Why do some chip amps / gainclones sound so bad? (updated 2012-03-30)
Posted 29th March 2012 at 04:56 AM by dunndatt
Updated 31st March 2012 at 12:23 AM by dunndatt (clarification fix)
Updated 31st March 2012 at 12:23 AM by dunndatt (clarification fix)
Years ago when getting started into DIY audio, I picked up three different chip amps from my local electronics shop. They were cheap and easy enough to build according to the data sheets, so I had a go at them. To my dismay, they sounded HORRIBLE! I couldn't figure out what went wrong until I really started digging into the data sheets.
While I didn't build the LM3886, I'm going to pick on it since it is a very common chip amp:
National Semiconductor LM3886 (from national.com, bought out by ti.com)
Claims 68watts into 4ohm speakers with 0.1% THD+N from 20-20kHz
National did pretty good with most of their claims, but I'll focus on something that stands out: The Distortion Vs. Power graphs. (As a comparison, I discovered that one of my other chip amps did their power testing at 10% distortion... which is pitiful. Every chip amp data sheet I've looked at has some kind of failing with this graph.)
The first graph is pretty accurate showing around 60watts. When the line goes vertical, this is where the output starts clipping.
The second graph shows SMPTE distortion. This one tells a different story. The line goes vertical at about 40watts.
The third graph shows the second SMPTE distortion test. The line goes vertical at about 30watts.
So what does all this mean?
If you want true clarity of sound out of an LM3886 the best that it can provide, the gain resistors will need to be throttled back to about 20watts (at least with the 4ohm graphs shown, 8ohm would probably be a similar ratio).
This is about 3.4x less than the advertised power output.
As a side note, my 3 test chip amps suddenly became usable when I throttled their gain resistor ratios back. I doubt many people really do this, and I wouldn't be surprised if many of the quality arguments (read flame wars) have to do with this.
----------------------------------------
Addendum. Shaun brought up a few points I should have initially addressed and clarified. So here goes...
When I first wired up the chip amps for testing and got such horrible levels of distortion, I wasn't really sure what to do about it (I was still very much a beginner back then). My voltage rails were stable and clean and within the part's SOA. I checked for shorts and mis-wirings and couldn't find any. The test speakers didn't have any damage (checked with my home stereo) and were metered to about 8ohms DC. I changed the input volume levels from very low to very high. Nothing made a difference.
I began searching through the data sheets thinking "all this couldn't be right". That's when I found the distortion graphs mentioned above. Since chip amps are essentially op-amps on steroids, the only real way to throttle their power back (that is while keeping the power supply unchanged) is to reduce the gain resistor ratios.
After going back and spending a little time with the soldering iron, the lower gain ratios were set. I did the hook ups again, and, lo and behold, they sounded really good. It was a proud moment for a beginner. To be clear in a confusing world, the power supply voltage and current levels were not changed from before, and the volume levels sounded clean at any level from low to high. The only things changed to make this possible were the gain setting resistors (when encountering problems, test one thing at a time).
Now some of y'all are probably thinking that this doesn't fully add up. As Shaun pointed out, if the gain is lowered, power can be maintained by increasing the input signal levels. This is quite true.
My current theory is that high to full gain on chip amps (or at the very least, my chip amps) somehow stresses the op-amp input equilibrium action to the point of distortion. Reducing the gain proportionally reduced the stress and distortion. Given the simplicity of the chip amp's package, there's not much else left for testing.
It's also my general practice in amplifier building (now in the present) to not set the gain so high that I have to reduce the line level signal to tiny values at the power amplifier's input, just to have the power amplifier circuit increase it again for the output. Since I have some seriously bad health problems at the moment, I can't really afford and don't have the energy to build a rig to accurately measure the power output at the reduced gain settings but using higher signal level inputs to just before clipping. I would be curious, though, to hear about other's experiences regarding the issues brought up in this blog.
While the distortion graphs buried in the data sheets may show a calculated sweet spot, I'm fully aware that there will be listening tests and other tweaks performed by the builders. This is pretty much a given in DIY.
I'd also like to make it clear that I'm not bashing chip amps, but the sales & marketing people who push misleading results in the name of shameless sales and profits. This tends to negatively influence proper setup. (In reality, just about every part imaginable is a bit overrated at the maximum defined settings in the data sheets.) I quite like chip amps for what they are. They make great learning tools and physically small amplifiers with minimal parts. Someday when my health lifts up a bit, I want to use my chip amps to build some small bi-amp'd speakers with a mini-sub and active crossovers. I've been wanting to test out this configuration for a LONG time.
Final Note: What this blog is not about is the overall sound quality of chip amps vs. discrete amps. Each has their place. Each has their advantages and disadvantages. I do not want this blog to become a flame war over those. (but, thanks for reading if you've made it this far
While I didn't build the LM3886, I'm going to pick on it since it is a very common chip amp:
National Semiconductor LM3886 (from national.com, bought out by ti.com)
Claims 68watts into 4ohm speakers with 0.1% THD+N from 20-20kHz
National did pretty good with most of their claims, but I'll focus on something that stands out: The Distortion Vs. Power graphs. (As a comparison, I discovered that one of my other chip amps did their power testing at 10% distortion... which is pitiful. Every chip amp data sheet I've looked at has some kind of failing with this graph.)
