QCW coil
Dr. ISOTOP, Thu Dec 22 2011, 01:13PMWell, the start of one, anyway. I've decided to jump on the bandwagon and build one of these.
Blog post here:
Currently tested at low powers on a resistive load; whether the controller can deal with real Tesla coil load conditions remains to be seen.
Re: QCW coil
Gregory, Thu Dec 22 2011, 10:24PM
Hello man. I'm making one too, and what I learned is that you need less LC capacitance to get a more fast response of the output. Linear control will not work too, now I'm using a delta modulator same of hysteresis modular, bang bang modulator inspired on steve/eric modulator. Good look!
Gregory, Thu Dec 22 2011, 10:24PM
Hello man. I'm making one too, and what I learned is that you need less LC capacitance to get a more fast response of the output. Linear control will not work too, now I'm using a delta modulator same of hysteresis modular, bang bang modulator inspired on steve/eric modulator. Good look!
Re: QCW coil
Dr. ISOTOP, Fri Dec 23 2011, 01:41AM
I don't see why less capacitance will improve response times; so long as your switches can cope with charging the output cap through the inductor that quickly it should be fine, especially with delta modulation.
I was designing this for a coil that would pull 600Apk, hence the air-cored inductor and huge output cap. Has it been experimentally confirmed that QCW's do not work at <300KHz? If so, I guess I need to scale things down a bit.
Dr. ISOTOP, Fri Dec 23 2011, 01:41AM
Gregory wrote ...
Hello man. I'm making one too, and what I learned is that you need less LC capacitance to get a more fast response of the output. Linear control will not work too, now I'm using a delta modulator same of hysteresis modular, bang bang modulator inspired on steve/eric modulator. Good look!
Hello man. I'm making one too, and what I learned is that you need less LC capacitance to get a more fast response of the output. Linear control will not work too, now I'm using a delta modulator same of hysteresis modular, bang bang modulator inspired on steve/eric modulator. Good look!
I don't see why less capacitance will improve response times; so long as your switches can cope with charging the output cap through the inductor that quickly it should be fine, especially with delta modulation.
I was designing this for a coil that would pull 600Apk, hence the air-cored inductor and huge output cap. Has it been experimentally confirmed that QCW's do not work at <300KHz? If so, I guess I need to scale things down a bit.
Re: QCW coil
Gregory, Fri Dec 23 2011, 02:45AM
Well, I think that people have already tryed to use a < 300khz ressonator at vttc or half retified sstcs and what they got is a "normal spark" no the sword like spark.
Gregory, Fri Dec 23 2011, 02:45AM
as it been experimentally confirmed that QCW's do not work at <300KHz
Well, I think that people have already tryed to use a < 300khz ressonator at vttc or half retified sstcs and what they got is a "normal spark" no the sword like spark.
Re: QCW coil
Steve Ward, Sun Jan 01 2012, 03:28AM
About bandwidth and power supplies using LC output filters. The LC resonance frequency (called a pole, actually its a double pole at that freq) limits the bandwidth of the power supply, after the Fres, you got a -20dB per decade drop off, unless you compensate for it some way. So this is mainly how i'd look at sizing my LC components, or at least their Fres. The other thing to consider is the output impedance of the power supply vs the input impedance of the tesla driver. You dont want the bus voltage to sag too much based on the intverter current, so you need some minimal amount of capacitance.
About control schemes: you can do a more classical power supply control (a "PWM control"), but the compensation for the resonant output filter is required. Look for documentation on forward and buck converter controls from TI, fairchild, etc... There are always short comings of this type of control with relation to the filter, and you may just need to avoid feeding the system any control signal with frequency near the Fres as this could make the thing go unstable. This is why i opted for non-linear control laws for a high speed buck conversion. The hysteretic control is both unstable (it oscillates at its own, much higher carrier frequency, which is not fixed but controllable) and very stable (as in, tracks a reference signal) at the same time. It merely turns the output on and off so that on average it meets its goals, but at any instant in time, its either too high or too low. If you make it self-oscillate faster, the regulation is improved for a given output filter. Loading the thing can alter the self oscillation frequency too, but i dont think ive ever seen it do anything undesirable. This kind of control will gladly accept references that contain frequency content above the LC resonance, and the frequency response of such a controller is simply dominated by the LC network, except its a roll off instead of a big peak at the Fres, so we say its stablized.
Steve Ward, Sun Jan 01 2012, 03:28AM
About bandwidth and power supplies using LC output filters. The LC resonance frequency (called a pole, actually its a double pole at that freq) limits the bandwidth of the power supply, after the Fres, you got a -20dB per decade drop off, unless you compensate for it some way. So this is mainly how i'd look at sizing my LC components, or at least their Fres. The other thing to consider is the output impedance of the power supply vs the input impedance of the tesla driver. You dont want the bus voltage to sag too much based on the intverter current, so you need some minimal amount of capacitance.
About control schemes: you can do a more classical power supply control (a "PWM control"), but the compensation for the resonant output filter is required. Look for documentation on forward and buck converter controls from TI, fairchild, etc... There are always short comings of this type of control with relation to the filter, and you may just need to avoid feeding the system any control signal with frequency near the Fres as this could make the thing go unstable. This is why i opted for non-linear control laws for a high speed buck conversion. The hysteretic control is both unstable (it oscillates at its own, much higher carrier frequency, which is not fixed but controllable) and very stable (as in, tracks a reference signal) at the same time. It merely turns the output on and off so that on average it meets its goals, but at any instant in time, its either too high or too low. If you make it self-oscillate faster, the regulation is improved for a given output filter. Loading the thing can alter the self oscillation frequency too, but i dont think ive ever seen it do anything undesirable. This kind of control will gladly accept references that contain frequency content above the LC resonance, and the frequency response of such a controller is simply dominated by the LC network, except its a roll off instead of a big peak at the Fres, so we say its stablized.
Re: QCW coil
Goodchild, Sun Jan 01 2012, 05:26AM
I wanted to add a little of my experience in as well.
<300KHz operation will be non-sword sparks.
For me 320KHz seemed to be the magic frequancy where sparks became nice swords.
The output capacitance is important in a couple of ways, it dose have to be sized properly because it plays a larger part in the performance of the buck to regulate. As Steve said the cap should be big enough not to have the inverter sag, but at higher power levels the use of a feed-forword capacitor in the feedback network will probably be needed to speed up the switching frequancy to the correct level. However the cap doesn't have to handle supper large amounts of current, not as much as the inductor at least.
