LME49810/11/30, LM3886 hum issues & grounding

[tomchr]

Folks,

As I seem to have hi-jacked the "Comparing LME49810, LME49811, LME49830 thread" with my grounding issues, I have started a new thread.

The post below seems to be a good starting point for this.

---

So as promised, here are the noise plots (in dBV/rt(Hz)) for an LM3886 amp with a gain of 20 V/V (26 dB). When building the amp, I made sure to have proper star grounding.

I am fairly certain that the remaining 60 Hz hum is electrostatic coupling into the input and/or feedback. I suppose I could shield the circuit with some aluminum foil to find out for sure, but I haven't bothered.
Initially, I had issues with multiple harmonics of 60 Hz showing up on the output. I got rid of those by grounding the heat sink the LM3886 is mounted to. I'm using the TF version (isolated package).

There's still a very, very faint hiss in the speaker. But the hum is gone. The hiss is only audible when I put my ear 1-2 cm from the tweeter dome so I can live with it.

Bottom line: Star ground does matter.

~Tom

[tomchr]

Latest update: As I broke down the setup following my post, I discovered that the output of the amp wasn't connected to the load but rather to a distortion analyzer that was turned off. If I connect the amp to the load, I get more harmonics of the 60 Hz.

My load resistor is set up on a rather large aluminum heat sink with fans. I wonder if it is acting as an antenna. I'll need to figure this one out...

I now have two channels working. I don't hear any hum in the speakers just a very, very faint hiss when I press my ear against the speaker. I got it done late last night so I haven't spent a lot of time listening to it, but my initial impression is good.

~Tom

[peranders]

You must be aware of the power supply rejection ratio. How much is the hum related to the output really? -90 dBV = 30 uV

Every amp has a noise and your result is very good.

[tomchr]

Originally Posted by peranders

You must be aware of the power supply rejection ratio. How much is the hum related to the output really? -90 dBV = 30 uV

Every amp has a noise and your result is very good.

I know. But the engineer in me can't help being curious as to where the little bit of hum that is left is injected.

~Tom

[tomchr]

I got it figured out. The 60 Hz "spur" was indeed caused by electrostatic coupling into the circuit. I wrapped the circuit in aluminum foil (after insulating it with a plastic bag) and measured the noise again. All ripple components are now gone.

Interestingly, grounding the load resistor heat sink knocked 6 dB off the 60 Hz spur. That was definitely an unexpected result.

Figures below show the output noise voltage (dBV/rt(Hz)) with the amp input grounded with a 50 Ohm terminator. The amp gain is 20.6 V/V (26.3 dB). The noise floor measures approx -130 dBV/rt(Hz), corresponding to 316 nV/rt(Hz) - or an input referred noise voltage spectral density of 15.3 nV/rt(Hz). Quite nice!

~Tom

[Speedskater]

Tom, what hardware and software do you use to make those measurements?

[tomchr]

For the noise measurements, I use an HP 3562A dynamic signal analyzer. For the distortion, an HP 8903A audio analyzer.

I control the instruments via a Prologix USB-GPIB controller using KE5FX's GPIB toolkit to get the plots from the 3562A. And I use a MATLAB script that I've developed to make the distortion curves. Eventually, those scripts will make it on to my Audio Website that I'm in the process of setting up.

~Tom

[kasey197]

Tom, as I am in the process of setting up a bench, could I ask why you'd use the HP8903A for the distortion instead of the HP3562A ? If I had the budget/time to learn to use only one, which would you think I should get ?

[tomchr]

I find the 8903A to be easier and more flexible in day-to-day use. Its built-in source has extremely low distortion. It's nice to have a distortion number readout so one can tweak the circuit to achieve minimum distortion (bias adjust for example). Its AC voltmeter is nice and makes conversion from voltage to power into 8 ohms a snap (just enter: 19.0 SPCL).

But if I want to measure the noise floor or figure out which harmonic is the dominant contributor to the THD, I use the 3562A. It's also useful for IMD measurements or to figure out how much 60/120 Hz ripple makes it through the amp. It's a nice, does-it-all box but the much underappreciated HP 3581A Wave Analyzer is a realistic substitute if you know the frequency of the signal of interest.

~Tom

[infinia]

120 Hz should dominate you should see it above noise floor esp at load! from the 60 Hz reading being highest means you have a measurement ground loop. on the DUT yank the chassis to earth to check or better yet connect it to the SA chassis with a piece of braid or Cu tape. Something is not right if adding shielding chages the noise reading.

I use a big aluminum plate sitting on the bench connecting the DUT (PS +amp +load) and SA together.