HI....
I add a 1n+6.8k , parallel with output and input of error-amp.
And connect the feedback from after LC filter.
Parallel a 5.1k+100p with feedback resistor (47k).
Add two 3v zener diode ,back to back , parallel with error-amp.
I probe the output , it get a unstable output waveform.
But i do it like you designed in your website.
I don't know where did i do something wrong??
When i increase the zener diode's voltage to 3.6 , the maximum amplitude of output become more bigh to about 40v.
I will try to modify some component value to test what different !
I add a 1n+6.8k , parallel with output and input of error-amp.
And connect the feedback from after LC filter.
Parallel a 5.1k+100p with feedback resistor (47k).
Add two 3v zener diode ,back to back , parallel with error-amp.
I probe the output , it get a unstable output waveform.
But i do it like you designed in your website.
I don't know where did i do something wrong??
When i increase the zener diode's voltage to 3.6 , the maximum amplitude of output become more bigh to about 40v.
I will try to modify some component value to test what different !
titanchen68 said:
You seem to have copied the components I have used for the 'modified' circuit linnew.asc . However if you have not changed the output filter components as suggested then this will not work.
Do you have a copy of LTspice?
I would recommend that you download a copy. It is free and will let you simulate the circuits I have given by clicking on the LT logo on the web page.
In the meantime get the software and try these circuits.
testa.asc
This is your present circuit with your original output filter. Run it and plot V(vvea)/V(vac). This will give you the loop gain and phase. Notice that the amplitude curve does not reach 0dB until above 200KHz. This is much too high.
You might try adjusting your components to try and crossover at about 40KHz. Have a look at this circuit for some values.
testb.asc
Again run it and plot V(vvea)/V(vac). You will see that the crossover is at about 40KHz.
Once you have got LTspice then you can try changing some component values to see how they affect the response then it should become clearer what each one does.
It is still going to be the case that your present output filter might cause you problems unless it is loaded with a 4R load. Try running the second circuit with RLOAD set to 100K and you will see the filter resonance and how it is affecting the phase margin.
In the mean time I'm having a look at the sources of problems during clipping and it should be possible to clean things up without the clamps. However the circuit becomes slightly more complex. Sorry about this but I am learning as well.
Hope this helps
Cheers
DNA
So thank you a lot ,,,
The LTspice is a good tool , I'm trying to play it !
so , you mean the problem that the circuit i copied in your webside is the output LC filter.
Because the LC filter didn't reduce to 0db at 200kHz , causing the output get something wrong??
today i try the my circuit with clamp zener diode and without.
When i used the zener diode to clamp the error-amp output.
The Vout after LC-filter get a very good waveform , even clip.
But the output amplitude always clamped within about +/-30v , using 3v zener diode back to back, triangle wave amplitude is 4Vp.
When i remove the zener diode , the output amplitude after LC filter , rise up to +/-50v.
i input the music signal , but , if the input signal too large , output get some pulse , it sound like "Tizz" , very strange.
I will try 3.6v zener diode , maybe output amplitude will get higher.
The LTspice is a good tool , I'm trying to play it !
so , you mean the problem that the circuit i copied in your webside is the output LC filter.
Because the LC filter didn't reduce to 0db at 200kHz , causing the output get something wrong??
today i try the my circuit with clamp zener diode and without.
When i used the zener diode to clamp the error-amp output.
The Vout after LC-filter get a very good waveform , even clip.
But the output amplitude always clamped within about +/-30v , using 3v zener diode back to back, triangle wave amplitude is 4Vp.
When i remove the zener diode , the output amplitude after LC filter , rise up to +/-50v.
i input the music signal , but , if the input signal too large , output get some pulse , it sound like "Tizz" , very strange.
I will try 3.6v zener diode , maybe output amplitude will get higher.
Originally posted by titanchen68
So thank you a lot ,,,
The LTspice is a good tool , I'm trying to play it !