Lately however I have been playing around with a new kind of delta modulation that uses internal integration rather than relying on the integration of the output capacitor of the buck. I'm hoping that this helps with both noise immunity and also making the switching frequency controlled internally rather than being dependent on the output C and load.
I would also like to point out that the origonal BB modulator I used was Steve ward's design, and not mine. Steve's deserves the credit for a well desighned modulator.
Goodchild, Sun Jan 01 2012, 05:26AM
I wanted to add a little of my experience in as well.
<300KHz operation will be non-sword sparks.
For me 320KHz seemed to be the magic frequancy where sparks became nice swords.
The output capacitance is important in a couple of ways, it dose have to be sized properly because it plays a larger part in the performance of the buck to regulate. As Steve said the cap should be big enough not to have the inverter sag, but at higher power levels the use of a feed-forword capacitor in the feedback network will probably be needed to speed up the switching frequancy to the correct level. However the cap doesn't have to handle supper large amounts of current, not as much as the inductor at least.
Lately however I have been playing around with a new kind of delta modulation that uses internal integration rather than relying on the integration of the output capacitor of the buck. I'm hoping that this helps with both noise immunity and also making the switching frequency controlled internally rather than being dependent on the output C and load.
I would also like to point out that the origonal BB modulator I used was Steve ward's design, and not mine. Steve's deserves the credit for a well desighned modulator.
Re: QCW coil
teravolt, Mon Jan 02 2012, 10:16AM
happy new years all, isn't there a balence of how much capacitance you can hang on the output inverter like Gregory specified? if there is to little capacitance it won't it won't filter the PWM and to much will be a burden and may afect the rate of rise that the inverter can handle? I would think that you wouldn't want any sort of resonance in the filter.
Dr. Isotop I noticed that you said that you are using a MBED to generate the controle hows that working out? I know that it has a 100Mhz core and I would think that it would be a exalent controle device. that is a interesting method of modulation that you have used to track the input with the output VS standard delta modulation. hope it works.
I to want to try building the ilusive QCW and I have gone with the SLR it is still in the consrution phase. Eric gave me the idea through this simulation inadvertantly.
I to want to rectifie and filter the output. I'm hoping 70uf will be enough to filter but not alow the output to be the ramp wave to track the input. right now it is open loop but may have to closed to controle nonlinearity like you have pausibly done with the MBED.
teravolt, Mon Jan 02 2012, 10:16AM
happy new years all, isn't there a balence of how much capacitance you can hang on the output inverter like Gregory specified? if there is to little capacitance it won't it won't filter the PWM and to much will be a burden and may afect the rate of rise that the inverter can handle? I would think that you wouldn't want any sort of resonance in the filter.
Dr. Isotop I noticed that you said that you are using a MBED to generate the controle hows that working out? I know that it has a 100Mhz core and I would think that it would be a exalent controle device. that is a interesting method of modulation that you have used to track the input with the output VS standard delta modulation. hope it works.
I to want to try building the ilusive QCW and I have gone with the SLR it is still in the consrution phase. Eric gave me the idea through this simulation inadvertantly.
I to want to rectifie and filter the output. I'm hoping 70uf will be enough to filter but not alow the output to be the ramp wave to track the input. right now it is open loop but may have to closed to controle nonlinearity like you have pausibly done with the MBED.
Re: QCW coil
Gregory, Tue Jan 03 2012, 12:10AM
You need some closed loop control. The bang bang control is the bestest for that situation because is very simple to implement, very stable, and will do a good job with the fast transients of the qcw bridge. For SLR maybe a PD control will work very well
Gregory, Tue Jan 03 2012, 12:10AM
You need some closed loop control. The bang bang control is the bestest for that situation because is very simple to implement, very stable, and will do a good job with the fast transients of the qcw bridge. For SLR maybe a PD control will work very well
Re: QCW coil
Dr. ISOTOP, Tue Jan 03 2012, 04:56AM
Happy new year!
With an LC filter like on the buck, too much capacitance can be compensated for with less inductance. I use 100 uH + 400 uF so the inductor has less wire and less Vdrop, but as Eric pointed out on my blog, too much capacitance causes an impedance mismatch with the coil, reducing the effectiveness of regulation.
The Mbed has worked well so far but this has only be with a resistive load. The only thing I don't like about it is that its rather fragile and expensive to replace. Also, it has the risk of latching up around Tesla coils and dying...optocoupling the feedback signal would probably be a good idea.
A good middle ground would probably be analog comparators to do the comparisons with the bounds, and a micro to do the actual control loop. A nice thing about micros when fiddling with control loops is that you can change the parameters or the control scheme by just changing a couple of defines in the code; no need for pot-twiddling or new boards
Dr. ISOTOP, Tue Jan 03 2012, 04:56AM
teravolt wrote ...
happy new years all, isn't there a balence of how much capacitance you can hang on the output inverter like Gregory specified? if there is to little capacitance it won't it won't filter the PWM and to much will be a burden and may afect the rate of rise that the inverter can handle? I would think that you wouldn't want any sort of resonance in the filter.
Dr. Isotop I noticed that you said that you are using a MBED to generate the controle hows that working out? I know that it has a 100Mhz core and I would think that it would be a exalent controle device. that is a interesting method of modulation that you have used to track the input with the output VS standard delta modulation. hope it works.
I to want to try building the ilusive QCW and I have gone with the SLR it is still in the consrution phase. Eric gave me the idea through this simulation inadvertantly.
I to want to rectifie and filter the output. I'm hoping 70uf will be enough to filter but not alow the output to be the ramp wave to track the input. right now it is open loop but may have to closed to controle nonlinearity like you have pausibly done with the MBED.
happy new years all, isn't there a balence of how much capacitance you can hang on the output inverter like Gregory specified? if there is to little capacitance it won't it won't filter the PWM and to much will be a burden and may afect the rate of rise that the inverter can handle? I would think that you wouldn't want any sort of resonance in the filter.
Dr. Isotop I noticed that you said that you are using a MBED to generate the controle hows that working out? I know that it has a 100Mhz core and I would think that it would be a exalent controle device. that is a interesting method of modulation that you have used to track the input with the output VS standard delta modulation. hope it works.
