kartino said:
Hi,
I think at this point in time we no longer need to mention that spice is not realistic enough ... etc. 🙂 That comparator being at the transistor level though will be fairly accurate.
It's not the high side driver that's much of a concern in itself. It's the comparator output for the high side driver, operating at ~500mW as opposed to ~25mW for the low side output. Now let's say it does fail.. I think it will take alot with it. Current hogging is also not protrayed by spice, but if there's any way it can be hogged, that sort of difference is going to cause it, no doubt in the least it will cause an odd timming/distortion error.
Even though it's "self correcting" the less it has to contend with the better.
Once again I'd also like to point out the PSRR of the lower rail... zero?
As per the drivers in general, they're just too weak to handle the parasitics associated with higher voltage very well. They're too weak to drive parallel high voltage mosfets too, which would present even more parasitic/miller effect. Go ahead and simulate that at 20kHz input @~5V and look how much your gates are modulated!! Those will be some hot mosfets, with a nice DC offset because of unequal Ron. While you're at it look at the lack of control over deadtime. It's all over the place, out of control.
BTW, why is it that one gate-source waveform has a good diode drop more than the other?
I think 12V zeners still stand a chance at conducting in this circuit. I'd make them between 18 and 24. That'll let you experiment with 15V gate drive if you like and still protect the driver easily. However I think if the mosfets blow, they're going to see full rail voltage with nothing to limite the current, those better be some high power zeners. I'd just remove them myself.
You have to admit the circuit I gave you guys long ago is nicely refined, though maybe not "perfect", but probably not very far off. It's a long road reinventing the wheel, and you're sure to learn alot, but don't go saying it's a 100V circuit until you've at least built it... know what I mean? Some kid is going to say "wow cool" and it'll blow up in his face.
Even the one I gave you is good for no where near that.
Regards,
Chris
I've added shortcircuit protection to the schematic. It works both for a short of the load (even to the rails) and for a temporary mosfet short which could happen at startup etc.. Have a look at it and tell me what you think.
It's the same short circuit protection that I'm using on my own amp. It's pritty simple and very fast. I was inspired by the bistabil multivibrators of the old days.
Attached is a screendump of a 3xshortcircuit simulation and the LTspice .asc file. Have fun!
Regards Kaspar S. Meyer
It's the same short circuit protection that I'm using on my own amp. It's pritty simple and very fast. I was inspired by the bistabil multivibrators of the old days.
Attached is a screendump of a 3xshortcircuit simulation and the LTspice .asc file. Have fun!
Regards Kaspar S. Meyer
Hi Chris,
Actually last night I recheck the posibility to use this proposed amp. The result is the high gate driver looked OK.
The only one that might have problem is VCE for Q6, is almost (100+100+12) = ~212V, wow....that's very high!
Using cascode to share IMHO would not help since the big portion still at the BJT that connected to high side gate driver (upper one).
Any idea?
The both gate driver voltage also shown not so equal. I don't know why, I will check again.
For PSRR problem, ok I agree with you, I have tried to use 1.3mA current source placed at R27 and R29, the result is the output very stable at any voltage level. Too complicated is it, but not so costly.
For Kaspar, Wow... thanks to sovadk for the cool short circuit protection....!
Lastly, we have 2 problem now, parasitic and Q6 VCE. With this problem, is it still have possible to make this amp or not? I am so pessimistic now. But I think for 70-80V rails will fine.
Thanks and best regards,
kartino
Actually last night I recheck the posibility to use this proposed amp. The result is the high gate driver looked OK.
The only one that might have problem is VCE for Q6, is almost (100+100+12) = ~212V, wow....that's very high!
Using cascode to share IMHO would not help since the big portion still at the BJT that connected to high side gate driver (upper one).
Any idea?
The both gate driver voltage also shown not so equal. I don't know why, I will check again.
For PSRR problem, ok I agree with you, I have tried to use 1.3mA current source placed at R27 and R29, the result is the output very stable at any voltage level. Too complicated is it, but not so costly.
For Kaspar, Wow... thanks to sovadk for the cool short circuit protection....!
Lastly, we have 2 problem now, parasitic and Q6 VCE. With this problem, is it still have possible to make this amp or not? I am so pessimistic now. But I think for 70-80V rails will fine.
Thanks and best regards,
kartino
To the Q6 problem:
I use the BSS131 250V in my amp. It's rated at 250V and has a very low miller capacitance -> fast switching. It's a MOSFET, although I don't see that changing anything.
I first tryed a BJT in my design, but it was way to slow when turning off (I was driving into saturation though). Anyway the same breakdown voltage problem applies to the high side current sense BJT, if you choose to implement it.
I have a spice model for BSS131 (attached) and there probably is a model out there for BF420 also.
About the shortcircuit protection:
It works, it's simple and I'm glad you like it 🙂
The downside is that it can be triggered by supply ripple. Therefore it should be tested on a non bidirectional supply like what you get with a rectifier bridge. This could simulate the actual stress IRL.
One good thing about the protection circuit, is that it might be possible to make it even simpler. If the amp still osccilates when the load is shorted, only low side sensing is needed. A post filter Mueta design does this, I dunno about UCD.
