JLH1969 MOSFET Class A/AB amp

HIFI2008

继续请教millwood 、qjdj 等多位老师：

图中一共需要5个电容。输入、输出电容的作用我明白，但其他三个电容（4.7uF、470uF）的作用是做什么的呢？？
需要分别选择什么类型的电容？？

从原理上来说，2N5401和和2N5550的Cob对音质影响大吗？？2N5550有没有必要换成2SC3423（Cob=1.8pF）？

qjdj

输入基极通过电阻连接的4.7uf是电源退耦电容。我理解可以滤除电源通过偏置传过来的杂波。实际我用的是120uf的电容。
反馈对地的220uf就是交流反馈电容。另外一个220uf电容我理解是自举电容。

millwood

yes, the 4.7uf is for decoupling, to improve PSRR.

the cap on the feedback loop is a DC blocking cap, to maintain the DC working point for the input transistor.

the other cap is a bootstrap cap. it can be omitted.

you can use 2sc3423 if you want. the input transistor should be a small signal PNP transistor; the phase splitter should be a medium power NPN transistor. I have actually used a P-channel small signal mosfet as input transistor and a N-channel power mosfet as the phase splitter without any problem.

HIFI2008

多谢qjdj老师和millwood老师指导

学生还有些地方不是很明白，以本帖第13页#254楼，millwood老师发的图为例，qjdj老师说的分别是哪个电容？？容量没有明确的要求吗？

millwood老师说有两个自举电容可要可不要？？能不能从电路原理上解释一下？？另外，实际音质会不会受到影响？？

qjdj

不客气！请不用称呼我“老师”，不太习惯，论坛交流为好。谢谢！
先将13页254楼的图片转过来说明。

输入基极通过电阻连接的4.7uf是电源退耦电容就是C6
反馈对地的220uf就是交流反馈电容就是C2
另外一个220uf电容我理解是自举电容就是C3
“millwood”朋友说的是C3可以忽略。我是按照220uf做了的。具体原理我的能力无法解释清楚。抱歉！

HIFI2008

感谢qjdj~~

C2和C3在这个全电流反馈的图上容量是470uF，用220uF会不会有什么区别？？

我用仿真测试了一下，如果供电电压为41V，那Q3的集电极电流为28mA左右，如果Q3使用的是2SC3423，根据特征频率，应该是在20mA的时候效果最好。
所以我试了一下改R2和R15的阻值，大概在400欧的时候，2SC3423的集电极电流为21mA，这样修改对于电路的其他部分有什么影响？反馈量小了，是不是相对的控制力度就弱了？？这两个反馈电阻多大才合适？？怎么推算出来的？？

qjdj

我理论分析不了，不过我做过，因为供电电压的变化，调整过R2和R15的阻值，电路工作正常，音质听不出变化。20ma对于中功率管比较大，需要加散热器。

HIFI2008

顶起来，请各位大侠继续分析~~用理论武装到牙齿~~

qjdj

做出来听着再来分析就更有劲哦。
已经几次推荐大家试试了。

millwood

hifi2008, good questions.

I haven't found the addition of R15 to be that beneficial. But it does change the output coupling from voltage coupled to current coupled and some people find that appealing.

1) bootstrap network (R1/R14/C3): this works to increase gain on the phase splitter. It is essentially a passive constant current source. one unique thing bout capacitors is that voltage over a capacitor cannot change suddenly (current going over an inductor cannot change suddenly). So by connecting the left of R14 to the output through C3, the left of R14 goes up and down with the output. the right of R14 also goes up and down with the output, through Vgs of M3 (almost). so in theory, if the left and right of R14 fluctuate together and are identical (almost), there shouldn't be any (ac) current going through R14. Thus, to Q3, R14 has infinite impedance. This greatly increases the gain on Q3.

the amp runs just fine without C3 (thus the boostrap network).

the bootstrap network is also an effective stability tool. As C3 is typically an electrolytic capacitor which turns more resistive than capacitive at high frequencies, the effectiveness of the bootstrap network goes down at high frequencies (typically way out of the audio band), which reduces the open loop gain at high frequencies -> something you want for stability reasons.

