[yejun]

I saw an implementation of low pass filter by adding a capacitor in feedback loop on headwize's project.
Why no one use this design to filter frequency higher than 200kHz?
Will the low pass make sound more clean?

 

[guzzler]

Quote:

Originally posted by yejun I saw an implementation of low pass filter by adding a capacitor in feedback loop on headwize's project. Why no one use this design to filter frequency higher than 200kHz? Will the low pass make sound more clean?

I suppose it's unnecessary cost and space, and most of us here are of the "small is beautiful" shape of mind... Also, theres not a whole lot of point as it shouldn't have any effect of the audible frequencies...

g

 

[Tomo]

Hello,

I beg to differ. It is not the cost and the space we are worried about.

Capacitor in the feedback path (Thank you for correcting) is very bad for certain type of opamps since it becomes short for radio frequency. This can cause severe instability. This applies to current feedback amplifiers.

Since many of you will simply "opamp-roll" without much care, I would NOT recommend that tactic. Someone will get into problems and it is simply a matter of time.

What I recommend instead is to attach Zobel Network / Snubber (and whatever else it is named) at the output. Although it will not prevent oscillation due to simply-bad-soldering or simply-bad-wiring, this will prevent oscillation due to RF-pickup from the outside and kills any RF spikes produced by ever more insane opamps we use.

Tomo

P.S. Seriously, if you are using insanely high power, wide bandwidth high speed opamps, you better make sure your amps don't oscillate. You might end up Jamming communications. Embarrassed by the fact that all the jamming is because of your amp will be the LEAST of your problems.

 

[yejun]

How is low pass at input stage?

 

[aos]

Tomo, I think he's saying about a capacitor in the FEEDBACK LOOP, not signal path.

And yejun, you're right, it is used purposefully by some people to cut the bandwidth of the amp. It will increase stability (being one of standard methods of compensation) and decrease noise by virtue of reduced bandwidth. It is especially true if you feed it signal that contains high frequency noise, such as EMI/RFI or stuff coming from a digital device such as D/A. I have noticed major improvements in noise floor when I was experimenting with my portable DAC, by adding a 10-20pF capacitor in the feedback loop.

HOWEVER, some people - ppl for example - maintain that doing this hurts the sound quality. And it makes sense, since you're reducing speed of the amplifier. What's the point of using a fast (and expensive) opamp to begin with if you're only going to slow it down? So if you can achieve stability and low noise without adding this capacitor (or keeping its value really low), so much the better.

 

[Tomo]

Aos -

That is correct. A capacitor in the feedback will become a short at high frequency. Now that the output is not properly terminated. This leaves space for RF pickup. This RF will go back into the amp through the feedback. Voltage FB opamps seem to care so little about that, but Current FB opamps do not like that.

(DAC stuff is entirely different because the source can produce that and you NEED it. Many people kill the garbles in DACs by using transformers and inductors also. But then I still wouldn't flirt with that idea, because that is not a good thing to do with CFB and there are excellent alternatives.)

Tomo

 

[aos]

You do not compensate a current feedback opamp with a capacitor in the feedback loop! In fact you don't have to - you do it by changing the value of the feedback resistor. CFB must have a resistor in the feedback loop- it can't be 0. Changing its value changes the bandwidth, and details are in particular device datasheets.

By the way I don't actually use feedback cap in the final portable DAC, because I managed to optimize layout and schematics of the digital part which have cut the noise at the amp input by two orders of magnitude (10mV to 100uV range). However almost all DACs use analog filters and in most configurations a cap in the feedback loop is part of that. But that's a filter, not an amp.

 

[ppl]

aos The reason for not using a cap around the feedback in a CFB op amp is the fact that with this topology the bandwidth of the device is determond by the Impedance from the out to inverting in. the lower the impedance the higher the bandwidth. this is why manufactures recomend a minimum resistor value for the feedback resistor. never ever put a capacitor in the feedback path of a CFB op amp.

next a capacitor around a voltage feedback op amp, called phase lead compensation will in addition to what tomo stated place the op amp in unity gain at some point and this will make any non unity gain stable opamp become unstable. so this technique can only be realised with unity gain op amps and unity gain opamps are internaly compensated anyway so why bother with external compensation. more over this cap will only improve stability over a limited range of capacitive loads at others it will present the load capacitence upon the sensitive non inverting input and cause more stability problems. some try and lessen these effects by placing a Resistor in series with the cap so as to have some real resistence or impedance and provide isolation of the capacitive load from the inverting input.

I could go on and include about the sound However if the above is not good enought reason to avoid this technique and obtain stability naturaly with proper layout and component selection, then, Hay! what can i say, try it for your self

 

[aos]

>> next a capacitor around a voltage feedback op amp, called phase lead compensation will in addition to what tomo stated place the op amp in unity gain at some point and this will make any non unity gain stable opamp become unstable

Wait, doesn't adding a cap do the opposite thing, i.e. make a non-unity-gain-stable opamp actually stable? Because you're introducing a pole that makes the gain go unity earlier than without the cap, where the phase margin is still high enough?

In general case, with capacitive load, things do get screwy as you said. A nice example is compensation of linear voltage regulators - I read an app note about this recently; most of them require output capacitor to be in a particular range of capacitance and ESR because as the poles/zeros keep moving around you can end up with an unstable loop. I personally like to isolate opamp outputs with a resistor or something so that they don't drive capacitive loads directly.

 

[ppl]

aos> No placing a cap around a non unity gain stable opamp will cause instability because at some point the cap will reduce the Open loop gain to unity and if this is in a area that is in wrong part of the transfer funtion the gain will not be reduced to unity prior to the Phase shift becoming excessive caused by the introduction of another pole in the mix. remember at High frequencies this cap liooks look a short circuit. the Phase lead caused by the cap will not improve the Phase margin enough to offset this.

there is alot of data on this subject in plenty of application notes. and most recomend that the cap be used to compensate for excessive capacitence on the inverting input, or to trim the transient response of a allready stable amp. and this type of compensation is not to be used to make an unstable amp stable.

it is true Alot of notable Analog designers like this cap but even those that do advocate a real small value <10pF.

For me the fewer capacitors in the signal path the better, No coupling capacitors, no intentional added inline capacitence.

 

[Tomo]

Hey,

I remember reading about this in an application note from some company. I seem to have misplaced it. Do you remember which one it was?

Tomo

 

[aos]

Tomo: read about what? Compensation of voltage or current feedback amps? If you tell me which one, I'm sure I can dig something out from my library of pdf's. There are some nice app notes I've read on these subjects.

 

[aos]

Just found this one - contains also a Java applet to demonstrate opamp stability, and has lots of Bode diagrams to show different methods of compensation.

http://www.analog.com/Analog_Root/st...stability.html

 

[Tomo]

Wow,

Aos, thank you. It also have many java calculators for many other things. I need few days to play with them all.

Tomo

 

[ppl]

thanks for the Link Tomo the Noise gain compensation IMHO really Hurt the sound by making the High end sound Bright and seams to make the Audio signal modulate the noise. As much as i dislike the Phase lead method the Noise gain mehod is far worse to my ears as rather than darken and slow down the apperent speed of the Music like Phase lead dose the Noise gain method takes the high end the other way.

This is still a real nice tool will have to book mark that one could save alot of time in the Future