amplifier analysis: open loop vs. closed loop performance
we listen to amps typically with some forms of feedback. Feedback is used to lower distortion and to improve stability.feedback wasn't widely used before 1950-1960s because tubes and early germanium transistors, only active devices available back then, were not high gain devices and using feedback would destroy however little gain those devices had. when silicon transistors became popular, and cheap, people started to use opamp like structure for audio amps. Soon, manufacturers discovererd that using feedback can drastically improve the specifications of their amps, making it much easier for marketing purposes.
through 1970 - 1980s, you would see lower and lower THD figures. However, soon people discovered that those amps did not sound better.
some audio designers started to analyze thd figures and found it limitations: it gave equal weighting to high order harmonics than low order harmonics. further analysis showed that that is in part caused by poor open loop performance, both in terms of distortion and phase stability / bandwidth.
from there, the whole high-end audio started to go back to low- or no- negative feedback design, and to strive for good "open loop" performance and not to rely on feedback to provide better specifications.
What I wanted to do is to show how one can analyze the open loop performance of his amps.
The example I am taking is the famed JLH1969, in mosfet configuration.
here is the schematics.
it is identical to the prior version I posted under a differrent thread, with the following modifications:
1) this amp was designed to drive low and capacitive impedance load (electrostatic amps). so I used irfp140 on the output.
2) because irfp140 has high gate capacitance (Cgs), I used irf510 as the phase splitter. irf510 has the advantage of small Cgs so no change for the front end is needed.
3) I inserted an inductor, L3, into the feedback loop to help me block any ac signal being fedback to the front stage but allow the same DC operating point. This is used for simulation only.
if you give the inductor a very small value (like 0.1uh here), it would be like a straight wire and the amplifier would work in closed loop. if you give the inductor a large value (like 10h), it would pretty much block any ac signal and the amp would run open loop.
and lastly the gain vs phase plot, from a/c analysis. Again, flat frequency response all the way to 2mhz, and no stability until about 3mhz. excellent numbers. next, I will do the open loop analysis, by adjusting the inductor to 10h. The amp will have much higher open loop gain so I will lower the input to 0.001v ac to product roughly the same output signal.
I will need to go out for breakfast and will finish this when i get back. sorry for this. 楼主能不能先翻译好再贴出来:victory: o man ! Well done:lol 楼主写得很认真，佩服！可以当作扫盲贴~ to do the open loop analysis, we will increase the inductor (L3) to a large value to block any ac signal from being fedback to the input transistor. We can achieve that with a 10h inductor - you can use a higher number that but doesn't do a whole lot.
when you do that, the (open loop) gain of the amp will be signficant so we are driving the amp using a 0.01v ac signal (see Vsin).
here is the schematic.
following by the output at 20khz.
the most interesting is the last chart, gain and phase shift vs. frequency, which will be in the next post. this is the gain / phase shift vs. frequency plot. The green line is the gain (in log form) plot, and the red line is the frequency plot.
1) gain vs. frequency: pretty much flat up to 30khz on the high end, and -3db to be at about 300khz. remember those figures.
the low-end drop is in part due to the use of an output capacitor and dc blocking capacitor in the feedback loop.
2) phase vs. frequency: about +/- 40 degrees through the audio band. and instability at about 3mhz.
Just how good those numbers?
Here is the open loop frequency response from KF358 (Fairchild's version of LM358/2904 opamp, a "high-performance" chip 20+ years ago). the gain starts to dip in less than 10hz (vs. JLH's 30KHz).
here is the open loop gain vs. frequency reponse from NE5532/34, another high performance chip20+ years ago. it significantly outperforms the KA358: the gain starts to dip after 100hz. Still no match to JLH's 30Khz.
another well known chip: LM3875. Gain starts to dip after 100hz.
those chips are fairly representative of the kind of open loop performance of "high performance" voltage feedback opamps. In the end, they all perform badly in open loop and have to rely on large amount of negative feedback to achieve good "specifications" for closed loop applications, resulting in good measurements (THD for example) but bad sound.
Because of that, people in high-end audio have largely abandoned the philosophy, and start to achieve foundamentally linear amps with good open loop performance and then either let it run open loop or use moderate amount of negative feedback.
so JLH1969 may be an old design, and it really is 40 years ahead of its time. We are now just catching up to it and to fully understand it and appreciate it. So when you look for your next amp design, ask yourself how well the amp performs open loop before you go out and build one.
By theway, tubes are inherently high speed devices and a lot of tubes amps run with noor very little feedback, delivering great open loop performance. you may ask that if there are chip opamps with good open loop performance.
Actually, they do exist, and almost all of them are high speed current feedback amps, of which the JLH is one.
here are a few examples.
1) AD815: some of the earliest and best known high-speed current feedback amps. Widely used for preamp or headphone amp. 400ma output current, 900v/us slew rate, 120mhz bandwidth. sometimes you will find this in SIP package and they are highly desirable.
2) opa2674: 500ma output current, 2000v/us slew rate and 250mhz bandwidth (at unity gain).