Bootstrap'ng is making a come back, and AKSA and P3A, two of the most successful amps of our times, both use such technics.

b) 220u vs. 470u: not much difference. 220u yields less bass but barely noticeable to me.

c) Q3's idle current: R2 and M4's Vgs determines Q3's idle current. I typically set it at least 10x of M4's gate current, which is about 2 - 3ma at 20khz, depending on the mosfets you use. as mosfets typically open up at Vgs = 3.5v, R2=220ohm yields 3.5v/220ohm=15ma. Now, if your mosfets open up at higher (or lower) voltages, you can adjust R2 to get you there.

I will make sure that R15 = R2 if R15 is used - Otherwise, the amp has different gains during the positive and negative cycles.

if R15 isn't used, I will make sure that R14 = R2.

d) what impact does Q3's idle current have? well, if you dial up Q3's idle current, voltage drop over R1/R14 will go up, this will turn off Q1 and M3/M4. so your output DC voltage will go down.

You will need to readjust at least R1 to make sure that M3/M4 idle right.

e) feedback network: the gain is set by 1+R5/R4. Higher R5/R4 will yield better bass response, but will also cause the output voltage to go down (because the voltage drop over R5 will increase). This is also a current feedback amp and they are more stable and have lower noise with lower value feedback resistors. I typically use 110ohm or 220ohm for R4 and then choose R5 based on the amount of gain I need.

How do you set the gain? I typically make sure that with a 1Vp input, the amp is at full power. for linear applications, mosfets typically drop 5 - 6v when "fully" open. so with Vcc=40v, your output swing, peak to trough, is about (40 - 2*5)=30v, or 30v/2=15v peak. This means that your amp needs to have a gain of 15Vp/1Vp = 15x. That means R5=(15-1)*R4=14*110=1.56k max.

you can use smaller resistors and I happen to like 1k resistors.

hope it helps.

q1688

请问那个图是这个贴的主角

jsgcc

原帖由 millwood 于 2008-5-30 22:31 发表
hifi2008, good questions.

I haven't found the addition of R15 to be that beneficial. But it does change the output coupling from voltage coupled to current coupled and some people find that appealing.

1) bootstrap network (R1/R14/C3): this works to increase gain on the phase splitter. It is essentially a passive constant current source. one unique thing bout capacitors is that voltage over a capacitor cannot change suddenly (current going over an inductor cannot change suddenly). So by connecting the left of R14 to the output through C3, the left of R14 goes up and down with the output. the right of R14 also goes up and down with the output, through Vgs of M3 (almost). so in theory, if the left and right of R14 fluctuate together and are identical (almost), there shouldn't be any (ac) current going through R14. Thus, to Q3, R14 has infinite impedance. This greatly increases the gain on Q3.

the amp runs just fine without C3 (thus the boostrap network).

the bootstrap network is also an effective stability tool. As C3 is typically an electrolytic capacitor which turns more resistive than capacitive at high frequencies, the effectiveness of the bootstrap network goes down at high frequencies (typically way out of the audio band), which reduces the open loop gain at high frequencies -> something you want for stability reasons.

Bootstrap'ng is making a come back, and AKSA and P3A, two of the most successful amps of our times, both use such technics.

b) 220u vs. 470u: not much difference. 220u yields less bass but barely noticeable to me.

c) Q3's idle current: R2 and M4's Vgs determines Q3's idle current. I typically set it at least 10x of M4's gate current, which is about 2 - 3ma at 20khz, depending on the mosfets you use. as mosfets typically open up at Vgs = 3.5v, R2=220ohm yields 3.5v/220ohm=15ma. Now, if your mosfets open up at higher (or lower) voltages, you can adjust R2 to get you there.

I will make sure that R15 = R2 if R15 is used - Otherwise, the amp has different gains during the positive and negative cycles.

if R15 isn't used, I will make sure that R14 = R2.

d) what impact does Q3's idle current have? well, if you dial up Q3's idle current, voltage drop over R1/R14 will go up, this will turn off Q1 and M3/M4. so your output DC voltage will go down.