The first graph is pretty accurate showing around 60watts. When the line goes vertical, this is where the output starts clipping.
The second graph shows SMPTE distortion. This one tells a different story. The line goes vertical at about 40watts.
The third graph shows the second SMPTE distortion test. The line goes vertical at about 30watts.
So what does all this mean?
If you want true clarity of sound out of an LM3886 the best that it can provide, the gain resistors will need to be throttled back to about 20watts (at least with the 4ohm graphs shown, 8ohm would probably be a similar ratio).
This is about 3.4x less than the advertised power output.
As a side note, my 3 test chip amps suddenly became usable when I throttled their gain resistor ratios back. I doubt many people really do this, and I wouldn't be surprised if many of the quality arguments (read flame wars) have to do with this.
----------------------------------------
Addendum. Shaun brought up a few points I should have initially addressed and clarified. So here goes...
When I first wired up the chip amps for testing and got such horrible levels of distortion, I wasn't really sure what to do about it (I was still very much a beginner back then). My voltage rails were stable and clean and within the part's SOA. I checked for shorts and mis-wirings and couldn't find any. The test speakers didn't have any damage (checked with my home stereo) and were metered to about 8ohms DC. I changed the input volume levels from very low to very high. Nothing made a difference.
I began searching through the data sheets thinking "all this couldn't be right". That's when I found the distortion graphs mentioned above. Since chip amps are essentially op-amps on steroids, the only real way to throttle their power back (that is while keeping the power supply unchanged) is to reduce the gain resistor ratios.
After going back and spending a little time with the soldering iron, the lower gain ratios were set. I did the hook ups again, and, lo and behold, they sounded really good. It was a proud moment for a beginner. To be clear in a confusing world, the power supply voltage and current levels were not changed from before, and the volume levels sounded clean at any level from low to high. The only things changed to make this possible were the gain setting resistors (when encountering problems, test one thing at a time).
Now some of y'all are probably thinking that this doesn't fully add up. As Shaun pointed out, if the gain is lowered, power can be maintained by increasing the input signal levels. This is quite true.
My current theory is that high to full gain on chip amps (or at the very least, my chip amps) somehow stresses the op-amp input equilibrium action to the point of distortion. Reducing the gain proportionally reduced the stress and distortion. Given the simplicity of the chip amp's package, there's not much else left for testing.
It's also my general practice in amplifier building (now in the present) to not set the gain so high that I have to reduce the line level signal to tiny values at the power amplifier's input, just to have the power amplifier circuit increase it again for the output. Since I have some seriously bad health problems at the moment, I can't really afford and don't have the energy to build a rig to accurately measure the power output at the reduced gain settings but using higher signal level inputs to just before clipping. I would be curious, though, to hear about other's experiences regarding the issues brought up in this blog.
While the distortion graphs buried in the data sheets may show a calculated sweet spot, I'm fully aware that there will be listening tests and other tweaks performed by the builders. This is pretty much a given in DIY.
I'd also like to make it clear that I'm not bashing chip amps, but the sales & marketing people who push misleading results in the name of shameless sales and profits. This tends to negatively influence proper setup. (In reality, just about every part imaginable is a bit overrated at the maximum defined settings in the data sheets.) I quite like chip amps for what they are. They make great learning tools and physically small amplifiers with minimal parts. Someday when my health lifts up a bit, I want to use my chip amps to build some small bi-amp'd speakers with a mini-sub and active crossovers. I've been wanting to test out this configuration for a LONG time.
Final Note: What this blog is not about is the overall sound quality of chip amps vs. discrete amps. Each has their place. Each has their advantages and disadvantages. I do not want this blog to become a flame war over those. (but, thanks for reading if you've made it this far
Total Comments 4
Comments
-
You're not clear on what you mean. Are limiting the amount of power you are driving the amps to, or you changing the global feedback? Or are you changing the latter in an attempt to control the former?
The problem is that lowering the amplifier gain won't guarantee a lower operating power level, as one can simply drive it with a higher input signal and still achieve those higher operating power levels.Posted 30th March 2012 at 09:24 AM by Shaun
-
Yeah, you're right. Since the clarification is kinda long, I made and update to the end of the blog. How about that?Posted 31st March 2012 at 12:25 AM by dunndatt
-
It's still difficult to understand, because you are using the wrong words.
"Since chip amps are essentially op-amps on steroids, the only real way to throttle their power back (that is while keeping the power supply unchanged) is to reduce the gain resistor ratios."
You are decreasing the op amp gain by increasing the amount of negative feedback applied. This will result in increased bandwidth and lower noise and distortion.
While your datasheet observations are interesting, its unlikely that there is a causal link to the improvements you noted on lowering the gain.Posted 3rd April 2012 at 02:32 PM by rjm
-
Negative feedback is directly proportional to gain. Lower gain is lower negative feedback. And there was a direct correlation between all 3 chip amps. The reasons were tracked down to the graphs mentioned above.Quote:Originally Posted by rjm
You are decreasing the op amp gain by increasing the amount of negative feedback applied. This will result in increased bandwidth and lower noise and distortion.Posted 10th April 2012 at 05:56 AM by dunndatt