I to want to try building the ilusive QCW and I have gone with the SLR it is still in the consrution phase. Eric gave me the idea through this simulation inadvertantly.
I to want to rectifie and filter the output. I'm hoping 70uf will be enough to filter but not alow the output to be the ramp wave to track the input. right now it is open loop but may have to closed to controle nonlinearity like you have pausibly done with the MBED.
Happy new year!
With an LC filter like on the buck, too much capacitance can be compensated for with less inductance. I use 100 uH + 400 uF so the inductor has less wire and less Vdrop, but as Eric pointed out on my blog, too much capacitance causes an impedance mismatch with the coil, reducing the effectiveness of regulation.
The Mbed has worked well so far but this has only be with a resistive load. The only thing I don't like about it is that its rather fragile and expensive to replace. Also, it has the risk of latching up around Tesla coils and dying...optocoupling the feedback signal would probably be a good idea.
A good middle ground would probably be analog comparators to do the comparisons with the bounds, and a micro to do the actual control loop. A nice thing about micros when fiddling with control loops is that you can change the parameters or the control scheme by just changing a couple of defines in the code; no need for pot-twiddling or new boards
Re: QCW coil
teravolt, Tue Jan 03 2012, 05:52AM
hi Gregory did you mean PID controller? I was thinking that it may be possible to measure the vout and feed it back through a op amp circuit to my VCO. it also may be possible to do this with Dr. Isotop's delta modulator. I thought about using a sg3525 then realized that I need a vco for the type of modulation I want to do even though both would work. I set the SLR up so that I am sending energy packets into a resonant circuit at a rate of my choice then rectifying and filtering it to average the total. it will be hard to tell how well it will track with a dynamic load like a tesla.there still is a lot of tweaking. A thought, since the modulator what ever type it is, is supposed to ramp the current or energy to a tesla. How about instead of treating the QCW as two separate systems treat it as one system and make it closed loop. Would it be possible to enter a voltage at the front end and make the secondary current track so a sparks current can be controlled in a more linear manor? would it mater?
teravolt, Tue Jan 03 2012, 05:52AM
hi Gregory did you mean PID controller? I was thinking that it may be possible to measure the vout and feed it back through a op amp circuit to my VCO. it also may be possible to do this with Dr. Isotop's delta modulator. I thought about using a sg3525 then realized that I need a vco for the type of modulation I want to do even though both would work. I set the SLR up so that I am sending energy packets into a resonant circuit at a rate of my choice then rectifying and filtering it to average the total. it will be hard to tell how well it will track with a dynamic load like a tesla.there still is a lot of tweaking. A thought, since the modulator what ever type it is, is supposed to ramp the current or energy to a tesla. How about instead of treating the QCW as two separate systems treat it as one system and make it closed loop. Would it be possible to enter a voltage at the front end and make the secondary current track so a sparks current can be controlled in a more linear manor? would it mater?
Re: QCW coil
Gregory, Tue Jan 03 2012, 11:29PM
This is one thing that I was saying.
Yeah! it will work, but I THINK that you need some derivative and integrative therm to get ir right. The drsstc bridge has very fast transients and the modulator need a very fast response. Compensation on feedback, like a simple capacitor, can be see as a small integral/derivative therm that will make your system get a very better response.
Gregory, Tue Jan 03 2012, 11:29PM
Eric pointed out on my blog, too much capacitance causes an impedance mismatch with the coil, reducing the effectiveness of regulation.
This is one thing that I was saying.
I was thinking that it may be possible to measure the vout and feed it back through a op amp circuit to my VCO
Yeah! it will work, but I THINK that you need some derivative and integrative therm to get ir right. The drsstc bridge has very fast transients and the modulator need a very fast response. Compensation on feedback, like a simple capacitor, can be see as a small integral/derivative therm that will make your system get a very better response.
Re: QCW coil
Dr. ISOTOP, Wed Jan 04 2012, 12:31AM
How does the math for the impedance matching work?
In particular, how do you calculate the input impedance of the bridge?
Dr. ISOTOP, Wed Jan 04 2012, 12:31AM
How does the math for the impedance matching work?
In particular, how do you calculate the input impedance of the bridge?
Re: QCW coil
Gregory, Wed Jan 04 2012, 03:09AM
I think that the impedance isn't of the bridge, because she aways is connecting the primary LC to the modulator output so she "isn't here as a load". So, I THINK that the impedance is the impedance of the primary tank side SQRT(L / C).
Gregory, Wed Jan 04 2012, 03:09AM
In particular, how do you calculate the input impedance of the bridge?
I think that the impedance isn't of the bridge, because she aways is connecting the primary LC to the modulator output so she "isn't here as a load". So, I THINK that the impedance is the impedance of the primary tank side SQRT(L / C).
Re: QCW coil
Goodchild, Wed Jan 04 2012, 03:17AM
You know my delta modulator consist of a single comparator and a single 5V supply, there is really no need to put a 100MHz MCU in there.
All you need for a good bang controller is a comparator with a small amount of hysteresis (say 5mV to 20mV). You feed the ref ramp into one input and the output of your feedback divider into the other and they are internal subtracted and squared off for the gate drive. The speed of integration (switching speed) is controlled simply by the LC (mainly the inductance of the L and ESR of the cap), load and delays in the control circuitry including the hysteric window on the comparator. Switching speed can also be manipulated by the addition of a feed-forword capacitor and series resistor before the HV feedback divider.
My system is a little fancier in that it internally generates the integration making switching speed independent of the output filter parasitics and the load. But it still only used one comparator
The use of a fast precision comparator is a must, any switching error in the comparator can result in regulation error and unwanted operation. The LM311 is NOT a precision comparator
It seems like a lot of EE's out there stay away from bang bang control because of the dependance on temp ESR and load dynamics controlling switching speed. But it's a really awesome control topology because of it's simplisity and ability to go from 0 to 100% duty cycle giving it excellent transieant response. No control loop compensation has to be added ether which makes it just all the more awesome.
Goodchild, Wed Jan 04 2012, 03:17AM
You know my delta modulator consist of a single comparator and a single 5V supply, there is really no need to put a 100MHz MCU in there.