You will need to readjust at least R1 to make sure that M3/M4 idle right.

e) feedback network: the gain is set by 1+R5/R4. Higher R5/R4 will yield better bass response, but will also cause the output voltage to go down (because the voltage drop over R5 will increase). This is also a current feedback amp and they are more stable and have lower noise with lower value feedback resistors. I typically use 110ohm or 220ohm for R4 and then choose R5 based on the amount of gain I need.

How do you set the gain? I typically make sure that with a 1Vp input, the amp is at full power. for linear applications, mosfets typically drop 5 - 6v when "fully" open. so with Vcc=40v, your output swing, peak to trough, is about (40 - 2*5)=30v, or 30v/2=15v peak. This means that your amp needs to have a gain of 15Vp/1Vp = 15x. That means R5=(15-1)*R4=14*110=1.56k max.

you can use smaller resistors and I happen to like 1k resistors.

hope it helps.

hifi2008，一个好问题。

我未发现增加R15那么有效。但它确实将输出耦合由电压耦合改为电流耦合，一些人喜欢这个。

1)自举网络(R1/R14/C3): 增加了分相器的增益。
它实际上是一个无源的常值电流源。
电容的一个特点是它两端的电压不能突变（电感两端的电流不能突变）。（bout->about）
所以通过C3将R14的左端与输出相连，R14左端的电压与输出电压同步变化。
R14的右端同样随输出电压上下变化，叠加了M3的Vgs（几乎，差不多）。
理论上，如果R14两端电压同步变化且几乎一致，不应该有任何（交流）电流通过R14。
这样对Q3而言，R14有无限大阻抗。这大大增加了Q3的增益。

这个放大器没有C3（即自举网络）也能很好工作。

自举网络也是一个有效的稳定工具。因为C3通常是个电解电容，阻抗在高频是会增大，自举网络的效率在高频时下降（通常在音频以外），这就减小了高频时的开环增益，正是稳定性所需要的。

自举是正在重新流行的技术，如AKSA和P3A，两个当今最成功的放大器，都用了这个技术。

b) 220u vs. 470u: 没有太大区别。220u 低频稍欠，但我几乎无法察觉。

c) Q3的静态电流：R2和M4的Vgs决定了Q3的静态电流。
我通常取为M4栅极电流的10倍，在20khz约为2 - 3ma，看你用什么场管，典型的导通 Vgs = 3.5v，取R2=220ohm 得到电流：3.5v/220ohm=15ma。
如果你的场管导通电压不同，调整R2即可。

如果用了R15，我会取R15 = R2。除非放大器正半周的增益与负半周不同。
如果不用R15，我会确保R14 = R2。

d) Q3的静态电流有什么影响？
如果调高Q3的静态电流，R1/R14上的压降会增加，这将使Q1和M3/M4截止，所以输出直流电压会下降。

你需要一个最小的R1，以保证M3/M4的静态正常。

e) 反馈回路：增益取决于 1+R5/R4 。
R5/R4较高会得到更好的低频响应，但会使输出电压降低（因为R5的压降会增加）。
因为这是个电流反馈的放大器，如果反馈回路电阻值较小，会更稳定和更低噪。
R4我通常用110ohm 或 220ohm，再根据需要的增益确定R5的值。

如何设计放大倍数？我通常使得1Vp输入信号时，放大器满功率。
对线性应用，场管的饱和压降可取5 - 6v。所以当Vcc=40v, 你的输出摆幅大约是(40 - 2*5)=30v，或峰值30v/2=15v。这意味着需要的增益为：15Vp/1Vp = 15倍。所以 R5 = (15-1)*R4 = 14*110 = 1.56k （最大）

