澳門客 發表於 2014-4-24 03:55 static/image/common/back.gif
建議你用效率高一點喇叭，或分析力強一點喇叭，再仔細對比，就知呆不呆板...
小弟用audio physics tempo 25,
個效率已經89db...{:1_352:}
本帖最後由 johnme 於 2014-4-24 10:51 編輯
澳門客 發表於 2014-4-24 04:05 static/image/common/back.gif
甚至輸出用220uf二支也呆板，那裡我用得很細，視負載而定，3a內我只會用10uf，3至5a我就會用100uf而已，這 ...
小弟已經做左兩個LT1084 regulators,
一個Cadj有用100u同一個冇用100u,
分別唔大,
不過分低頻析力係好左d.
至於Cout,
小弟朋友試過唔同combo之後,
recommend小弟用220 x 2,
小弟之後做過一個用兩粒220既試過,
發覺整體改善明顯,
只是中頻比較緊,
所以決定再試其他combo.
正式用印好既pcb做regulator時,
心中有以下方案:
1. 用1u
2. 用100u
3. 用220u // 1u
4. 用oscon 47u (不過RS個價都好有威嚇性)
5. 用100u//10u mlcc
(最後個combo係LT1963個APPLICATION NOTE上,
見個TRANSIENT RESPONSE效果最好)
http://cds.linear.com/docs/en/application-note/an104f.pdf
最後用220u,
因為方便同之前做比較.
唔知各位有何睇法?
澳門客 發表於 2014-4-24 04:05 static/image/common/back.gif
甚至輸出用220uf二支也呆板，那裡我用得很細，視負載而定，3a內我只會用10uf，3至5a我就會用100uf而已，這 ...
這個原則我同意.
billymud 發表於 2014-4-24 06:18 static/image/common/back.gif
師兄美言，實不敢當 。其實小弟跟隨您各個帖子從中學習，得益斐淺哩！
客氣~{:1_253:}
johnme 發表於 2014-4-24 10:12 static/image/common/back.gif
小弟用audio physics tempo 25,
個效率已經89db...
未到95db還算不上高效果喇叭{:1_253:}
billymud 發表於 2014-4-24 06:09 static/image/common/back.gif
師兄，綠性PSU按負載調整輸出級電容值, 高見！深表贊同.
想、請問用在NAS供電, 果度肉E-cap...
旦電容雖然細粒好用，但我不認同用在低頻供電裡會發揮得好，反而用充放電快高級的電解電容會更好，由於一般都是用10uf/16v，我建議最好用blackgate也不會太貴{:1_253:}
澳門客 發表於 2014-4-24 14:13 static/image/common/back.gif
旦電容雖然細粒好用，但我不認同用在低頻供電裡會發揮得好，反而用充放電快高級的電解電容會更好，由於 ...
Ack and thx.
本帖最後由 johnme 於 2014-6-14 15:59 編輯
Eventually, I have built the first prototype with the Sigma-TMC board described in earlier post.
In this prototype, I made the following changes (compared with the original sigma 11):
1. add CRC snubber (C1, C2 and R1) for damping oscillation caused by the leakage inductance of xformer and capacitance of rectifiers when the rectifiers are turning off.
2. omitted R6 (0.47ohm) due to my careless mistake,
the space on PCB is too small to fit such a resistor
3. add C13 and use Oscon here
4. add C9, C10 and R4 to form a TMC help tp stabilize the PSU.
1. CRC snubber
The most commonly used snubbers are those small caps connected to rectifiers.
However, people miss the key: it is the resistor that snub not the capacitor. In other words,
we need to use RC, a resistor connected to a cap in series and then connect them to a rectifier.
I connected a pulse generator (set tp 120Hz) connected to a mosfet to simulate the oscillation caused by the sudden cut of current in a rectifier(Figure 1). And measure the value of resistor that critically damp the oscillation.(Figure 2)
Here are some typical results for Nuvotem transformers (All 230 primary 15V secondary (parallel) )-
For a 120VA xformer, C1= 0.01u, C2 =0.15u, R1 ~42.5 ohm
For a 80VA xformer, C1= 0.01u, C2 =0.15u, R1 ~39 ohm
For a 50VA xformer, C1= 0.01u, C2 =0.15u, R1 ~38.6 ohm
Use the above values as reference only,
as I cannot 100% identify critically damping correctly by eyeballs.
Basically, the smaller the resistor value, the more is the damping.
How does the use of CRC benefit in terms of sonic performance?
I suggest you implement it and listen by yourself.
The difference is very obvious, I mean the improvement.
本帖最後由 johnme 於 2014-6-14 15:58 編輯
3 & 4. replacing 1u cap and adding TMC
In the original sigma design,
the Miller compensation cap is 3900pF.
3900pF may be too large.
First, a large Miller cap means low slew rate.
Second, a large Miller cap introduces early
negative feedback at HF. From the 1st graph,
you may see the attenuation at 10kHz is about -94dB.
If we reduce the value to 220pF,
the atenuation at 10kHz is now 111dB! (see figure 2)
However, you can see that the phase exceeds 180 degrees.
It is likely to oscillate above 1Mhz.
An unstable circuit is obtained.
In order to stabilize the PSU,
I use TMC method to replace
the Miller compensation method.
However, there is a trade-off, as mentioned by VJ.
The ripple rejection ratio drops about 7~12dB
depending on the frequency range concerned.
What if we replace the 1u output cap by a 100u Oscon? (see figure 3)