Yes, it might take a little time to get used to using it but it is very good for trying things out and exchanging ideas.
so , you mean the problem that the circuit i copied in your webside is the output LC filter.
Because the LC filter didn't reduce to 0db at 200kHz , causing the output get something wrong??
Yes, if you look at that circuit then you will see that the filter inductor is 32uH and the capacitor is 2u2. The damping components are 6u8 and 4R7. The filter corner frequency with those components is 20KHz. With your filter components it is 50KHz.
In the circuit I set things for an 80KHz loop crossover frequency. With your filter components the crossover becomes 200KHz. That's 2.5 times the original..... a factor of 50KHz/20Khz.
today i try the my circuit with clamp zener diode and without.
When i used the zener diode to clamp the error-amp output.
The Vout after LC-filter get a very good waveform , even clip.
But the output amplitude always clamped within about +/-30v , using 3v zener diode back to back, triangle wave amplitude is 4Vp.
I cannot be certain at the moment but I know you would expect something like 41V. It is likely that deadtime is robbing you of the extra output voltage.
When i remove the zener diode , the output amplitude after LC filter , rise up to +/-50v.
Sounds good.... did you make the component changes suggested in testb.asc ? If so then it seems that you finally have a circuit that is small signal stable so that should let you investigate further.
i input the music signal , but , if the input signal too large , output get some pulse , it sound like "Tizz" , very strange.
I will try 3.6v zener diode , maybe output amplitude will get higher.
Sounds.... well. Now you are getting into looking at other things. There are two things I have noticed which both relate to the relative timing of the mosfet drive compared to the PWM signal from the PWM comparator.
Intoduction of the deadtime, depending on how it is done, and delay times in both the IR2110 and gate interface add up to a delay that, when you are operating at high output levels will shift the switching transitions from being 'both' sides of the triangle wave to being after its peak (positive or negative).
At that point the waveforms are no longer properly related and the feedback starts misbehaving. I'm still thinking of some ways of improving that.
Cheers
DNA
So thank you a lot ,,,
The LTspice is a good tool , I'm trying to play it !
Yes, it might take a little time to get used to using it but it is very good for trying things out and exchanging ideas.
so , you mean the problem that the circuit i copied in your webside is the output LC filter.
Because the LC filter didn't reduce to 0db at 200kHz , causing the output get something wrong??
Yes, if you look at that circuit then you will see that the filter inductor is 32uH and the capacitor is 2u2. The damping components are 6u8 and 4R7. The filter corner frequency with those components is 20KHz. With your filter components it is 50KHz.
In the circuit I set things for an 80KHz loop crossover frequency. With your filter components the crossover becomes 200KHz. That's 2.5 times the original..... a factor of 50KHz/20Khz.
today i try the my circuit with clamp zener diode and without.
When i used the zener diode to clamp the error-amp output.
The Vout after LC-filter get a very good waveform , even clip.
But the output amplitude always clamped within about +/-30v , using 3v zener diode back to back, triangle wave amplitude is 4Vp.
I cannot be certain at the moment but I know you would expect something like 41V. It is likely that deadtime is robbing you of the extra output voltage.
When i remove the zener diode , the output amplitude after LC filter , rise up to +/-50v.
Sounds good.... did you make the component changes suggested in testb.asc ? If so then it seems that you finally have a circuit that is small signal stable so that should let you investigate further.
i input the music signal , but , if the input signal too large , output get some pulse , it sound like "Tizz" , very strange.
I will try 3.6v zener diode , maybe output amplitude will get higher.
Sounds.... well. Now you are getting into looking at other things. There are two things I have noticed which both relate to the relative timing of the mosfet drive compared to the PWM signal from the PWM comparator.
Intoduction of the deadtime, depending on how it is done, and delay times in both the IR2110 and gate interface add up to a delay that, when you are operating at high output levels will shift the switching transitions from being 'both' sides of the triangle wave to being after its peak (positive or negative).