3) ths6012 / tpa6120a2: 400ma output current, 300mhz bandwidth and 1300v/us slew rate. tpa6120 is designated as a headphone amp by TI.
you will find a lot more. Most of those chips were originally developed as either video amplifier or xDSL drivers so you will find quite a few of them in old ADSL modems.
However, those chips are not without fault:
a) being a current feedback amp, they tend to have low input impedance. You may need to have a buffer infront of them.
b) they tend to oscillate, due to their high speed. they don't work well with reactive load so you may need to use an output resistor to slow them down when driving into a reactive load.
c) they are very picky about layout.
I have used the opa2674 and tpa6120 as headphone amps and preamps and they are some of the most accurate amps I have ever heard.
I would encourage you all to give them a try. in case people are interested, here are some open loop gain plots for some well known chipamps.
1) LM1875: something interesting about its open loop gain, vs. those for LM3875/3886.
2) OPA227/2132: those used to be called "ultra high performance" opamps before people discovered why they don't sound so great.
3) OPA54x family: they are also used to make "gainclones" and some reported their "warm" sound. But those are very slow amps (Gain bandwidth 1mhz vs. 8mhz for the LM3886).
4) LM49810: notice how bad its open loop gain is.
a few interesting opamps are lm4562 (and the associated LME family audio opamps) from National. They look very good on paper at least.
enjoy. Rod Elliott has an article http://sound.westhost.com/highspeed.htm on high speed amp and THS6012, a driver for ADSL modem.Probably you can apply the same to TPA6120.
Cheers, JLH和运放两者用途不同，不能踩着运放来提高JLH
什么是运放？就是能进行运算放大的，需要有两个输入端、并且符合输入端虚短、高输入阻抗、低输出阻抗、直流特性要好。要实现运放基本功能，就需要牺牲一些东西。如果不要求这些直流特性，很多电路都能实现很好的开环带宽，
要对比就应该和差不多年代的对比，例如磁带前置LA3210，他的开环特性也是很好的，直流特性也和JLH一样，但现在有人用它吗？
现在的趋势是追求直流化，但LA3210、JLH最大的缺点是这里。所以难以得到通用。
Is there any audio-specific operating amp?
Most of the illustrated chips are designed for universal purpose. Maybe the latest specialized audio chip can exhibit better open loop performance.I agree with such point: tubes are inheretly high speed devices. Though JLH has a output capacitor, it is conceptually advanced in the aspect of how to apply feedback. In other words, the primary target for an audio amplifier, is to realize optimal open loop performance and then, if necessary (as always necessary to solid state amp), to apply modest feedback to achieve maximum performance. But anyway NF is not a panacea that is able to cure all circuitry and element defects. The reason is that the existing distortion evalution system fails to reflect the full range of distortions that may affect sound quality. Large amount feedback can perfect some the of distortion performance indicators, but in the same time may severely deteriorate the others, which has been already been incorporated, or been reluctantly incorporated into the evaluation system. Some vendor has recognized them and align their amps to these "private" standards, but the industry as a whole, has not openly recognized them, not to mention to publish the guiding standards for them.
Sorry.
"which has NOT been already been incorporated" 大致能看懂 这外籍华人还真多 Kenny,Would you please let us know more about the other distortion performance indicators?
Thanks in advance.
Well, one of the key issues concerning other distortion types...
is that they are hard to measure by any single approach. Just take audio-specific vacuum tube for example: ordinarily the data sheet will give the specifications for universal purpose, such as the typical operating points, ri, plate voltage, and so on. Altogether there would be at most a dozen of checking items. But according to inside source, the tube selection company or the reputed audio-gear vendors will check at least a dozen more of items. And these items are kept confidential. Sorry that I don't know them, or I would start a factory myself.回复 #16 kenny 的帖子
:lol 《子不语》其实大家可以用中文回答，millwood也是华人，能看懂中文，只是他不方便输入。 la3210 does have negative (current) feedback, done through r4/r8. given the big difference between the frequency responses when the compensation capacitor is used, I would venture to guess that those transistors in la3210 donot have great high frequency response either.
原帖由 locky_z 于 2007-10-28 10:55 发表 http://bbs.hifidiy.net/images/common/back.gif
现在的趋势是追求直流化，但LA3210、JLH最大的缺点是这里。所以难以得到通用。
you are absolutely right about JLH's lack of DC stability. However, that can be addressed with a slight reconfiguration of the output stage -> making it into a NAD3020 for example.
But most importantly, I haven't met anyone who can listen to DC, or seen a speaker that can reproduce DC signals. No CD has any content below 16hz. so why do we require our amplifiers to reproduce DC? 直流化并不是仅仅为了听到更低频率的声音,而是为了取消耦合电容,
耦合电容不仅仅对低频的下限产生影响,对高频音色影响更甚 原帖由 millwood 于 2007-10-28 19:50 发表 http://bbs.hifidiy.net/images/common/back.gif
la3210 does have negative (current) feedback, done through r4/r8. given the big difference between the frequency responses when the compensation capacitor is used, I would venture to guess that t ...
LA3210的PDF里面标准电路，R8是被一个电容旁路的，所以也算开环。
页:
[1]
2