ostripper said:This will not do, as I must know the "secret".
The first of the compensations is the NFB loop capacitor, I've
seen this one on the majority of class B amps, both DIY
and commercial.
Notice how it shoves the UG point into the MHZ range and
changes the phase considerably. Andy C. noted that It
produced a "2 pole response" and I should dump it. BUT ,
why is it used so widely in designs??
Hi OS,
The capacitor in the feedback loop is commonly called "lead compensation". If you do a plot of the loop gain of FA without this capacitor, you'll see its phase is zero at very low frequencies, then it curves down to -90 deg, where it stays for a while. Then as frequency increases, the phase begins to "sag", becoming more negative than -90 deg. What this lead compensation cap can do is "prop up" that phase, so it stays near -90 deg up to a much higher frequency than it would otherwise. It's not that this technique doesn't work, it's that the capacitor value required, given the 33k feedback resistor, may be impractically small. You can try it in the simulator. Try putting, say, 1pF in this spot and see if you can show this. If the phase starts to become less negative than -90 deg at some high frequency, it's overcompensated and a smaller cap is needed. The optimum value of this capacitor will keep this phase as close to -90 deg as possible over the widest possible bandwidth.
That's why it's called "lead compensation". The cap puts some phase lead in the loop gain to (at least partially) cancel out some of the natural phase lag in excess of 90 deg.
Well the idea is, if you're in the high-feedback camp, you can increase the unity loop gain frequency for a given phase margin. This gives more feedback, reducing distortion. The tradeoff is stability will be more sensitive to the load. We had earlier picked an 80 degree phase margin (that is, phase of loop gain is -100 degrees at the unity loop gain frequency). I believe the unity loop gain freq was around 500-600 kHz originally? So at what frequency is the phase of the loop gain -100 deg now? That would be the new target unity loop gain frequency if you wanted to play around with this a bit. This could be adjusted by reducing the comp cap or reducing the emitter degeneration in the input stage, or both.
Regarding sound, that's in the ear of the behearer, so I'll limit my comments to technical things.
Regarding sound, that's in the ear of the behearer
, so I'll limit my comments to technical things.
4Mhz+ at -100Db..
An externally hosted image should be here but it was not working when we last tested it.
I did
and with LPT Re=22R and 27p Cdom.. 4mhz+ at -100dbas well as the "little hump" of slight overcomp is gone.
An externally hosted image should be here but it was not working when we last tested it.
ostripper said:
Yikes! That's way too high for it to be stable with real-world loads - especially if there's no output inductor. Max unity loop gain frequency should be 1 MHz or so, though amps with FET output stages can go higher.
If you're not using an output inductor, less than 1 MHz is prudent for stability reasons.
Anyway, that demonstrates the lead compensation idea. Another alternative is to keep the unity loop gain frequency unchanged and just let the lead comp give you a better phase margin (better stability).
I am using an RCL network on a "daughtercard" but I agree
4mhz is too high.. Sim crashed with only slightly lower Cdom.
Here is 2pf lead with 68p Cdom.. .98mhz ULG -88db phase
from 15khz to 1.5mhz ruler flat..
you answered my primary question.. this is a stability factor..
A ??? ,how do you use the FFT utility on LT??
OS
4mhz is too high.. Sim crashed with only slightly lower Cdom.
Here is 2pf lead with 68p Cdom.. .98mhz ULG -88db phase
from 15khz to 1.5mhz ruler flat..
An externally hosted image should be here but it was not working when we last tested it.
you answered my primary question.. this is a stability factor..
A ??? ,how do you use the FFT utility on LT??
OS
That is the Zoble network.. Just in case some nasty Hf is present.
Very commonly used on just about all hifi amplifiers.
Why is it not on board.. 3 reasons.. 1. I have run the "U know what" out of both frugals without any Inductor/ zoble with no
problems, but for safety reasons or difficult loads, best to be safe
than sorry. 2. wanted the dirty zoble return to go back to main star ground,not board ground.
3. I did not put it on board as the inductor could
introduce noise into the circuit and I could build a much heftier
"industrial" version offboard..(see pix).
OS
Very commonly used on just about all hifi amplifiers.
Why is it not on board.. 3 reasons.. 1. I have run the "U know what" out of both frugals without any Inductor/ zoble with no
problems, but for safety reasons or difficult loads, best to be safe
than sorry. 2. wanted the dirty zoble return to go back to main star ground,not board ground.
3. I did not put it on board as the inductor could
introduce noise into the circuit and I could build a much heftier
"industrial" version offboard..(see pix).
