bias spreader parasitic oscillation
When a transistor of a bias spreader is remotely located on a heat sink or power transistor, it is especially important to use appropriate bypassing to prevent the possibility of oscillation due to inductance in those connections. This is especially important in the case of a CFP bias spreader, since CFPs have a greater tendency to parasitic oscillation in the first place. Sometimes a collector-emitter bypass capacitor should be placed right at the remoted transistor.
Cheers,
Bob
When a transistor of a bias spreader is remotely located on a heat sink or power transistor, it is especially important to use appropriate bypassing to prevent the possibility of oscillation due to inductance in those connections. This is especially important in the case of a CFP bias spreader, since CFPs have a greater tendency to parasitic oscillation in the first place. Sometimes a collector-emitter bypass capacitor should be placed right at the remoted transistor.
Cheers,
Bob
I don't see any technical reason for this although it is supposed to be a form of voicing. If the two resonate, it might affect the feedback loop in various ways, probably not for the better. In all likelihood though the lytic inductance is not much larger than the film inductance and is insignificant if you look at the LR corner frequency with the feedback shunt resistor. And then there is the trace inductance to ground which is probably larger than the cap inductances.
10nF is 0.01uF, it's 1000:1.
BTW, in the LTSpice, LT supply the model for LTC6268, so either way, it's all theirs.
I have the CFP layout pretty much point to point. I did not see any sign of oscillation. I did put 0.1uF from collector of NPN to base, across the spreader, it made no difference. It's been rock solid for me. I also adjust TPOT2 to 0 to completely turn off the PNP, not change in turns of stability. It's just rock solid.
One thing, I use low Vbe high current transistors in the spreader, they are low speed transistors, they usually are quite forgiving. We don't need any speed in the spreader circuit, it's is thermal speed, not signal speed. The slower the transistor, the better. Just need to find low Vbe to get more headroom.
One thing, I use low Vbe high current transistors in the spreader, they are low speed transistors, they usually are quite forgiving. We don't need any speed in the spreader circuit, it's is thermal speed, not signal speed. The slower the transistor, the better. Just need to find low Vbe to get more headroom.
I don't intend it as a form of voicing. It is to make sure that the HF impedance stays low even if the electrolytic would have resonated by itself - this is extra insurance for FB stability at high frequencies. Maybe it is guilding the lily.
Cheers,
Bob
Agreed we don't need speed in the spreader. That was not the point. Most people use transistors in the spreader with 100MHz ft without even trying; your arrangement may be the exception.
Never underestimate the potential for parasitic oscillations that you may not even be able to see. Often, a parasitic oscillation may only be a brief burst when the circuit passes through a particular operating point.
Cheers,
Bob
Anything that would upset the stability of the feedback loop can be measured. I don't recall ever seeing any measurements of lytics that showed any resonances other than the SRF, which is swamped by the feedback shunt resistor, and can't be any more significant than the ground inductance to the feedback node.
If bypassing the feedback shunt cap has an effect, it is not due to it resonating. But a resonant bypass of the feedback shunt could perhaps be a sketchy form of compensation.
Bob,
If the circuit only goes through a very short period of oscillation what is the consequences of that? I can see a problem with an oscillation that is constant but what does the very short burst do that is a worry? Would it cause the circuit to fail or stress a component to failure or is it just showing that the circuit has stability problems?
If the circuit only goes through a very short period of oscillation what is the consequences of that? I can see a problem with an oscillation that is constant but what does the very short burst do that is a worry? Would it cause the circuit to fail or stress a component to failure or is it just showing that the circuit has stability problems?
You may know how radio detectors work. They conduct in one direction, so they develop a DC voltage when there are RF signals present.
Every transistor has 2 potential detectors in it. A small period of oscillation will probably shift operating points or cause a DC offset in the output, making an audible tick or POW depending on the nature of it. Or it may shift the output stage idle up dangerously.
If this happens for instance every time the output crosses zero, it will cause a very unpleasant kind of distortion, worse than crossover distortion.
Short bursts of slight oscillation probably won't hurt anything, but if sustained for long enough or if strong enough, it can fry the output zobel, which will further worsen stability, causing the problem to escalate.
Slight oscillation can also have interesting audible effects. If it is a relatively constant oscillation it may even sound very nice. However the FCC will not be happy about it.
