Please bear with me guys, I know the feedback horse has been beaten to death so thoroughly so often that you probably groan from the sight of the thread title alone.
Still I think I have a valid question.
Assume a standard topology amp - input diff pair, gain stage, output stage. We know that the differential input signal at the input stage increases with frequency as the open loop gain decreases with frequency.
Another way to say this is that the amp itself (which of course always works open loop) needs more and more effective input signal for a given output level, as frequency increases.
This ever increasing diff input signal may presumably become so large that it could be causing internal overdrive and 'sticking' and gross distortion. Some pundits argue that this is the cause for 'bad' sound of feedback amps.
Now, I'm not saying this is true or not; and in any case it would depend on a lot of unmentioned factors. But maybe there is a simple test to find out if this internal overdrive can happen withing the foreseen use space of the amp:
If you run the amp open loop, with no feedback loop, you can check whether it can deliver the required (max) output level over the whole frequency band without obvious overdrive problems (which would immediately be seen in this situation, even with just a simple scope probably).
If the open loop amp passes this test, you can close the feedback loop and do not have to worry about this issue.
Does this make sense?
Jan
Still I think I have a valid question.
Assume a standard topology amp - input diff pair, gain stage, output stage. We know that the differential input signal at the input stage increases with frequency as the open loop gain decreases with frequency.
Another way to say this is that the amp itself (which of course always works open loop) needs more and more effective input signal for a given output level, as frequency increases.
This ever increasing diff input signal may presumably become so large that it could be causing internal overdrive and 'sticking' and gross distortion. Some pundits argue that this is the cause for 'bad' sound of feedback amps.
Now, I'm not saying this is true or not; and in any case it would depend on a lot of unmentioned factors. But maybe there is a simple test to find out if this internal overdrive can happen withing the foreseen use space of the amp:
If you run the amp open loop, with no feedback loop, you can check whether it can deliver the required (max) output level over the whole frequency band without obvious overdrive problems (which would immediately be seen in this situation, even with just a simple scope probably).
If the open loop amp passes this test, you can close the feedback loop and do not have to worry about this issue.
Does this make sense?
Jan
Yes agreed, the clipping behaviour will most probably (surely) be worse with feedback.
Jan
If I understand correctly, you have an output stage with a climbing demand of drive with higher frequenties.At some moment the demand exeed the capabilities of the diff.input+gain stage.Since the output doesn't follow no more there is no feedback to stop the diff. input to go up and the driver output goes skyhigh.
Playing around with composite amp's are you?
Mona
Playing around with composite amp's are you?
Mona
If I understand correctly, you have an output stage with a climbing demand of drive with higher frequenties.At some moment the demand exeed the capabilities of the diff.input+gain stage.Since the output doesn't follow no more there is no feedback to stop the diff. input to go up and the driver output goes skyhigh.
Playing around with composite amp's are you?
Mona
No the output stage is jus a follower, it's the miller on the gain stage.
You have an answer to my question?
Jan
Dosnt the differential input stage have a much higher diff. Input voltage at the highest frequencies when the feed back loop is closed. So testing it open loop is the least taxing?
If we look at the extreme when the freq is so high the open loop gain is unity, the differential input voltage will be close to the output voltage which will overload it.
If we look at the extreme when the freq is so high the open loop gain is unity, the differential input voltage will be close to the output voltage which will overload it.
Ah, the gain stage can't handle the high frequencies and the diff.input is incapable of giving the extra drive.Then you get intermodulation with the lower (audible) frequenties, probably not nice.Well limit the bandwide before entering the amp.Since the overdrive don't reach the output feedback isn't a solution.
Mona
Mona
Jan,
One of the ways you can test this is to look at the differential error signal of the front end stage. If you have an overload taking place anywhere in the forward path of the amplifier, the error signal will spike.
Further, any error signal from the front end stage should remain within the linear range of the input diff stage if the system is to remain linear with feedback.
One of the ways you can test this is to look at the differential error signal of the front end stage. If you have an overload taking place anywhere in the forward path of the amplifier, the error signal will spike.
Further, any error signal from the front end stage should remain within the linear range of the input diff stage if the system is to remain linear with feedback.
CFA is a topology where is quite easy to have enough Loop Gain even at 200 kHz to prevent to overload diff pair(complementary in this case) and low pass filter at the input will prevent unwanted high frequencies to enter the amp.
This is the Loop Gain of an CFA with HEXFET OPS. The Loop Gain at 200 kHz is 45 dB.
Damir
This is the Loop Gain of an CFA with HEXFET OPS. The Loop Gain at 200 kHz is 45 dB.
Damir
Attachments
I don't think so. For a given Vout, there is a required Vin on the two diff input pins. Feedback or not doesn't make a difference - the amp itself is always running ol of course.
Jan
Yes that's a good test in feedback mode. But what I am after is a test on the ol amp to predict that within the expected input signal level and frequency, it will have internal overshoot or not when I close the fb loop.
Example: I have an LM3886 and I test it with a freq sweep with a level that gives 20VRMS Vout at any audio frequency with a representative load (yes Vin then needs to rise with freq but the AP or dScope can Regulate that).
Suppose the amp passes this with no problems. Can I then say, when I close the fb loop, that there will never be an internal overload as long as I limit Vout to 20V at the freq range I did the test?
My reasoning is that since the amp itself will always be running open loop, feedback or not, seen from the +V and -V input pins of the input pair, feedback or not should make no difference in this context, assuming we stay within the tested Vout and freq range.
jan
Perhaps you are right, I don't see a real problem here.
To take the LM3886 as an example.From the datasheet you find that the output drops past 80kHz an the there gain is dropped to 20dB.
Output 25Vt with a gain of 10x the input is 2.5Vt, clearly the max. for that IC.So if you limit the diff. input to 5V there is no internal clipping.
Mona
Attachments
Measuring an amplifier circuit such as LM3886 without feedback isn't easy though. Do-able, but touchy, and will still usually need feedback at DC at least to keep things centered.
Jan, I'm assuming you are only concerned about the linearity aspects of the open loop amplifier, right? Because there is also stability to be dealt with, you might still get an internal overload if the amp+fb goes unstable (making its own input, in effect).
Actually this is one of the arguments I was looking for Bill. Now I know two conditions that may cause internal overload even if the OL amp would not: clipping and instability.
Is there more?
Jan
Member
Joined 2009
Paid Member
Ketje, you have compared at two different frequencies, 80kHz and 800kHz.
At 80kHz there is shown +40dB of open loop gain, requiring an input of 0.25Vpk for an output of 25Vpk.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.