OS
They are the zobel.. also known as the Boucherot cell.
when the output sees too inductive of a reactance
R1/C1 shorts out any HF instability before it can enter the
NFB loop of the amp.
On poorly designed amps I have felt R1 get hot (shorting
out HF instability no doubt so even in the physical world
it can be seen to actually 'do" something.
L1/R2 also have a job, they reduce damped ringing at
the output.
All four components together ensure a good interface
between the electrical(amp) and the physical (speaker)
In the repair of 100's of consumer amps these components
are nearly universally utilized with similar values...
OS
The transfer function of the diff amp-VAS is a function of the load impedance seen at each gain stage. R11/C12 provide an HF load to the diff pair, with R11 providing a zero or break at HF. Remember that the diff outputs are out of phase so the RC is just a balanced load rather than one from each to ground.
Many would probably argue that more CDOM comp would be better since the BW reduction is traded off for a more linear VAS. In fact, this type of compensation is more often seen in open loop amps such as those used in instrumentation where CDOM is not used. However, most of the analysis offered is based on the small signal response, and this type of comp _might_ make sense when looking to solve a large signal problem such as loop overshoot under dynamic conditions. It is hard to say since it is difficult to model the large signal behavior.
You should sim and observe the behavior when clipping at say 10 or 20 kHz. Look at the diff pair, VAS outputs, and also peak currents under fault conditions such as a shorted output. With precautions of course for real hardware.
Have you considered a compensation cap or RC across the emitters of the diff pair?
Pete B.
Many would probably argue that more CDOM comp would be better since the BW reduction is traded off for a more linear VAS. In fact, this type of compensation is more often seen in open loop amps such as those used in instrumentation where CDOM is not used. However, most of the analysis offered is based on the small signal response, and this type of comp _might_ make sense when looking to solve a large signal problem such as loop overshoot under dynamic conditions. It is hard to say since it is difficult to model the large signal behavior.
You should sim and observe the behavior when clipping at say 10 or 20 kHz. Look at the diff pair, VAS outputs, and also peak currents under fault conditions such as a shorted output. With precautions of course for real hardware.
Have you considered a compensation cap or RC across the emitters of the diff pair?
Pete B.
R29 - R12 bring back memories of Otala and Leach. They are from the time when it was thought that too much open loop gain was bad as suggested by the TIM camp. Leach later changed his view on this. The caps further reduce the open loop gain at HF, again, many would argue that more CDOM is the better way to go.
Don't know if you are familiar with Otala's work, but if you read it please also read Bob Cordell's counter position and Leach's later evolution of his Low TIM amp. The wise professor that I studied Audio Engineering under pointed out that TIM was nothing new, as video guys knew about it for years as slewing induced distortion. He didn't even have to think about it.
Pete B.
Don't know if you are familiar with Otala's work, but if you read it please also read Bob Cordell's counter position and Leach's later evolution of his Low TIM amp. The wise professor that I studied Audio Engineering under pointed out that TIM was nothing new, as video guys knew about it for years as slewing induced distortion. He didn't even have to think about it.
Pete B.
That is the first real answer to the purpose or r11/c12 I have
ever received in DIYland
... bravo. I can see its effects atHF along with changes in phase margin.
I tried that just now , not much affect at all.
These comps. can really be heard on the real thing. I"ll have to
"find" otala's book. I lean toward the increased Cdom ,as
shunting HF "after the fact" seems to be a crude approach.
Most fortunate, I have to be my own "professor"..
This amp runs good now but I just want to share the best
with my fellow DIY'ers and hash out the "voodoo" so
any explanation can be given Concisely without "I got it in a book"
or ??? "Dunno", etc... thanks
OS
I don't think Otala wrote a book on the subject. IIRC it was an IEEE transactions paper, then in the AES. I have it around here somewhere.
We discussed it a bit here:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=48873&highlight=
We discussed it a bit here:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=48873&highlight=
Pete,
Check your mail again.
Are you having stability problems with the amp?
When I changed the emitter resistors to 220R, mine ran well with the 47pF caps. This is outside in the open air though. As soon as you go to the trouble of nicely fitting it in a case, that's when the problems start (see my Patchwork )
Check your mail again.
Are you having stability problems with the amp?
When I changed the emitter resistors to 220R, mine ran well with the 47pF caps. This is outside in the open air though. As soon as you go to the trouble of nicely fitting it in a case, that's when the problems start (see my Patchwork
)- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.