Every transistor has 2 potential detectors in it. A small period of oscillation will probably shift operating points or cause a DC offset in the output, making an audible tick or POW depending on the nature of it. Or it may shift the output stage idle up dangerously.
If this happens for instance every time the output crosses zero, it will cause a very unpleasant kind of distortion, worse than crossover distortion.
Short bursts of slight oscillation probably won't hurt anything, but if sustained for long enough or if strong enough, it can fry the output zobel, which will further worsen stability, causing the problem to escalate.
Slight oscillation can also have interesting audible effects. If it is a relatively constant oscillation it may even sound very nice. However the FCC will not be happy about it.
Hi Alan0354,
If your mouth is always making sound, your ears cannot hear ... anything but the sound of your own voice. In this case it is your fingers.
I have a passing interest in what your project is, but that is it. I am very interested in what was going on in this thread before you arrived on the scene. Your project should move to it's own thread where it can be examined with the attention it deserves. You tend to steam roller over whatever is going on at the time. This is very much a selfish trait that you seem to have in spades.
Whatever you know in this field, it isn't the equal of several other folks around here. You do not have the right to halt the proceedings until someone kindly explains things to your satisfaction. Your voice has equal weight to those other members, not more. It's time you recognize that your way of interacting with others is not acceptable.
Please understand that there is no problem with asking relevant questions. The second you wish to debate the issue to the extent that you usually do, it is time to start your own thread. I'm pretty sure that if you follow those guidelines, you will notice that other folks around here will warm up to you.
Normally we don't moderate in threads we participate in, but for you I'll make an exception. You seem resistant to the idea you have to follow the rules.
In the list of rules we are all subject to, you will find a bit about thread jacking. If you continue to do this, you will probably find out what our version of a "time out" is. So please take this as an official warning. Be considerate of your fellow members and play nice in the sandbox.
-Chris
Okay Alan, to be blunt about it then.
If your mouth is always making sound, your ears cannot hear ... anything but the sound of your own voice. In this case it is your fingers.
I have a passing interest in what your project is, but that is it. I am very interested in what was going on in this thread before you arrived on the scene. Your project should move to it's own thread where it can be examined with the attention it deserves. You tend to steam roller over whatever is going on at the time. This is very much a selfish trait that you seem to have in spades.
Whatever you know in this field, it isn't the equal of several other folks around here. You do not have the right to halt the proceedings until someone kindly explains things to your satisfaction. Your voice has equal weight to those other members, not more. It's time you recognize that your way of interacting with others is not acceptable.
Please understand that there is no problem with asking relevant questions. The second you wish to debate the issue to the extent that you usually do, it is time to start your own thread. I'm pretty sure that if you follow those guidelines, you will notice that other folks around here will warm up to you.
Normally we don't moderate in threads we participate in, but for you I'll make an exception. You seem resistant to the idea you have to follow the rules.
In the list of rules we are all subject to, you will find a bit about thread jacking. If you continue to do this, you will probably find out what our version of a "time out" is. So please take this as an official warning. Be considerate of your fellow members and play nice in the sandbox.
-Chris
That's why there is no one sure method to test for stability and local parasitic oscillation. I do square wave test, make sure no excessive overshoot, more importantly, the ringing dies down after one ring, not continue for 3 or 4 cycles. Then I sweep frequency from low to as high as possible. Look for strange shape or some sensitive frequency. Then use screw driver to poke around to see whether you can start something. You can also increase or decrease the rail voltage, increase and decrease bias current. Also, turn the power on and off to see whether there is any burst of oscillation.
That's the reason I sweep to 2MHz, not that I need to run at that, but I want to see how the wave shape transition. If it is going smoothly from sine to triangular at slew rate limit and roll off gradually, it is more promising. some times, you see the waveform do some funny things when transition from sine to triangular, one should pay more attention.
Those are the few basic methods to try first, there might be more subtleties. As I said, I don't know how to put it any better, you see enough of these problems, you see a trend, it's like touching a nerve. If you see smooth transition, no abrupt changes, chances are good you have a stable circuit.
I also like to test OPS and IPS/VAS separately, I want to avoid dealing with GNFB. so far, I only look for local and parasitic oscillation in OPS, tame it first like what I did before putting them together.
Last edited:
Memories, those were the good old days...
What's wrong with this picture? Even though our precocious newbie only started posting back in Jan. 2015, perhaps not even the Somali pirates are such blantant hijackers...