At that point the waveforms are no longer properly related and the feedback starts misbehaving. I'm still thinking of some ways of improving that.
Cheers
DNA
1. I change the clamp zener diode to 3.9v , But i probe the output of the error-amp.
The output amplitude still only get 3v to saturate, I don't know why.
Doesn't the output of the error-amp limit by zener diode ??
Do you have any ideal about it ??
So , even i change higher voltage zener diode , my output amplitude still get about 30v peak to saturate.
(I sure the input signal higher than 4v.)
2. Now , I am not adding the dead time control.
Do you mean if i add the dead time control will help ??
3. Do you mean when the input signal near the triangle amplitude,
the triangle wave will get some phase shift ??
The output amplitude still only get 3v to saturate, I don't know why.
Doesn't the output of the error-amp limit by zener diode ??
Do you have any ideal about it ??
So , even i change higher voltage zener diode , my output amplitude still get about 30v peak to saturate.
(I sure the input signal higher than 4v.)
Genomerics said:
2. Now , I am not adding the dead time control.
Do you mean if i add the dead time control will help ??
3. Do you mean when the input signal near the triangle amplitude,
the triangle wave will get some phase shift ??
titanchen68 said:
Strange, are you running the TL082 shown in your circuit diagram as the error amplifier off +/-9V rails or have you connected it to +/-5V rails?
No, the deadtime will make things worse.
Not exactly, normally one mosfet is turned on (the other off) on the triangle upslope. The other mosfet is turned on (the previous off) on the triangle downslope. At high output levels deadtime and or delays move the switching from either side to one side and that causes the problems.
DNA
Genomerics said:
I supply the TL082 +/-9v.
Oh! In my mind , many people were suggesting me add the dead time control before IR2110.
So , why it will get worse ???
could you please just say more about it ...
I still get some confuse ,,, thanks!!!
Yes , I did control the triangle amplitude into +/- 4.2Vp.
Didn't higher than LT1016's power supply.
so , I think it's safe.
Today , i change the power supply for trangle wave generator to +/-5V.
Generate a 2.5Vp triangle wave. f=140KHz,
I use twice LC filter , L=20uH and C=1.5u.
Rf = 47K , Rin = 2.2k , Zener diode is 2v.
When the input amplitude is 2v,F=100Hz , the output amplitude after LC filter is 20Vp.
And increase the input amplitude to 2.5v, the output amplitude become 40Vp.
It looks like not a linear function.
And when the output amplitude is 40Vp, the output waveform looks like a triangle wave , Its seems get very high distortion.
Did i set wrong component value ???
Or something i can improve ???
Generate a 2.5Vp triangle wave. f=140KHz,
I use twice LC filter , L=20uH and C=1.5u.
Rf = 47K , Rin = 2.2k , Zener diode is 2v.
When the input amplitude is 2v,F=100Hz , the output amplitude after LC filter is 20Vp.
And increase the input amplitude to 2.5v, the output amplitude become 40Vp.
It looks like not a linear function.
And when the output amplitude is 40Vp, the output waveform looks like a triangle wave , Its seems get very high distortion.
Did i set wrong component value ???
Or something i can improve ???
titanchen68 said:
OK, it was just a thought since the data sheet for a TL082 shows the maximimum output with +/-5V supplies and a 2K load (your input resistor driven through the zeners) is +/-3V. Just looked too close to the problem you are getting.
Zener diodes, some types are worse than others, have a 'soft' turn on characteristic so don't make the best possible clamps. Yours might be limiting prematurely if you cannot supply enough input drive. I assume that if you leave them out then the error amplifier can swing much higher.
However you still need to be concerned about the LT1016 differential input voltage, see later.
Yes, you see many circuits where the gate drive resistor is used with a parallel diode to give a slow on and fast off signal to achieve the dead time. Unfortunately this relies on the mosfet device capacitances to achieve the appropriate delays and these are not tightly controlled parameters both during manufacture and operation.