你可以用更小一点的电阻，我喜欢用 1k。

希望能有所帮助。

感谢millwood提供这么多设计经验，翻译出来希望更多人得到启发。

jsgcc

给个仿真的结果，参数调整到静态电流2A左右，比较清凉一点。



失真很小：0.001%
感谢楼上qjdj兄提供的ms10仿真文件，非常有用，穷烧们尽可以虚拟大功率！

音响DSP

EWB10很容易出错。
有谁可以提供做好的板子给我上AP测试，请PM。
理论联系实际，光做个仿真是不够的。

millwood

有谁可以提供做好的板子给我上AP测试，请PM。

even if an amp tests great on AP, so what? does that make it sound better than another amp that doesn't measure as well on ap?

your train of thoughts is exactly what the Japanese had in the 1970s and 1980s and that has been largely abandoned by the 1990s.

a lot of today's high-end amps do NOT measure well. Most of Pass's designs fall into that category. By all account, they sound pretty good, good enough for people to fork over $40K for each of them.

j10

楼上所言的确是实话。

音响DSP

就许你说你的好，别人说的好，你就说别人是作假和骗子。
3%的失真，听的人估计都要吓跑。晒下你的听音设备吧。
贵的不等于好。PASS机器，软绵绵的，没推力。去年11月在深圳圣廷苑的雅乐音响展，听过MBL的前级、柏林之声顶级CD机、B&W的音箱、Classe的功放以及NuForce的D类，都很一般。
要玩所谓的音色，单端恒流源负载缓冲输出级前级，我早就试过了，danielyang、工仔、FUMAC和偶然等朋友都听过，不喜欢。
JLH小甲功放，风景依旧也有。
依照你的逻辑，你说指标没用，听感很好，可是你自己也说批判过我们的PK听感主观性太强，不可信。典型的双重标准！
那你的JLH电路客观测试指标不好--3%的失真。
至于主观，我的2级组合功放会让更多的人品尝，也会跟不同机器进行对比。好东西不怕对比和测试的。
欢迎有JLH功放的朋友，本周六下午到我家进行对比，请PM。

millwood

就许你说你的好，别人说的好，你就说别人是作假和骗子。

not really. if you say that A is better than B, I expect you can back it up with facts and data.

3%的失真，听的人估计都要吓跑。晒下你的听音设备吧。

speakers routinely distort in the order of 10%. so unless you can listen to electric signals, you are listening through a 10% thd speaker regardless of how good your amps are.

think about that.

贵的不等于好。PASS机器，软绵绵的，没推力。去年11月在深圳圣廷苑的雅乐音响展，听过MBL的前级、柏林之声顶级CD机、B&W的音箱、Classe的功放以及NuForce的D类，都很一般。
要玩所谓的音色，单端恒流源负载缓冲输出级前级，我早就试过了，danielyang、工仔、FUMAC和偶然等朋友都听过，不喜欢。
JLH小甲功放，风景依旧也有。[quote]

that is a great example of what sounds good to one person doesn't mean it will sound good for the rest. So when you proclaim how good the as2604 is to you, without having any data to back it up, think about your own experience above.

[quote]依照你的逻辑，你说指标没用，

I never said that and you couldn't be more wrong on that.

听感很好，可是你自己也说批判过我们的PK听感主观性太强，不可信。

blind listening test

how many times do we have to repeat that to you so you understand what this discussion is all about?

那你的JLH电路客观测试指标不好--3%的失真。

first of all, it is not mine JLH. it is JLH's jlh.

2ndly, many tests are available to show that it does well into the <1% thd territory.

至于主观，我的2级组合功放会让更多的人品尝，也会跟不同机器进行对比。

blind listening test

how many times do we have to repeat that to you so you understand what this discussion is all about?

好东西不怕对比和测试的。

then why are you trying to hard to avoid subjecting your "great sounding" as2604 to a scientific, verifiable blind test?

what are you afraid of?

欢迎有JLH功放的朋友，本周六下午到我家进行对比，请PM。

I hope this time, you set it up as a blind test and you could describe for us how it is done and what the results are.

BTW, don't forget to level match the two devices you are comparing.

民主

gao1942

有人出pcb板或套件吗？