A feature not yet present is undervoltage shutdown. This would avoid startup pops etc. It can be added fairly simply if the interest is there??
I use the BSS131 250V in my amp. It's rated at 250V and has a very low miller capacitance -> fast switching. It's a MOSFET, although I don't see that changing anything.
I first tryed a BJT in my design, but it was way to slow when turning off (I was driving into saturation though). Anyway the same breakdown voltage problem applies to the high side current sense BJT, if you choose to implement it.
I have a spice model for BSS131 (attached) and there probably is a model out there for BF420 also.
About the shortcircuit protection:
It works, it's simple and I'm glad you like it 🙂
The downside is that it can be triggered by supply ripple. Therefore it should be tested on a non bidirectional supply like what you get with a rectifier bridge. This could simulate the actual stress IRL.
One good thing about the protection circuit, is that it might be possible to make it even simpler. If the amp still osccilates when the load is shorted, only low side sensing is needed. A post filter Mueta design does this, I dunno about UCD.
A feature not yet present is undervoltage shutdown. This would avoid startup pops etc. It can be added fairly simply if the interest is there??
BFN26 is not discontinued by Infineon. It's rated at 300V 200mA Ccb=1.5pF and it's housed in a SO23-3 package (I like SMD). This is the model (actually this one is made by Siemens). Just add it to "SwCADIII\lib\cmp\standard.bjt"
Attached is a circuit where it's included. I'm made a cascode cupling to enusre the same temperature of Q6 and Q7.
Attached is a circuit where it's included. I'm made a cascode cupling to enusre the same temperature of Q6 and Q7.
Hi,
Well, this simplified version is becomming more and more complex. I don't think it will be long before you go back to my version of it 🙂 Looks like its' slowly getting there already. It's quite simple, far more robust, precise, and balanced. Bit more like the original I'm assuming. None of this gets you out of the patent, and being a worse circuit I fail to see the motivation for it?
As far as the gate voltages not being balanced, I already pointed out the cure for that at least three times, so this time I just present you with the problem, and let you investigate to come up with the cure yourself 🙂 I think at 100V you'd probably want them well balanced. You also don't want them modulating down to ~9Volts at full power. These drivers become insufficient at these power levels, and that, is the area you should really be working on improving.
Regards,
Chris
Well, this simplified version is becomming more and more complex. I don't think it will be long before you go back to my version of it 🙂 Looks like its' slowly getting there already. It's quite simple, far more robust, precise, and balanced. Bit more like the original I'm assuming. None of this gets you out of the patent, and being a worse circuit I fail to see the motivation for it?
As far as the gate voltages not being balanced, I already pointed out the cure for that at least three times, so this time I just present you with the problem, and let you investigate to come up with the cure yourself 🙂 I think at 100V you'd probably want them well balanced. You also don't want them modulating down to ~9Volts at full power. These drivers become insufficient at these power levels, and that, is the area you should really be working on improving.
Regards,
Chris
sovadk said:
With pleasure.
http://www.diyaudio.com/forums/showthread.php?threadid=55385
and another,
http://www.diyaudio.com/forums/showthread.php?postid=786795#post786795
I have some posted all over, even a full bridge or two. Non are entirely complete but the hard stuff is done.
Regards,
Chris
I'm sorry Chris. I don't see any reson for DIY's to switch version. The only real difference I can spot is that you're using a slightly more complicated comperator with some cascode coupling.
The propogation delay in "UCD_High Rails" is large. In the configuration I'm using (see attachment) I've done everything to minimize propogation delay and even sacrificed a lot of gain to achieve this. This is because I'm using hysteresis modulation, which approaches a UCD modulation scheme when the delay becomes large.
Anyway this was just something I've been thinking about, it has not so much to do with my comment on your comperator configuration.
By that way, I actually come to like the gate drivers you guys use instead of a totempole configuration. The reason is, that high side level shifter isn't driven into saturation which really slows things down in my design.
Regards Kaspar
The propogation delay in "UCD_High Rails" is large. In the configuration I'm using (see attachment) I've done everything to minimize propogation delay and even sacrificed a lot of gain to achieve this. This is because I'm using hysteresis modulation, which approaches a UCD modulation scheme when the delay becomes large.
Anyway this was just something I've been thinking about, it has not so much to do with my comment on your comperator configuration.
By that way, I actually come to like the gate drivers you guys use instead of a totempole configuration. The reason is, that high side level shifter isn't driven into saturation which really slows things down in my design.
Regards Kaspar
Hi. I'm sorry that I'm floating this thread with posts.
I couldn't help it and got inspired by UCD's gate driver to do this: attachment (it's just a compressed png file)
The deadtime is below 50ns. The total power dissapation in X8 and X10 (the level shifters) is below 200mW for +/-30V rails.
If you look at the power MOSFET currents I(X9:d) and I(X7:s), there's no sign of shoot through.
Regards Kaspar
I couldn't help it and got inspired by UCD's gate driver to do this: attachment (it's just a compressed png file)
The deadtime is below 50ns. The total power dissapation in X8 and X10 (the level shifters) is below 200mW for +/-30V rails.
