- 积分
- 1255
- 在线时间
- 6435 小时
- 最后登录
- 2024-4-26
- 阅读权限
- 70
- 精华
- 0
 
- UID
- 3284
- 帖子
- 3192
- 精华
- 0
- 经验
- 1255 点
- 金钱
- 1272 ￥
- 注册时间
- 2005-3-7
|
楼主 |
发表于 2019-7-15 10:25
|
显示全部楼层
这个图片，失真号称几乎测不出。请看作者的描述：
As the next experiment, I decided to get rid of the crossover distortion. For this purpose, I added a traditional adjustable bias circuit with a transistor and a trimpot. Now I also had to add a current source, because with the bias circuit there is no single point into which the OpAmp could put its drive current into both bases! I adjusted the bias for the best distortion, and this was really a good one! The second harmonic was down right where the test oscillator delivered it, about -80dB, so I couldn't really measure it! The third harmonic was at -84dB, and the best improvement was that the higher harmonics had simply disappeared! They were all below the noise floor, which stayed at -120dB. Actually, this noise floor seems to come from the soundcard A/D converter, so that the actual noise of this and the above amplifier may even be better! With music, this amplifier sounded perfect - clean and smooth. And I'm pretty confident that the THD is well below the limits of my measurement setup, which is 0.01%.
The quiescent current was around 10mA. When lowering it to about 3mA, the high harmonics started to rise out of the noise floor. If you want to adjust the bias for the exact best quiescent current, there is a simple trick: Lift R4 from the output, and connect it to pin 6. Now the output stage has been left outside the feedback loop, and all its distortion will show up at the output. Watching the signal on an oscilloscope, or even better on a real time spectrum analyzer (soundcard and software), adjust the trimpot to the lowest distortion level. Have a current meter in the supply line and make sure that you don't exceed 30mA or so of quiescent current, in order to keep the small transistors cool. But most likely the best distortion will be at a current lower than that. Once the adjustment is complete, return R4 to its normal position. Now the full gain and slew rate of the operational amplifier is used to correct the small remaining cross-over distortion of the output stage, and the distortion will certainly disappear from the scope screen, from your ears, and possibly fall below the detection level of the spectrum analyzer!
This circuit can also be run from a split power supply, by exactly the same mods as for the previous circuit. And since the transistors are properly biased, there isn't any significant distortion increase when using larger transistors. Be sure to use some that have enough gain - you have only a few mA of driving available, and with a +-15V power supply and an 8 Ohm speaker, there can be almost 2A of output current! So, you need a gain of 300 at least. There are power transistors in the 4A class that provide such gain, and these are good candidates. The other option is using Darlington transistors, which far exceed the gain needed here. But they will again increase the distortion, not very much, but perhaps enough to make it audible again.
The conclusions, drawn from this day of applied science, are the following:
- Integrated circuits, even if they are of the same type, are not all born alike;
- One can easily outperform a simple IC by using a few discrete parts;
- It pays to measure how a circuit performs, and not just blindly trust a textbook;
- Simple, small, inexpensive amplifiers of high performance can be easily built;
- A computer is a great laboratory instrument!
|
-
|
发表于 2019-7-14 21:36