All you need for a good bang controller is a comparator with a small amount of hysteresis (say 5mV to 20mV). You feed the ref ramp into one input and the output of your feedback divider into the other and they are internal subtracted and squared off for the gate drive. The speed of integration (switching speed) is controlled simply by the LC (mainly the inductance of the L and ESR of the cap), load and delays in the control circuitry including the hysteric window on the comparator. Switching speed can also be manipulated by the addition of a feed-forword capacitor and series resistor before the HV feedback divider.
My system is a little fancier in that it internally generates the integration making switching speed independent of the output filter parasitics and the load. But it still only used one comparator
The use of a fast precision comparator is a must, any switching error in the comparator can result in regulation error and unwanted operation. The LM311 is NOT a precision comparator
It seems like a lot of EE's out there stay away from bang bang control because of the dependance on temp ESR and load dynamics controlling switching speed. But it's a really awesome control topology because of it's simplisity and ability to go from 0 to 100% duty cycle giving it excellent transieant response. No control loop compensation has to be added ether which makes it just all the more awesome.
Re: QCW coil
Dr. ISOTOP, Wed Jan 04 2012, 04:47AM
I briefly thought about implementing the controller this way, but decided against it. I suppose my excuse for using a fancy micro in the controller is so I can implement "optional" features like overcurrent detection without reworking the entire circuit. I really want to avoid the fault condition where the bridge fails short, which will take out the buck and god knows what other bits of silicon down the line.
Dr. ISOTOP, Wed Jan 04 2012, 04:47AM
Goodchild wrote ...
You know my delta modulator consist of a single comparator and a single 5V supply, there is really no need to put a 100MHz MCU in there.
You know my delta modulator consist of a single comparator and a single 5V supply, there is really no need to put a 100MHz MCU in there.
I briefly thought about implementing the controller this way, but decided against it. I suppose my excuse for using a fancy micro in the controller is so I can implement "optional" features like overcurrent detection without reworking the entire circuit. I really want to avoid the fault condition where the bridge fails short, which will take out the buck and god knows what other bits of silicon down the line.
Re: QCW coil
Goodchild, Wed Jan 04 2012, 03:12PM
Well even with a MCU you would still have to add some sort of extra circuit to sense the current in the first place. Even at 100MHz and interrupts your OCD response time will probably still be in the hundreds of u second range unless your uC spends all it's time checking the OCD. A comparator is really the better way to go for speed, as you want to get that silicon shut off in <10uS if possible. MBED also runs a rater complex firmware and library architecture that will eat up that 100MHz rather fast making it not much faster than a 8bit 16MHz uC.
In my opinion 100MHz uC is not the right way to go with this sort of controller, it's will be a lot a slower than a single dedicated comparator and a lot more susceptible to noise and such.
If you wanted a configurable option that was fast and had better noise immunity you should look into a FPGA or CPLD. There is a reason uCs are not widely used for SMPS control.
Eric
Goodchild, Wed Jan 04 2012, 03:12PM
Dr. Isotop wrote ...
I briefly thought about implementing the controller this way, but decided against it. I suppose my excuse for using a fancy micro in the controller is so I can implement "optional" features like overcurrent detection without reworking the entire circuit. I really want to avoid the fault condition where the bridge fails short, which will take out the buck and god knows what other bits of silicon down the line.
Goodchild wrote ...
You know my delta modulator consist of a single comparator and a single 5V supply, there is really no need to put a 100MHz MCU in there.
You know my delta modulator consist of a single comparator and a single 5V supply, there is really no need to put a 100MHz MCU in there.
I briefly thought about implementing the controller this way, but decided against it. I suppose my excuse for using a fancy micro in the controller is so I can implement "optional" features like overcurrent detection without reworking the entire circuit. I really want to avoid the fault condition where the bridge fails short, which will take out the buck and god knows what other bits of silicon down the line.
Well even with a MCU you would still have to add some sort of extra circuit to sense the current in the first place. Even at 100MHz and interrupts your OCD response time will probably still be in the hundreds of u second range unless your uC spends all it's time checking the OCD. A comparator is really the better way to go for speed, as you want to get that silicon shut off in <10uS if possible. MBED also runs a rater complex firmware and library architecture that will eat up that 100MHz rather fast making it not much faster than a 8bit 16MHz uC.
In my opinion 100MHz uC is not the right way to go with this sort of controller, it's will be a lot a slower than a single dedicated comparator and a lot more susceptible to noise and such.
If you wanted a configurable option that was fast and had better noise immunity you should look into a FPGA or CPLD. There is a reason uCs are not widely used for SMPS control.
Eric
Re: QCW coil
teravolt, Wed Jan 04 2012, 04:34PM
I don't know if any body is interested but I found this IC that I stumbled on and thought that this may be pertinent
teravolt, Wed Jan 04 2012, 04:34PM
I don't know if any body is interested but I found this IC that I stumbled on and thought that this may be pertinent
Re: QCW coil
Goodchild, Wed Jan 04 2012, 06:26PM
teravolt that is an awesome little chip, it dose bang bang control a little bit differently than normal seems like it samples rather than running all the time. From my initial glance at the datasheet it looks like all timing is done internally rather than using external integration, as I said earlier this is a wonderful way to improve noise and load immunity on the switching times.
I may have to give one of these a try some time in the future.
Goodchild, Wed Jan 04 2012, 06:26PM
teravolt wrote ...
I don't know if any body is interested but I found this IC that I stumbled on and thought that this may be pertinent
I don't know if any body is interested but I found this IC that I stumbled on and thought that this may be pertinent
teravolt that is an awesome little chip, it dose bang bang control a little bit differently than normal seems like it samples rather than running all the time. From my initial glance at the datasheet it looks like all timing is done internally rather than using external integration, as I said earlier this is a wonderful way to improve noise and load immunity on the switching times.
I may have to give one of these a try some time in the future.
Re: QCW coil
Gregory, Wed Jan 04 2012, 06:47PM
So, are you using a single histeresy comparator? only a comparator with the -input at the feedback and the +input at the reference ramp with some histeresy at comparation? This is my original desing that worked very well.
Gregory, Wed Jan 04 2012, 06:47PM
ou know my delta modulator consist of a single comparator and a single 5V supply
So, are you using a single histeresy comparator? only a comparator with the -input at the feedback and the +input at the reference ramp with some histeresy at comparation? This is my original desing that worked very well.
Re: QCW coil
Goodchild, Wed Jan 04 2012, 07:45PM
Basically yep that's it. I however also sum my own integrated ramp back onto the feedback so that the integration is set internally rather than by external LC.
