Made THD measurements on my GC.

[batcat]

Hello all,

I have not posted much but have learned a lot by reading other posts on this forum. I recently constructed a GC amp using the LM3886. I used point to point wiring on the amp and the filters for a bi-amp system driving my OB speakers.

I got a chance to make THD measurements on just the amp section, which most people would be interested in.

Measurement instrument: ST1700B with a residual THD of .003%



Max. Power @ 2khz into 8ohms 60W with 36V rails

THD at 50W into 8ohms 20hz .006%
2khz .004%
20khz .007%

Total noise: -91db

I did not perform the tests into 4ohms because of instability at 4ohms. The amp goes into oscillation as soon as the 4ohms is switched in on all 4 channels. I have two channels driving a 4ohm load. The gain on the amps is 19.

Does anyone think lowering the gain will stop the oscillation at 4 ohms? This is the same schematic as posted in the spec. sheet.

The amp sounds very good and during listening I don’t detect an oscillation. I’m comparing the sound to vintage McIntosh, in some respects this GC sounds better. Although I don’t listen to the McIntosh on the same speakers.

[Fenris]

Originally posted by batcat

I did not perform the tests into 4ohms because of instability at 4ohms. The amp goes into oscillation as soon as the 4ohms is switched in on all 4 channels. I have two channels driving a 4ohm load. The gain on the amps is 19.

Does anyone think lowering the gain will stop the oscillation at 4 ohms? This is the same schematic as posted in the spec. sheet.

The LM3886 is supposed to be used at a gain of 20+. It is not unity gain stable, so lowering the gain probably won't help. Do you have a zobel on the output? Do you have a sufficiently robust power supply?

[markiemrboo]

Originally posted by Fenris



The LM3886 is supposed to be used at a gain of 20+. It is not unity gain stable, so lowering the gain probably won't help. Do you have a zobel on the output? Do you have a sufficiently robust power supply?

The datasheet actually says it is designed to be stable with a gain >= 10, so I imagine that a gain of 19 should be just fine.

LAYOUT, GROUND LOOPS AND STABILITY
The LM3886 is designed to be stable when operated at a
closed-loop gain of 10 or greater, but as with any other
high-current amplifier, the LM3886 can be made to oscillate
under certain conditions. These usually involve printed circuit
board layout or output/input coupling.

Do you have an output Zobel (capacitor in series with resistor from the output to ground)? If you do, are you connecting the return from this (the ground part) to the same ground as the low level ground(s)?

I recently had a problem similar to what you describe, because I had done this. From what I remember I think it was OK in to a 10R resistor load (I don't have an 8R resistor, may have been stable there too), but when I tried a 4.7R resistor it burst in to oscillation. As soon as I separated the input ground from the Zobel ground it went away and has been fine since

Just a suggestion! Hope you can follow what I am on about.

[batcat]

I do not have a Zoble on the ouputs. I am no driving into a xover, but directly to the drivers. Is there a universal Zoble to start with?

Thanks

[markiemrboo]

Originally posted by batcat
I do not have a Zoble on the ouputs. I am no driving into a xover, but directly to the drivers. Is there a universal Zoble to start with?

Thanks

EDIT: Sorry, misread your post. You said you do NOT have one, not that you do I think perhaps adding one might help? Worth a try. I have used a 2W 4.7R carbon resistor and a 100nF cap.


I'm not sure if it matters if you're driving in to a crossover or not. Perhaps someone else can answer that one for you.

Do make sure that you keep the Zobel return / ground separate from the input returns / grounds though. Mine goes to the power ground and seems to be OK.

I run four wires from each amp back to a central point (in my case the disconnecting / loop breaker network).

- The power ground / Zobel
- The input ground for LM3886 stuff (i.e the resistor from input to ground, feedback loop resistor to ground etc)
- The input RCA barrel ground tab
- The speaker return

It could be something else entirely of course, but it does sound suspiciously like the problem I had before With this grounding set up I don't get any (visible) oscillation, no hum and no buzz (at all, even right up close to the speakers), so I can only assume it's somewhere near good grounding!

I know I am still a bit rubbish at explaining, sorry!



I think you should be able to solder this resistor and cap across the speaker terminals on the amp (and from what I gather it might actually be better to this). I have mine on the PCB.

Something like this...

PHP Code:

Speaker Red -----------------------------
                           |
                       [4.7R]
                           |
                      [100nF]
                          |
Speaker Blk  ------------------------------ 


[Nordic]

What voltage were you useing at 4 ohms?

[batcat]

Nordic, I'm using 36v rails if that is what you mean.

[davidallancole]

Are you using a 4 ohm power resistor or a 4 ohm speaker for your load? I mean when it is oscillating?

[Fenris]

Sorry about the confusion, I was working off of old memory. Gain should be fine at 19. Adding a zobel can't hurt stability. A good place to start for values is somewhere between 2 and 5 ohms and a 0.1uF cap. Most people put it on the circuit board, but some (usually those who do P2P wiring) will put it across the output terminals of the amp. Either's fine.

A zobel in a crossover network is designed to work with the high inductance of the driver and give a constant impedance throught the crossover region. If the impedance is not constant, then the crossover components will only be correct at a single resistance, but will be incorrect as the frequency (and hence impedance) changes. A zobel counteracts the rising impedance due to the driver's inductance and creates a more constant impedance.

A zobel on the amp is slightly different. If you assume that a properly designed loudspeaker will have a crossover that will keep the impedance relatively close to constant, then the zobel will have a fixed resistance to work with. It will then form a damped RC network (sometimes refered to as a snubber) that will bleed off high frequency energy (first order f3 pole of approximately 200KHz given 8 ohm speaker load and 0.1uF).

Also, 36v might be a bit high to use with 4 ohms. There's a download from National that helps with designing power supplies for chipamps.

[jackinnj]

You violated one of the boundary conditions when you used the LM3886 with a low impedance load and high rail voltage. the heat sink gets impossibly large (at least for non-forced air convection) at 4R/36V

use National's on-line calculator:

http://www.national.com/appinfo/audi...gn_Guide15.xls