As a result it should be better to use all the available drive current you have to minimise variations in switching times and implement the dead time control somewhere else in the circuit. That is best done before the inputs to the driver IC and ideally at circuit ground level.
The result should be more precise and controllable. My example with the level shifting and extra transistors isn't very good. I prefer the one later on on the web page where I use an extra pair of comparators to generate the appropriate signals.
I'll try, it might take a little bit of time to set some things up so I can show some pictures. Expect some more words later....
If you look on page 17 of the LT1016 data sheet from Linear Technologys website then they show a simplified schematic. D1 and D1 are diode connected transistors across the input terminals.
The base-emitter juction of a transistor can withstand a reverse voltage of about 5V before it breaks down, hence the +/-5V maximum rating. If the current is not limited then the device may be damaged. If your triangle wave was at +4V and your error amplifier output at -1V then you exceed that rating.
Ideally feedback might avoid that but it is not guaranteed under start up or transient conditions.
If you have not limited the input current then the LT1016 will be damaged. Even with some form of limiting its performance might be degraded.
Cheers
DNA
I don't understand why you say that ...
No , i still use the zener diode to clamp .
Because if i didn't use it , the output amplitude of error amp will higher than the trangle wave amplitude.
(Trangle wave is 2.5v , use +/-5v supply for all opamp)
It will get "Tizz" sound causing by some pulse in output of error amp.
If i use the zener clamp the output of error-amp, the output of error-amp will not get the unneed pulse .
do you mean add the dead time control before IR2110,
or in output of LT1016 ??
The ground level you said is 0v or -55v ???
do you mean two diode , one connect from input to +Vcc,
and one connect to -Vcc???
do you mean because i supply +/-5v to LT1016.
SO , the |V+| + |V-| must < 5v ???
the input of LT1016 , i have added a 10k before input.
do you mean that ??
Since you are having problems with the output level of your error amplifier being limited even when you use different value zener diodes then I was trying to find another source for the problem. I was just suggesting a possibility, which turned out to be wrong.
The only other cause I can think of is that low voltage zeners do not have very sharp characteristics and as a result your limit will be poorly defined and you may well see non-linearities.
You might try the 'clamp' shown here
clamp.gif.
It still has a soft characteristic but should not affect normal operation when it is 'not working'.
Where and how you implement the dead time is a matter of choice however it is likely that it will be 'easiest' to do at 0V since this is where the rest of the supplies for your other electronics are referenced.
When I said D1 and D1 I meant D1 and D2 and these are internal to the LT1016 as an input protection clamp. They are not in your circuit and do not need to be added.
Ignore me..... my fault, I didn't notice your 10K input resistors however such large values may give you problems with noise pickup.
Sorry for the delay. I am tidying up the website before it gets too messy.
Cheers
DNA
The only other cause I can think of is that low voltage zeners do not have very sharp characteristics and as a result your limit will be poorly defined and you may well see non-linearities.
You might try the 'clamp' shown here
clamp.gif.
It still has a soft characteristic but should not affect normal operation when it is 'not working'.
Where and how you implement the dead time is a matter of choice however it is likely that it will be 'easiest' to do at 0V since this is where the rest of the supplies for your other electronics are referenced.
When I said D1 and D1 I meant D1 and D2 and these are internal to the LT1016 as an input protection clamp. They are not in your circuit and do not need to be added.
Ignore me..... my fault, I didn't notice your 10K input resistors however such large values may give you problems with noise pickup.
Sorry for the delay. I am tidying up the website before it gets too messy.
Cheers
DNA
Hi
Just like you said , it looks still will work in linear area.
But i will try it , and add a dead time control before level shifting.
In these days i try , if i didn't use the zener diode be the clamp component.
When output amplitude too higher , the sound will get some strange "Tizz sound".
But i don't remember does it get clip .
