr.jpg]

It is completely unaudible,of course.
Don't mind the 50/100 Hz hum introduced on the graph,as it's because of crappy cables.

Note:That's present only with LM4562.No problems if using OPA2132P,but I don't like the sound.
 
#8
Hey I'm glad that helped solve the crosstalk problem! 🙂

I don't have any ideas unfortunately about that high frequency ringing just on one channel. If it were both channels I would say try unsoldering the BW pin on the BUF634 to drop the bandwidth and see if that helps. But your other channel is OK. The other thing I would usually recommend is start replacing parts looking for a noisy cap, resistor, or op amp, but you have built a whole second board which rules out bad parts. Inputs too close to the output can cause oscillations - may be worth checking the wiring path on that one board half.

The link isn't working in your last post.
 
#9
[img=http://s3.postimage.org/1y3r43gck/Screenshot_11_cr.jpg]
It was an attachment of final measurement showing extremely low THD.Crosstalk is good enough now.

I've even followed completely different layout.I don't think there's anything wrong with it,traces are thick and well separated.I think LM4562 is just nervous in this schematic..for some reason.
What about compensating caps?I've got some 33pF NPO,where should I put them,if possible?
I will order OPA1612,hoping to like them🙄
 
#10
-78 dB crosstalk? Shirley you jest. Have a look at the very best phono cartridge crosstalk numbers or look up what's considered sufficient. -78 dB is vastly better than needed. That said, I don't think you'll do a lot better with both channels in the same IC. Possibly lowering the impedance of all lines that might pick up signals, but that would degrade other things. You might measure the crosstalk over the entire audio band and beyond to get a clue about what the path is. HF, look for capacitive coupling. LF, look for common paths on the wiring or supply issues.
 
#11
GTrenchev said:
What about compensating caps?I've got some 33pF NPO,where should I put them,if possible?
Click to expand...

Take a look at figure 4 on page 9 in the BUF634 data sheet:

http://focus.ti.com/general/docs/li...enericPartNumber=buf634&fileType=pdf&track=no

They show the cap from op amp output to the inverting input to reduce the gain to unity at high frequency. In that note on the figure they say "NOTE: (1) C1 not required for most common op amps. Use with unity-gain stable high speed op amps." Well the LM4562 is unity gain stable from the data sheet and would probably fit their definition of "high speed". Worth a try! They used 220pF and left the BW pin connected. I'm kind of fond of multilayer ceramic X7R caps these days.

Other than that, I would agree with Conrad that two chips are better than one, one for each channel, for separation. That alone might solve it. You could use LME49710s, or if you have a bunch of the LM4562s just one one per channel and don't use the other half of each one.
 
Last edited:
#13
Conrad Hoffman said:
-78 dB crosstalk? Shirley you jest. Have a look at the very best phono cartridge crosstalk numbers or look up what's considered sufficient. -78 dB is vastly better than needed. That said, I don't think you'll do a lot better with both channels in the same IC. Possibly lowering the impedance of all lines that might pick up signals, but that would degrade other things. You might measure the crosstalk over the entire audio band and beyond to get a clue about what the path is. HF, look for capacitive coupling. LF, look for common paths on the wiring or supply issues.
Click to expand...

For natural listening with headphones some crosstalk is actually desireable. With speakers there is air between the ears and the speakers which will allow for some degree of crosstalk. This is absent when using headphones and is often actually incorporated electronically in some high-end headphone amp designs.

http://sound.westhost.com/project109.htm
 
Last edited:
Status
Not open for further replies.
Top Bottom