What's wrong with this picture? Even though our precocious newbie only started posting back in Jan. 2015, perhaps not even the Somali pirates are such blantant hijackers...
I once had a prototype amplifier on the bench. It was measuring only 0.01% THD. That was more than I had expected. I finally discovered that it was due to a brief parasitic oscillation burst.
I designed what I called a "parasitic oscillation sniffer" that had some high-slope high-pass filtering in it that would kill a signal that was in the audio band. The sniffer was designed with RF wideband techniques with a bandwidth somewhere between 100-200MHz. It included a fairly sensitive RF detector at the end of the chain and a 555 trigger circuit that would illuminate an LED long enough to see it if there was an RF burst. It would connect to the output of the amplifier under test via a 50-ohm coax cable.
Unfortunately, this was just before I got in deep with the book, and I never completed the prototype.
The nice thing about this is that it can be used with sinewaves or even music. I often wonder how many commercial amplifiers would fare with this test.
Cheers,
Bob
Thank you Bob. I would think that these parasitic oscillations would be a good reason to have both an input filter going into the amp and also a practical bandwidth limit of an amp so those very high RF type things don't become an issue just so you can say you have some bandwidth that is unnecessary for a power amp or any real audio circuit. I guess it also goes with having a reasonable slew rate value that we really need and not shooting for the sky here. That is what makes me wonder why some want to go for extreme bandwidth and even using some really fast transistors in the output section of the amplifier.
Time to head back to reading your book and looking forward to your second edition.
Time to head back to reading your book and looking forward to your second edition.
The one in my class was covered in tattoos and solid gold earrings, sailed into university with a certificate of a merchant seaman school.
But he gave answers, in a question disguise.
The 1200 page thesis work he did at an offshore company in Monaco was rejected, he returned with 600 and refused to drop a single more, received his masters with honors.
Kean is the waving kid on this thread, and deservedly so.
(doesn't need to be told what ULGF is either)
I suppose if the question is addressed to Bob, then I don't really need to answer. I guess now I know why I get ignored sometimes. LOL! (no, no one ever taught me any manners)
Even if I can answer correctly, I don't have the amount of experience that Bob does backing my answers. So I guess it is rude to butt in when that kind of assurance is the aim of the question.
Even if I can answer correctly, I don't have the amount of experience that Bob does backing my answers. So I guess it is rude to butt in when that kind of assurance is the aim of the question.
There's always the one wise guy in class who already knows the material and wants to answer and ask all the questions so nobody else gets a chance. Takes over and becomes the teachers pet who everybody hates in the end. The people who want to ask questions hold back because they don't want to feel silly asking real questions that everyone else is also wondering about, the class learns little, the teacher thinks he has a room full of dolts and not much is really taught. Give me a class where everyone is asking questions, the teacher has to think before giving an answer and sometimes the teacher has to go back and do some serious thinking to answer some of these questions, that is where people really learn.
ps. Not directing any of this at you Kean, you do give well thought out answers or opinions. Some will just quash the conversation, others add to the discussion. I put you into the add category.
ps. Not directing any of this at you Kean, you do give well thought out answers or opinions. Some will just quash the conversation, others add to the discussion. I put you into the add category.
Last edited:
Actually that is not quite the point. An input filter does not do anything for a parasitic oscillation. In fact, often having a slow amplifier does not do anything for parasitic oscillations as long as it is properly conservatively compensated. Fast transistors are a good thing, as they create less excess phase lag in the forward path. Going for higher ULGF reduces HF distortion, it is not just to be able to say we have an amplifier with wider bandwidth.
In fact, as seen from the outside, the amplifier's bandwidth will usually be limited by the conventional input filter, whose -3dB frequency will often be 200-500kHz tops.
Fast output transistors are desirable to minimize secondary crossover distortion, sometimes called dynamic crossover distortion, caused by inability to sweep out the minority carriers in the base to shut the transistor off when output current is changing fast (current slew rate). Fast output transistors can sometimes increase the risk of local parasitic oscillations in the output stage, but this must be managed by good layout and circuit techniques like rail decoupling between output stages and appropriate base stopper resistors. It is also helps to have the Zobel network close to the output transistors to provide HF loading at high frequencies.
Of course, speed and bandwidth should never be stretched at the expense of stability, be it global or local.
Cheers,
Bob