If i use the zener diodes be the clamp.
If the zener diode voltage too low , the output amplitude will not get the highest amplitude output.
If the zener diode voltage too high , will lose the clamp ability.
Still get Tizz sound.
So , do you know the amp in commercial how to do to get stable output amplitude...??
Just like you said , it looks still will work in linear area.
But i will try it , and add a dead time control before level shifting.
In these days i try , if i didn't use the zener diode be the clamp component.
When output amplitude too higher , the sound will get some strange "Tizz sound".
But i don't remember does it get clip .
If i use the zener diodes be the clamp.
If the zener diode voltage too low , the output amplitude will not get the highest amplitude output.
If the zener diode voltage too high , will lose the clamp ability.
Still get Tizz sound.
So , do you know the amp in commercial how to do to get stable output amplitude...??
titanchen68 said:
It's me again, I was busy getting drunk.
Anyway... I've added what might be the last part to the end of your web page.
http://www.genomerics.org/titanchen/titanchen.html
I am getting lazy in my old age but the model is there along with the circuit if you want to try some things out for yourself. If I was going to do things this way then what you have there is close to what I would look to doing.
In fact, although my end goal won't be like that some of the control electronics may be useful for me.
It needs further investigation but hopefully there is enough there to help you out.
Cheers
DNA
Hello
I just read this post...Why do you use a hight level triangle wave? This will reduce the open loop gain of your amplifier...In my product, I use a triangle wave of about 100mv to have a high open loop gain, then I close feedback with very low volatge gain, about normally X12 for single ended amplifier and X24 for full bridge circuit. I limit the pulse with with narrow pulse generator added to the output of comparator. This will avoid the output coil saturation and higher the overal effiency of the amplifier. In this way, you will be too capable of detect the clipping of the amplifier when you compare the output of the comparator and the output with added pulse. It's the only way I have found to do not reach the 0-100% duty cycle limit, wich cause most of the lose on ''cracking'' soun due to inductor saturation.
About the high open loop gain, you have lot of advantage to do thing in that ways, so if you take feedback after output coil, you will avoid all probleme of frequency response dependent of load, faster compensation of phase, and lower distortion. A good stupid rull I use is the 5 and 95% limit on the pulse limiter. That give me maximum power versus effiency, without class d disadvantage.
I think taht you understand that this configuration is only good for fixed frequency amplifier...Not free running...
Have a nice night!
Fred
www.d-amp.com
I just read this post...Why do you use a hight level triangle wave? This will reduce the open loop gain of your amplifier...In my product, I use a triangle wave of about 100mv to have a high open loop gain, then I close feedback with very low volatge gain, about normally X12 for single ended amplifier and X24 for full bridge circuit. I limit the pulse with with narrow pulse generator added to the output of comparator. This will avoid the output coil saturation and higher the overal effiency of the amplifier. In this way, you will be too capable of detect the clipping of the amplifier when you compare the output of the comparator and the output with added pulse. It's the only way I have found to do not reach the 0-100% duty cycle limit, wich cause most of the lose on ''cracking'' soun due to inductor saturation.
About the high open loop gain, you have lot of advantage to do thing in that ways, so if you take feedback after output coil, you will avoid all probleme of frequency response dependent of load, faster compensation of phase, and lower distortion. A good stupid rull I use is the 5 and 95% limit on the pulse limiter. That give me maximum power versus effiency, without class d disadvantage.
I think taht you understand that this configuration is only good for fixed frequency amplifier...Not free running...
Have a nice night!
Fred
www.d-amp.com
fredos said:
I suppose a high amplitude triangle wave 'feels' good in terms of overall noise performance and linearity. Comparator delay and response times are also specied in terms of voltage step and overdrive. With higher drive levels these figures are reduced.
Like you say it does reduce the gain in that part of the system but that can be accounted for by the gain used in the error amplifier. There is a limit to the to the loop crossover frequency. I haven't pinned that one down yet but if that limit is designed to then the low triangle versus high triangle 'should' have the same overall loop gain.