Goodchild, Wed Jan 04 2012, 07:45PM
Gregory wrote ...
So, are you using a single histeresy comparator? only a comparator with the -input at the feedback and the +input at the reference ramp with some histeresy at comparation? This is my original desing that worked very well.
ou know my delta modulator consist of a single comparator and a single 5V supply
So, are you using a single histeresy comparator? only a comparator with the -input at the feedback and the +input at the reference ramp with some histeresy at comparation? This is my original desing that worked very well.
Basically yep that's it. I however also sum my own integrated ramp back onto the feedback so that the integration is set internally rather than by external LC.
Re: QCW coil
Gregory, Wed Jan 04 2012, 08:22PM
First order delta modulator :>
But, for your integrator/acumulator amplifier you need a simetrical -5v to 5v supply, no?
I'm using TL3116 as comparator and TL072 as amplifiers/integrators.
I have already tryed to sum a small derivation term at the feedback, it help to get a better ripple/transient response. It acts iqual to the feed-forword capacitor, bacause this capacitor can be see as a derivator.
Now I need to make a good inductor hehe. I need to buy a inductimetter, mesure the indutance using ressonance sucks.
Gregory, Wed Jan 04 2012, 08:22PM
I however also sum my own integrated ramp back onto the feedback
First order delta modulator :>
But, for your integrator/acumulator amplifier you need a simetrical -5v to 5v supply, no?
I'm using TL3116 as comparator and TL072 as amplifiers/integrators.
I have already tryed to sum a small derivation term at the feedback, it help to get a better ripple/transient response. It acts iqual to the feed-forword capacitor, bacause this capacitor can be see as a derivator.
Now I need to make a good inductor hehe. I need to buy a inductimetter, mesure the indutance using ressonance sucks.
Re: QCW coil
Goodchild, Wed Jan 04 2012, 09:50PM
You don't need a separate opamp or a split +5 -5 supply. I use a single TL3116 and a single +5V supply.
I actually use an RC network to generate the ramp that is summed back on the feedback ramp. The trick is to have the feedback ramp be larger in amplitude that the ripple of the output so that the comparator only looks at the RC ramp and not the ripple from the output.
Goodchild, Wed Jan 04 2012, 09:50PM
Gregory wrote ...
First order delta modulator :>
But, for your integrator/acumulator amplifier you need a simetrical -5v to 5v supply, no?
I'm using TL3116 as comparator and TL072 as amplifiers/integrators.
I have already tryed to sum a small derivation term at the feedback, it help to get a better ripple/transient response. It acts iqual to the feed-forword capacitor, bacause this capacitor can be see as a derivator.
Now I need to make a good inductor hehe. I need to buy a inductimetter, mesure the indutance using ressonance sucks.
I however also sum my own integrated ramp back onto the feedback
First order delta modulator :>
But, for your integrator/acumulator amplifier you need a simetrical -5v to 5v supply, no?
I'm using TL3116 as comparator and TL072 as amplifiers/integrators.
I have already tryed to sum a small derivation term at the feedback, it help to get a better ripple/transient response. It acts iqual to the feed-forword capacitor, bacause this capacitor can be see as a derivator.
Now I need to make a good inductor hehe. I need to buy a inductimetter, mesure the indutance using ressonance sucks.
You don't need a separate opamp or a split +5 -5 supply. I use a single TL3116 and a single +5V supply.
I actually use an RC network to generate the ramp that is summed back on the feedback ramp. The trick is to have the feedback ramp be larger in amplitude that the ripple of the output so that the comparator only looks at the RC ramp and not the ripple from the output.
Re: QCW coil
Gregory, Thu Jan 05 2012, 12:10AM
If you are using only a single comparator I think that he is the quantizer (feedback - reference + histeresy). How are you adding the integrated ramp with not other amp? oO You need to make the integration.
Gregory, Thu Jan 05 2012, 12:10AM
You don't need a separate opamp or a split +5 -5 supply. I use a single TL3116 and a single +5V supply.
If you are using only a single comparator I think that he is the quantizer (feedback - reference + histeresy). How are you adding the integrated ramp with not other amp? oO You need to make the integration.
Re: QCW coil
Dr. ISOTOP, Thu Jan 05 2012, 01:17AM
Sadly that chip seems a tad slow; it can't do over 10 KHz, which is borderline for a QCW modulator and still audible
Dr. ISOTOP, Thu Jan 05 2012, 01:17AM
Goodchild wrote ...
teravolt that is an awesome little chip, it dose bang bang control a little bit differently than normal seems like it samples rather than running all the time. From my initial glance at the datasheet it looks like all timing is done internally rather than using external integration, as I said earlier this is a wonderful way to improve noise and load immunity on the switching times.
I may have to give one of these a try some time in the future.
teravolt wrote ...
I don't know if any body is interested but I found this IC that I stumbled on and thought that this may be pertinent
I don't know if any body is interested but I found this IC that I stumbled on and thought that this may be pertinent
teravolt that is an awesome little chip, it dose bang bang control a little bit differently than normal seems like it samples rather than running all the time. From my initial glance at the datasheet it looks like all timing is done internally rather than using external integration, as I said earlier this is a wonderful way to improve noise and load immunity on the switching times.
I may have to give one of these a try some time in the future.
Sadly that chip seems a tad slow; it can't do over 10 KHz, which is borderline for a QCW modulator and still audible
Gregory wrote ...
If you are using only a single comparator I think that he is the quantizer (feedback - reference + histeresy). How are you adding the integrated ramp with not other amp? oO You need to make the integration.
I believe he is using a passive RC integrator.You don't need a separate opamp or a split +5 -5 supply. I use a single TL3116 and a single +5V supply.
If you are using only a single comparator I think that he is the quantizer (feedback - reference + histeresy). How are you adding the integrated ramp with not other amp? oO You need to make the integration.
Re: QCW coil
Gregory, Thu Jan 05 2012, 02:45AM
ohhhhh. Now I understand.. He is using the RC to integrate/acumulate. I was think that he was using the RC to generate the reference ramp with some PWM from a uC. Now all make sense.
Gregory, Thu Jan 05 2012, 02:45AM
ohhhhh. Now I understand.. He is using the RC to integrate/acumulate. I was think that he was using the RC to generate the reference ramp with some PWM from a uC. Now all make sense.
