AMPLIFIER TESTING: HOW-TO QUESTIONS

[AndrewT]

Locate the Main Audio Ground (MAG) somewhere near all the ground references/connections that the amplifier needs. None on these are on the chassis.
Now run a connection from that MAG to the PSU Zero Volts on that nice big useless buss bar.

Run another connection from MAG to all the exposed metal parts, ready for when you mount it all inside a box.

[CharlieLaub]

Originally Posted by RNMarsh

Might try testing with MultiTones -- If you can find the software/hardware to do it.

Also take a look at your whole system.... measure the thd/harmonic/Im from source to speaker terminals of the power amp. get the whole distortion data for your entire system -- as that is what you are actually hearing. That total distortion level can tell you a lot... like which harmonics are dominate and where the most improvement could be done. -RNMarsh

Hi RN in Cool,

I am only focusing on linear distortion testing at this point. I carry out testing on DIY loudspeakers that I build and I'm aware that the loudspeaker will likely have 10x more distortion that a middle of the road amplifier. As a result, I don't plan to get too serious about other parts of the reproduction chain until things are much farther down the road.

At the moment I am checking out this chip amp that was part of a PA (karaoke?) plate amp. I have a bunch of them, and they are convenient for my needs, so I want to explore the performance limits/limitations. I'd like see how I can clean up the output in various configurations and loads so that I know how to best put them to use.

-Charlie

[Zero Cool]

Originally Posted by RNMarsh

Might try testing with MultiTones -- If you can find the software/hardware to do it.

Also take a look at your whole system.... measure the thd/harmonic/Im from source to speaker terminals of the power amp. get the whole distortion data for your entire system -- as that is what you are actually hearing. That total distortion level can tell you a lot... like which harmonics are dominate and where the most improvement could be done. -RNMarsh

Where is the LIKE button??

[CharlieLaub]

Originally Posted by AndrewT

Locate the Main Audio Ground (MAG) somewhere near all the ground references/connections that the amplifier needs. None on these are on the chassis.
Now run a connection from that MAG to the PSU Zero Volts on that nice big useless buss bar.

Run another connection from MAG to all the exposed metal parts, ready for when you mount it all inside a box.

I perceive this as an unfriendly, unhelpful, and somewhat insulting post.

-Charlie

[CharlieLaub]

I moved on the the next revision of the power supply to see what changes and improvements could be gained. First, I added two more 10k caps, and used a 1.2 ohms series resistor on each rail in between the "first' set of caps, and the "second" set. I continue to use the "useless" buss bar, since it is at least convenient for making connections...

So, very different results than what I was expecting! The PS with the inline resistors (forming an RC lowpass filter with corner frequency no higher than 13Hz) did have much reduced ripple at 120Hz, but lots of higher frequency noise or harmonics shot up considerably! The mains noise (60Hz) was also higher. See attached plot of the amplifier on but with no input signal, and the measurement noise floor.

Not only that, but the distortion performance across the board was worse, too. For instance, at 1kHz the distortion increased to 0.0574% from 0.0385%. Similar changes happened across the board.

So, while ripple was reduced, everything else is worse. I had imagined that the RC filter would kill all higher frequency noise and that the additional caps would be beneficial.

Perhaps the additional wiring is just creating more "antennas" that can pick up noise? I did try to reposition the components in an attempt to improve this but it did not help. Everything is just sitting on the desk next to my computer, so noise pickup is a possibility...

Thoughts?

-Charlie

[CharlieLaub]

More changes produced "REV 3" of the PS.

I decided to remove the series resistors on the rails, but to keep the second set of caps in place. This produced the following changes:
the ripple harmonic (120Hz) returned
the mains noise (60Hz) was reduced with respect to REV 2
the higher frequency noise/harmonics increased with respect to REV 2
the 1k Hz distortion returned to the level found for REV 1 of the PS

So, to rank the performance in terms of noise and ripple, from best to worst:
REV 1 (single set of caps)
REV 2 (two sets of caps with a 1.2R series resistor on the rails)
REV 3 (two sets of caps directly connected at rails)

When two sets of caps were used, I connected the buss bars with wire. The transformer center tap is connected at the buss bar of the first set of caps. The speaker returns and a ground wire that leads to the PC board are connected on the second buss bar.

So, what's up here? Is layout a problem? As I add more "stuff" am I just picking up more noise or is it being generated in the/by the components that I am adding??? I have not found many good resources on the physical layout of a power supply like this, so I did not think it would be terribly important.

-Charlie

[Zen Mod]

connect those gnd wires as drawn

twist feeding lines to pcb

connect speaker gnd returns to gnd points on pcb

xformer's CT must be connected to that bus bar only , not to pcb

[whizgeek]

stick in a whole bunch of 1u and 0.1f bypass caps,, then put a bunch of ferrites and common mode chokes, faraday cages, twisted pairs, snubbers, MOVs, mains filters, dc blockers, , --,--, etc.
Then do it all again with the resistors back in place.

only joking, but you do have a large can of worms there...maybe it should be a can of Guinness. the very best of luck to you, it would be great to have a definitive answer.
(this thread is going to be better than the ryder cup!)

[CharlieLaub]

OK, I followed Zen Mod's suggestions (except for the one about running the ground wires to the PCB instead of to the star ground). Unfortunately there was no real change in the level of mains noise/harmonics.

I then tore down the PS and re-wired it in the "REV 1" configuration with a new bridge (in case I slightly toasted one of the diodes or something...). I also built a new shielded input cable from coax in case that was where noise was creeping in. When I connected everything and powered up again, the mains noise and harmonics were still there to some extent, not as low as I originally measured when testing REV 1. I actually did try to insert some 10-100nF ceramic caps in the secondary and this had absolutely no effect on the harmonics. Hmmm.

Then something dawned on me.. it got pretty hot out there today (100F) and maybe this is just due to dirty mains power??? Not sure, but as of right now it's early evening and people are likely hunkered down with lots of air conditioning on, etc. I will have to remeasure in the morning, and if this happens again see if a line filter will help.

This "REV 1" power supply (single pair of caps connected via a buss bar, with all grounds connected there) seems simple and effective, so why make things more complicated? I am getting lower distortion (THD) levels, such as:
100 Hz: 0.005%
300 Hz: 0.005%
500 Hz: 0.007%
1k Hz: 0.016%
5k Hz: 0.07%
10k Hz: 0.087%

These distortion figures meet the design spec of the amp, as specified in the data sheet, so I am feeling good about this result. This is at "full power". I actually measured the RMS voltage with a voltmeter (albeit a crappy one) and that was about 16V, which corresponds to 32W of power. I'll take it!

I will try to post more info when I have it.

-Charlie

[CharlieLaub]

Ooops, forgot to post plots. Here is with no signal but amplifier on, 100Hz and 1k Hz (see attachments).

-Charlie