How to measure signal crosstalk and SNR

[firstcolle]

HI,
I want to measure the crosstalk between left and right channels of an amplifier, the datasheet says -112db with 1KHz sine at 1Vrms.
how can I measure it??

Can I simply to put 1Vrms 1khz into left channel, measure the V on right channel and then calc the dB with 20*log(Vright/1Vrms) ??

and for snr?

I'm not looking for super precise values, it's just for try to measure this kind of information.


thanks

[ejeffrey]

Yes, you can do that.  Be careful, even at audio frequency it can be hard to actually measure -112 dB crosstalk.  That is a few microvolts rms.  So you scope is not going to cut it, you need a higher resolution instrument like an audio spectrum analyzer.  Also, be aware that crosstalk in your measurement apparatus can also show up.

For SNR, the problem is that few instruments have enough dynamic range to show your 1 V signal and the noise at the same time.  The normal solution is to use a notch filter that blocks 60+ dB of the 1 kHz test tone, then look at the result on an audio spectrum analyzer.

[firstcolle]

thanks for the super fast replay!

mm.. yes, -120db should be 1uV... definitely something that I can't measure with my scope.. 
but measuring higher values (in the mV order) could represent a problem in the IC or in the PCB layout.

[TheUnnamedNewbie]

You need a audio spectrum or audio network analyzer to do this. No scope is going to cut it there.

[ejeffrey]

A traditional oscilloscope won't have enough dynamic range, but a very good sound card line input might work, good audio CODECs have a SNR around -100 or -110 dB.  Make sure the signal levels are suitable -- don't blow it up with a speaker level signal.  You can also make a simple low noise preamp with a good opamp.

[CopperCone]

i dont see why you cant use a low noise preamplifier to take your measurement. Beware however that the scope has crosstalk to so only measure one channel at a time.

A few lt1028 would probobly work fine. Or lt1001.

I recommend the jim williams circuit in the app note about measuring ldo noise. You get a dual stage filter using lt1028 and ac coupling and a brickwall low pass filter at 100khz. Its caled design idea: performance verification of low noise low dropout voltage regulators.

Mine works well but i used silver mica caps i had on hand rather then polyester. I put it in a metal tin and have options for battery power or banana jack lab psu power. My modifications included a ferrite bead and low voltage gas discharge tube on the input (would add esd diode too if ihad one on hand) and some tantalum capacitors to ground (like 100uf) for the banana jack power input. I used a dual throw switch to change between battery power or lab supply power.

I also added another bnc connector on the output of the gain stage so i can look at the raw signal before the brickwall filter if i want.

Also works as a decent preamp for a microphone (including ultrasonic ones)

Of course the application is differnent then a real audio spectrum analyzer, which is sweapt, however it will offer a lower noise floor then even the best fft device from hp.. But it is a boat anchor that weighs like 50 lb. And the dynamic range is less.

Better to add a preamp to a top end analyzer, especially due to real timebw, but for your application a hp3582a would work well since its a stable neasurement. On the lowest attenuator the noise floor should be less then -120db. Kinda pricey still though.

Just using the preamp i mentioned and a regular multimeter would be very accurate at 1khz, for instance a hp34401a in db mode would beat the crap out of a scope for repeatability and accuracy. I would actually trust dmm over scopes at such a low frequency.. Just make sure your input sine wave ic clean from hf that could lead to ameasurement error