Re: QCW coil
Goodchild, Thu Jan 05 2012, 05:28AM
you got it.
Goodchild, Thu Jan 05 2012, 05:28AM
Gregory wrote ...
ohhhhh. Now I understand.. He is using the RC to integrate/acumulate. I was think that he was using the RC to generate the reference ramp with some PWM from a uC. Now all make sense.
ohhhhh. Now I understand.. He is using the RC to integrate/acumulate. I was think that he was using the RC to generate the reference ramp with some PWM from a uC. Now all make sense.
you got it.
Re: QCW coil
teravolt, Thu Jan 05 2012, 06:58AM
hi Dr. Isotop, when I looked at the specs I think you are right that 10Khz sample rate is a bit slow. How ever for the bang bang control you probably don't need much more resolution for a ramp wave form of 5 to 20mS to filter. using a mbed for delta modulation is over kill you could probly use a adruino. cirtanly a microcontroler combined with hardware would be a awsum way to controle feedback and reduce parts count. I will probly end up using all hardware at this point for feedback.
teravolt, Thu Jan 05 2012, 06:58AM
hi Dr. Isotop, when I looked at the specs I think you are right that 10Khz sample rate is a bit slow. How ever for the bang bang control you probably don't need much more resolution for a ramp wave form of 5 to 20mS to filter. using a mbed for delta modulation is over kill you could probly use a adruino. cirtanly a microcontroler combined with hardware would be a awsum way to controle feedback and reduce parts count. I will probly end up using all hardware at this point for feedback.
Re: QCW coil
Dr. ISOTOP, Thu Jan 05 2012, 07:17AM
Hmm I suppose a 10 mS ramp corresponds to just 100 Hz, so there's 10 KHz is a reasonable switching rate.
The micro doesn't really save parts, given that the comparator-with-hysteresis approach is a comparator and a few resistors. The Mbed is definitely *not* the way to go even with that approach; I used it because I happened to have one around at the time. If I do keep the software approach (which at this point looks less and less likely) I'd use this
Dr. ISOTOP, Thu Jan 05 2012, 07:17AM
teravolt wrote ...
hi Dr. Isotop, when I looked at the specs I think you are right that 10Khz sample rate is a bit slow. How ever for the bang bang control you probably don't need much more resolution for a ramp wave form of 5 to 20mS to filter. using a mbed for delta modulation is over kill you could probly use a adruino. cirtanly a microcontroler combined with hardware would be a awsum way to controle feedback and reduce parts count. I will probly end up using all hardware at this point for feedback.
hi Dr. Isotop, when I looked at the specs I think you are right that 10Khz sample rate is a bit slow. How ever for the bang bang control you probably don't need much more resolution for a ramp wave form of 5 to 20mS to filter. using a mbed for delta modulation is over kill you could probly use a adruino. cirtanly a microcontroler combined with hardware would be a awsum way to controle feedback and reduce parts count. I will probly end up using all hardware at this point for feedback.
Hmm I suppose a 10 mS ramp corresponds to just 100 Hz, so there's 10 KHz is a reasonable switching rate.
The micro doesn't really save parts, given that the comparator-with-hysteresis approach is a comparator and a few resistors. The Mbed is definitely *not* the way to go even with that approach; I used it because I happened to have one around at the time. If I do keep the software approach (which at this point looks less and less likely) I'd use this
Re: QCW coil
Gregory, Fri Jan 06 2012, 02:03AM
Eric
I tryed to make some simulation in LTspice with a RC integrator and I get only sheet, hehehe.
I tried a simples quantizer/comparator (amp with ramp at +input and feedback at -input) , the output of the comparator driven the switch and at the same output the RC that summing at the feedback (-input).
Gregory, Fri Jan 06 2012, 02:03AM
Eric
I tryed to make some simulation in LTspice with a RC integrator and I get only sheet, hehehe.
I tried a simples quantizer/comparator (amp with ramp at +input and feedback at -input) , the output of the comparator driven the switch and at the same output the RC that summing at the feedback (-input).
Re: QCW coil
teravolt, Fri Jan 06 2012, 04:35AM
hi Dr. Isotop that's a nice package lots of stuff on it. it seems that arm is prity popular to. I like MBED and Arduino because there is a large base of users. when it comes to software I'm prity much a newbie but when it comes to hardware I think of myself as not to bad. micros have a lot of potential.
Gregory, I kind of have the idea but do you have a schematic
teravolt, Fri Jan 06 2012, 04:35AM
hi Dr. Isotop that's a nice package lots of stuff on it. it seems that arm is prity popular to. I like MBED and Arduino because there is a large base of users. when it comes to software I'm prity much a newbie but when it comes to hardware I think of myself as not to bad. micros have a lot of potential.
Gregory, I kind of have the idea but do you have a schematic
Re: QCW coil
Goodchild, Fri Jan 06 2012, 06:07AM
Check for silly errors, it could just be that it's phased wrong.
Goodchild, Fri Jan 06 2012, 06:07AM
Gregory wrote ...
Eric
I tryed to make some simulation in LTspice with a RC integrator and I get only sheet, hehehe.
I tried a simples quantizer/comparator (amp with ramp at +input and feedback at -input) , the output of the comparator driven the switch and at the same output the RC that summing at the feedback (-input).
Eric
I tryed to make some simulation in LTspice with a RC integrator and I get only sheet, hehehe.
I tried a simples quantizer/comparator (amp with ramp at +input and feedback at -input) , the output of the comparator driven the switch and at the same output the RC that summing at the feedback (-input).
Check for silly errors, it could just be that it's phased wrong.
Re: QCW coil
Dr. ISOTOP, Fri Jan 06 2012, 06:52AM
Plus, it gets a tad costly after you kill two or three
Dr. ISOTOP, Fri Jan 06 2012, 06:52AM
Gregory wrote ...
Eric
I tryed to make some simulation in LTspice with a RC integrator and I get only sheet, hehehe.
I tried a simples quantizer/comparator (amp with ramp at +input and feedback at -input) , the output of the comparator driven the switch and at the same output the RC that summing at the feedback (-input).
I *think* (not sure on this) if your comparator wasn't set up with hysteresis, the controller will not oscillate.Eric
I tryed to make some simulation in LTspice with a RC integrator and I get only sheet, hehehe.