I'm not sure what the advantage of the low closed loop gain you use would be but perhaps it's a matter of noise and overall distortion figures.
You can see from the webpage that I have been trying a few tricks to improve overload recovery. The most important one seems to be limiting the error amplifier output voltage.
Initially I suggested some form of clamp below the triangle wave peak but that is hard to specify and reduces the maximum output amplitude. My latest effort actively detects the peaks of the triangle wave and uses those signals to limit the error amplifier output above the triangle level.
This does allow the amplifier to go into clip but the recovery is much cleaner. In producing those signals I also get a pair of digital levels which can be used to ensure that switching is still happening even if the original PWM signal has gone. Unfortunately it introduces additional switching edges during normal operation. Wether that helps or not is something I will have to investigate.
Either way the result is the loop is effectively open or inactive during clip so the filter is free to do its own thing. That introduces another set of variable behaviour. I suppose the ideal would be to
have some additional loop that was active during clip so the amplifier would still have control of the output and filter. I've looked at that in the past but it has proved difficult to achieve.
Perhaps I should go back and look again. In the meantime I'm trying to come to terms with the overall behaviour so far. Hurts my head.
Have a nice night too!
DNA
I try to increase the +/-RAIL voltage to 85v.
But when i turn on the power , the output of mosfet can not get good square wave.
It seems can not turn on / off quickly.
And get very higher pulse at rise/down eage.
I use IRFS23N20D , i think it's Vdss higher enought for 85v.
Does anyone has the same experience.???
Because I want to get 400Wrms in 4ohm.
So , i think 56v is enought.
But if the output amplitude get to 56v , the output will get some strange Tizz sound.
So , "Genomerics" tought me add a zener diode in error amp.
It will clamp the output amplitude of error-amp.
But the outut amplitude after LC-filter will only get 40v highest.
So , if i increase the +/-RAIL voltage to 85v , and clamp the output of error-amp.
I think i will get 56v amplitude to get 400Watt rms.
But when i turn on the power , the output of mosfet can not get good square wave.
It seems can not turn on / off quickly.
And get very higher pulse at rise/down eage.
I use IRFS23N20D , i think it's Vdss higher enought for 85v.
Does anyone has the same experience.???
Because I want to get 400Wrms in 4ohm.
So , i think 56v is enought.
But if the output amplitude get to 56v , the output will get some strange Tizz sound.
So , "Genomerics" tought me add a zener diode in error amp.
It will clamp the output amplitude of error-amp.
But the outut amplitude after LC-filter will only get 40v highest.
So , if i increase the +/-RAIL voltage to 85v , and clamp the output of error-amp.
I think i will get 56v amplitude to get 400Watt rms.
Today i tye some test.
whatever the amplitude of triangle wave is ,,,
When the output amplitude of error-amp get into highest amplitude , will only equare to top amplitude of triangle wave.
Is that causing by comparator ???
I saw every designed here , if feedback connect from output of mosfet , they just across a low-pass filter and connect back to error-amp.
I want to know , Is that will not get the problem i meat ??
I get ...
1. when the output get to +/-RAIL , it get unstable.
2. when the output get clip , It get strange Tizz sound.
3. when the output get ciip , It get higher DC offset.
It that causing by wrong component value designed ??
whatever the amplitude of triangle wave is ,,,
When the output amplitude of error-amp get into highest amplitude , will only equare to top amplitude of triangle wave.
Is that causing by comparator ???
I saw every designed here , if feedback connect from output of mosfet , they just across a low-pass filter and connect back to error-amp.
I want to know , Is that will not get the problem i meat ??
I get ...
1. when the output get to +/-RAIL , it get unstable.
2. when the output get clip , It get strange Tizz sound.
3. when the output get ciip , It get higher DC offset.
It that causing by wrong component value designed ??
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.