If you look at the power MOSFET currents I(X9:d) and I(X7:s), there's no sign of shoot through.
Regards Kaspar
OK, this is my solution for share stress for Q6.
And PSRR problem. And balanced problem. So this circuit is the only posibility for me. I use cheap component because the only available in here.
Of course we still have parasitic problem.
Like Chris said, it is more and more complicated.
Lastly I would like to ask you everybody, what voltage level that can be implemented with wise for this circuit? I have lack of expereience in higher voltage switching hence I can not make decision from my experience.
Thanks for your sopport,
Best regards,
kartino
And PSRR problem. And balanced problem. So this circuit is the only posibility for me. I use cheap component because the only available in here.
Of course we still have parasitic problem.
Like Chris said, it is more and more complicated.
Lastly I would like to ask you everybody, what voltage level that can be implemented with wise for this circuit? I have lack of expereience in higher voltage switching hence I can not make decision from my experience.
Thanks for your sopport,
Best regards,
kartino
Hi, Chris, Constantine,
Good news, I succesfully fix the minor thing for my 'UCD'. The problem solved when R1 is replaced with series 150ohm and trim pot 100ohm.
The second, the heat already solved. I was using small heatshink like Constantine. The problem is the the heat rise when played louder. Then I have experiment to use small fan to blow the amp. Then after that known that the heatshik pretty cool an no temp rise. Even at played very loud. Yup... the amp need good air circulation.
I got the idea from PC PSU. FYI, I use cheap and dirty rewinded PC PSU to supply the AMP at 30V+/-.
Now the amp became my favorite amp. I find new 'color' of sound at my amp.
Best Regards,
Kartino
Good news, I succesfully fix the minor thing for my 'UCD'. The problem solved when R1 is replaced with series 150ohm and trim pot 100ohm.
The second, the heat already solved. I was using small heatshink like Constantine. The problem is the the heat rise when played louder. Then I have experiment to use small fan to blow the amp. Then after that known that the heatshik pretty cool an no temp rise. Even at played very loud. Yup... the amp need good air circulation.
I got the idea from PC PSU. FYI, I use cheap and dirty rewinded PC PSU to supply the AMP at 30V+/-.
Now the amp became my favorite amp. I find new 'color' of sound at my amp.
Best Regards,
Kartino
classd4sure said:
Hey Chris,
Just comming up to speed on this very interesting thread.
Ever bothered to simulate with very high inductor ripple currents.
In real life, it drops the distortion down.
When the load current is less than the peak inductor current ripple the dead time doens't matter much.
The switching occurs when the switch turns off rather than being dependent on when the switch turns on.
I'd love to try this on a big UcD, and say set the ripple current to say 10 amps. then you could get 5^2 * Rload/2 or 50 Watts of very low distortion output.
My guess is the UcD400 is running at about 1.5~2A ripple current, so you only get about 1A*Rload/2 or 2Watts in this mode.
Notice the UcD distortion curves are very good at 1W!
The smaller inductance, say 5uH, will mean you need to increase the output filter cap to keep all the same though. Like 5.6uF
Not really an issue in UcD. This will raise the Q of the LC with nominal loads, but UcD is pretty invariant to this.
That, and the high Q means its out of band too.
Best Regards,
Mike
Best Regards,
Mike
Kartino, does R1 really trim the dead time?
Btw, here is idea of VERY cheap PSU. You need 2 (two!) power transfomer from ATX PSU (1 for negative and 1 for positive PS rail), one IR2153(D) and two IRF740/BUZ90/7N60/etc. Just it. Primary windings of power transformers connects in parallel, second windings can be commutated in MANY ways ("5"+"12", "12"+"12", etc.) for different amplifier`s power. IR2153 does not provide feedback for voltage regulation, so there is always ~50% duty cycle. Because of it, one "12V" winding would give 23..25V rectified output (does anyone understand what i am talking about?) 🙄
If you would satisfy of +/-25V or +/-11V rails, then you need just one power transfomer from ATX PSU. All of this can be done without rewinding transformers.
PS set IR2153`s clock to 60-80kHz, not 35kHz like typical ATX PSU (50% duty cycle, do you remember?..)
Btw, here is idea of VERY cheap PSU. You need 2 (two!) power transfomer from ATX PSU (1 for negative and 1 for positive PS rail), one IR2153(D) and two IRF740/BUZ90/7N60/etc. Just it. Primary windings of power transformers connects in parallel, second windings can be commutated in MANY ways ("5"+"12", "12"+"12", etc.) for different amplifier`s power. IR2153 does not provide feedback for voltage regulation, so there is always ~50% duty cycle. Because of it, one "12V" winding would give 23..25V rectified output (does anyone understand what i am talking about?) 🙄
If you would satisfy of +/-25V or +/-11V rails, then you need just one power transfomer from ATX PSU. All of this can be done without rewinding transformers.
PS set IR2153`s clock to 60-80kHz, not 35kHz like typical ATX PSU (50% duty cycle, do you remember?..)
- Status
- Not open for further replies.