I tried a simples quantizer/comparator (amp with ramp at +input and feedback at -input) , the output of the comparator driven the switch and at the same output the RC that summing at the feedback (-input).
teravolt wrote ...
hi Dr. Isotop that's a nice package lots of stuff on it. it seems that arm is prity popular to. I like MBED and Arduino because there is a large base of users. when it comes to software I'm prity much a newbie but when it comes to hardware I think of myself as not to bad. micros have a lot of potential.
Gregory, I kind of have the idea but do you have a schematic
The Mbed is cute and all, but I don't trust their library to be performance-optimized. Overhead on a micro is a no-no, IMHO.hi Dr. Isotop that's a nice package lots of stuff on it. it seems that arm is prity popular to. I like MBED and Arduino because there is a large base of users. when it comes to software I'm prity much a newbie but when it comes to hardware I think of myself as not to bad. micros have a lot of potential.
Gregory, I kind of have the idea but do you have a schematic
Plus, it gets a tad costly after you kill two or three
Re: QCW coil
Gregory, Fri Jan 06 2012, 04:12PM
I tried some simple histerese using a RC delayer at the quantizer output, and a only get more ripple, normal. The histeresys is here only to make the thing not oscilate and make the comparation more stable. I think the problem is the RC constant that I need to calculate
?
Gregory, Fri Jan 06 2012, 04:12PM
I *think* (not sure on this) if your comparator wasn't set up with hysteresis, the controller will not oscillate.,
I tried some simple histerese using a RC delayer at the quantizer output, and a only get more ripple, normal. The histeresys is here only to make the thing not oscilate and make the comparation more stable. I think the problem is the RC constant that I need to calculate
Gregory, I kind of have the idea but do you have a schematic
?
Re: QCW coil
Goodchild, Fri Jan 06 2012, 06:17PM
Sounds like your not feeding enough amplitude from the RC back onto the feedback, the RC ripple must be larger than the ripple from the feedback for the RC to take effect.
Goodchild, Fri Jan 06 2012, 06:17PM
Gregory wrote ...
I tried some simple histerese using a RC delayer at the quantizer output, and a only get more ripple, normal. The histeresys is here only to make the thing not oscilate and make the comparation more stable. I think the problem is the RC constant that I need to calculate
?
I *think* (not sure on this) if your comparator wasn't set up with hysteresis, the controller will not oscillate.,
I tried some simple histerese using a RC delayer at the quantizer output, and a only get more ripple, normal. The histeresys is here only to make the thing not oscilate and make the comparation more stable. I think the problem is the RC constant that I need to calculate
Gregory, I kind of have the idea but do you have a schematic
?
Sounds like your not feeding enough amplitude from the RC back onto the feedback, the RC ripple must be larger than the ripple from the feedback for the RC to take effect.
Re: QCW coil
Dr. ISOTOP, Sat Jan 07 2012, 07:56AM
I'm not sure I understand how the internal integrator works; since you're not taking feedback directly from the bus voltage, won't it fail to regulate sometimes?
Also, I simulated the LTC1041 and it seems to work decently:
For some reason the LTC1041 doesn't like it when the divider resistors are too high (at least the simulation says that).
Dr. ISOTOP, Sat Jan 07 2012, 07:56AM
I'm not sure I understand how the internal integrator works; since you're not taking feedback directly from the bus voltage, won't it fail to regulate sometimes?
Also, I simulated the LTC1041 and it seems to work decently:
For some reason the LTC1041 doesn't like it when the divider resistors are too high (at least the simulation says that).
Re: QCW coil
Goodchild, Sat Jan 07 2012, 03:42PM
Well yes and no; I do sample from the bus voltage, but not the bus voltage of the buck, rather the output of the buck which also happens to be the bus of the inverter.
Your using a bang bang controller there, that is a totally different story than what I was talking about with my internal integration. I use a single comparator and not a full BB controller. It also has the integration set internal using the delta pin.
As a result switching will be determined mainly by the delta set point and the sampling rate. I would suggest reading the datasheet, all of this info and more are already in it.
Goodchild, Sat Jan 07 2012, 03:42PM
Well yes and no; I do sample from the bus voltage, but not the bus voltage of the buck, rather the output of the buck which also happens to be the bus of the inverter.
Your using a bang bang controller there, that is a totally different story than what I was talking about with my internal integration. I use a single comparator and not a full BB controller. It also has the integration set internal using the delta pin.
As a result switching will be determined mainly by the delta set point and the sampling rate. I would suggest reading the datasheet, all of this info and more are already in it.
Re: QCW coil
Gregory, Sat Jan 07 2012, 03:43PM
I think that with a higher resistor values you get less high frequency gain. Try increase the resistor and use a feed-foward capacitor at your divider, it will "derivate" and will increase the gain for the high frequency (ripple). Some 270p capacitor in parallel with the upper resistor of the divider will make the job.
Gregory, Sat Jan 07 2012, 03:43PM
For some reason the LTC1041 doesn't like it when the divider resistors are too high (at least the simulation says that).
I think that with a higher resistor values you get less high frequency gain. Try increase the resistor and use a feed-foward capacitor at your divider, it will "derivate" and will increase the gain for the high frequency (ripple). Some 270p capacitor in parallel with the upper resistor of the divider will make the job.
Re: QCW coil
Dr. ISOTOP, Sun Jan 08 2012, 01:35AM
Though not in the way I expected it too.
Turns out I missed the bit on the datasheet about charging the input capacitor in the LTC1041
Mmm I'm liking that waveform...hopefully it works in real life!
Dr. ISOTOP, Sun Jan 08 2012, 01:35AM
Goodchild wrote ...
Well yes and no; I do sample from the bus voltage, but not the bus voltage of the buck, rather the output of the buck which also happens to be the bus of the inverter.
Oops, that's what I meant by "bus voltage".Well yes and no; I do sample from the bus voltage, but not the bus voltage of the buck, rather the output of the buck which also happens to be the bus of the inverter.
Gregory wrote ...
I think that with a higher resistor values you get less high frequency gain. Try increase the resistor and use a feed-foward capacitor at your divider, it will "derivate" and will increase the gain for the high frequency (ripple). Some 270p capacitor in parallel with the upper resistor of the divider will make the job.
And what do you know, the feed-forward capacitor works magic For some reason the LTC1041 doesn't like it when the divider resistors are too high (at least the simulation says that).
I think that with a higher resistor values you get less high frequency gain. Try increase the resistor and use a feed-foward capacitor at your divider, it will "derivate" and will increase the gain for the high frequency (ripple). Some 270p capacitor in parallel with the upper resistor of the divider will make the job.
Though not in the way I expected it too.Turns out I missed the bit on the datasheet about charging the input capacitor in the LTC1041
Mmm I'm liking that waveform...hopefully it works in real life!
Re: QCW coil
Gregory, Sun Jan 08 2012, 02:15AM
I think that you can increase even more the resistances of the divider, I think 100k -> 1k is a better value.
Gregory, Sun Jan 08 2012, 02:15AM
I think that you can increase even more the resistances of the divider, I think 100k -> 1k is a better value.
Re: QCW coil
Dr. ISOTOP, Sun Jan 08 2012, 03:39AM
10K@300V is 9W dissipated in the divider, which is a bit hot but is only a tiny fraction of the output power.
Dr. ISOTOP, Sun Jan 08 2012, 03:39AM
Gregory wrote ...
I think that you can increase even more the resistances of the divider, I think 100k -> 1k is a better value.
The LTC1041 datasheet says 10K is the highest input resistance at which the input capacitor can fully charge. It still works at 600K:10K, but the voltage at the feedback pin starts misbehaving. The datasheet says this can be worked around with a bypass cap, but I'm afraid the integrating effect of the capacitor will give me poor transient response.I think that you can increase even more the resistances of the divider, I think 100k -> 1k is a better value.
10K@300V is 9W dissipated in the divider, which is a bit hot but is only a tiny fraction of the output power.
Re: QCW coil
Goodchild, Mon Jan 09 2012, 02:08AM
Nope it's actually quite the opposite, if selected correctly the capacitor will actually improve your transient response.
Goodchild, Mon Jan 09 2012, 02:08AM
bwang wrote ...
The datasheet says this can be worked around with a bypass cap, but I'm afraid the integrating effect of the capacitor will give me poor transient response.
The datasheet says this can be worked around with a bypass cap, but I'm afraid the integrating effect of the capacitor will give me poor transient response.
Nope it's actually quite the opposite, if selected correctly the capacitor will actually improve your transient response.
Re: QCW coil
Gregory, Mon Jan 09 2012, 04:04AM
You can use some opamp at voltage-folower mode to decrease your divider ouput impedance.
Yeah! If you choose the correct value it will "acumulate" the error and use it to give you a fast response.
Gregory, Mon Jan 09 2012, 04:04AM
The LTC1041 datasheet says 10K is the highest input resistance at which the input capacitor can fully charge.
You can use some opamp at voltage-folower mode to decrease your divider ouput impedance.
Nope it's actually quite the opposite, if selected correctly the capacitor will actually improve your transient response.
Yeah! If you choose the correct value it will "acumulate" the error and use it to give you a fast response.
Re: QCW coil
Steve Conner, Mon Jan 09 2012, 11:27AM
Just be careful that the feedforward capacitor doesn't "feed forward" big enough voltage spikes to blow the controller chip. You won't see these spikes in your simulation because they come from filter capacitor ESR/ESL and wiring inductance. It's a wise move to have some series resistance.
Steve Conner, Mon Jan 09 2012, 11:27AM
Just be careful that the feedforward capacitor doesn't "feed forward" big enough voltage spikes to blow the controller chip. You won't see these spikes in your simulation because they come from filter capacitor ESR/ESL and wiring inductance. It's a wise move to have some series resistance.
Re: QCW coil
Gregory, Mon Jan 09 2012, 04:20PM
Something between 470ohm and 1k will work just fine
Gregory, Mon Jan 09 2012, 04:20PM
Just be careful that the feedforward capacitor doesn't "feed forward" big enough voltage spikes to blow the controller chip. You won't see these spikes in your simulation because they come from filter capacitor ESR/ESL and wiring inductance. It's a wise move to have some series resistance.
Something between 470ohm and 1k will work just fine
Re: QCW coil
Goodchild, Mon Jan 09 2012, 06:37PM
something I just noticed also you should use antiparallel diodes with your IGBTs in real life as well, most IGBT half bridge modules have them built in already so it probably won't be a big deal. I also added in the diode in my simulations as well, because you do get some switching action with the diodes even though it is a sync buck. The low side diode turns on at the same time as the low side switch.
On the note of a series resistor with the feedforword cap, also note that it will change the switching windows as well, so change the cap or resistor value accordingly.
Goodchild, Mon Jan 09 2012, 06:37PM
bwang wrote ...
Though not in the way I expected it too.
Turns out I missed the bit on the datasheet about charging the input capacitor in the LTC1041
Mmm I'm liking that waveform...hopefully it works in real life!
Goodchild wrote ...
Well yes and no; I do sample from the bus voltage, but not the bus voltage of the buck, rather the output of the buck which also happens to be the bus of the inverter.
Oops, that's what I meant by "bus voltage".Well yes and no; I do sample from the bus voltage, but not the bus voltage of the buck, rather the output of the buck which also happens to be the bus of the inverter.
Gregory wrote ...
I think that with a higher resistor values you get less high frequency gain. Try increase the resistor and use a feed-foward capacitor at your divider, it will "derivate" and will increase the gain for the high frequency (ripple). Some 270p capacitor in parallel with the upper resistor of the divider will make the job.
And what do you know, the feed-forward capacitor works magic For some reason the LTC1041 doesn't like it when the divider resistors are too high (at least the simulation says that).
I think that with a higher resistor values you get less high frequency gain. Try increase the resistor and use a feed-foward capacitor at your divider, it will "derivate" and will increase the gain for the high frequency (ripple). Some 270p capacitor in parallel with the upper resistor of the divider will make the job.
Though not in the way I expected it too.Turns out I missed the bit on the datasheet about charging the input capacitor in the LTC1041
Mmm I'm liking that waveform...hopefully it works in real life!
something I just noticed also you should use antiparallel diodes with your IGBTs in real life as well, most IGBT half bridge modules have them built in already so it probably won't be a big deal. I also added in the diode in my simulations as well, because you do get some switching action with the diodes even though it is a sync buck. The low side diode turns on at the same time as the low side switch.
On the note of a series resistor with the feedforword cap, also note that it will change the switching windows as well, so change the cap or resistor value accordingly.
Print this page