subwo1 said:
The comparison sounds interesting.
See what I can do
It might be asking a bit much for the first day with a new program..If I can manage it though it will say alot for it.Regarding being impatient...I already started skipping sections of the help files ...
I'd hate to have it cost me two coffees
Chris
By adding an LTF of the following form into the forward path of the circuit :
H = A0 * (jwT1 + 1)/(jwT2 + 1)
The netlist I sent to Bricolo already contains it
, hope he didn't get too confused. The additional circuit has a gain of 5 up to 20 kHz approx and then falls to 1 at about 100 kHz. Even though it is basically flat above that it has enough phase-shift at the switchibg frequency that the feedback cap has to be redimensioned if one doesn't want the switching frequency to drop too much.I wasn't able to find Johan's double loop topology but I assume he does something similar.
For better understanding I post the schematic here (PSPICE users can mail me for the .SCH file).
Regards
Charles
Attachments
Hi,
For your reference:
http://www.diyaudio.com/forums/attachment.php?s=&postid=427287
He's not sure if that can work, neither am I. That actually is somewhat similar, isn't it?
Regards,
Chris
For your reference:
http://www.diyaudio.com/forums/attachment.php?s=&postid=427287
He's not sure if that can work, neither am I. That actually is somewhat similar, isn't it?
Regards,
Chris
Aaaaaah this one !!!!
Yes, definitely !!!!!!
You just have to use a resistor in series with C2 and another resistor parallel to said series RC circuit.
This would give an elegant all-IC solution (apart from the output stage of course) when a gate-driver IC is used. I would use a dual Op-AMP so there is one left to be used as a differential input buffer.
BTW the very same circuit can be used for carrier-based class-d amps with feedback takeoff from the filter, if dimensioned differently (hint: PID controller) !
Regards
Charles
Yes, definitely !!!!!!
You just have to use a resistor in series with C2 and another resistor parallel to said series RC circuit.
This would give an elegant all-IC solution (apart from the output stage of course) when a gate-driver IC is used. I would use a dual Op-AMP so there is one left to be used as a differential input buffer.
BTW the very same circuit can be used for carrier-based class-d amps with feedback takeoff from the filter, if dimensioned differently (hint: PID controller) !
Regards
Charles
Congrats to Bruno P
Playing around a little with the simulator and using ideal switching stages (i.e. voltage controlled voltage source) I had difficulties to get an UcD running.
With nonideal components however this is quite easy.
It is never a bad idea to develop a circuit that makes proactive use of the inevitable non-idealities of real-world components.
Since the intrinsic circuit delays help you building up the oscillation in an UcD you don't have to fight for extremely small delays like in a carrier-based class-d amp.
Regards
Charles
Playing around a little with the simulator and using ideal switching stages (i.e. voltage controlled voltage source) I had difficulties to get an UcD running.
With nonideal components however this is quite easy.
It is never a bad idea to develop a circuit that makes proactive use of the inevitable non-idealities of real-world components.
Since the intrinsic circuit delays help you building up the oscillation in an UcD you don't have to fight for extremely small delays like in a carrier-based class-d amp.
Regards
Charles
Hey!
I was just fighting with that myself, if I understand your french (It's 5am), we came to the same conclusion, I wasted all night.
Also not exactly profficient on Ltspice yet "singular matrix" errors....that's new.
Here's a few thoughts I had maybe you can confirm them unless you already have.
-If we use a comparator IC instead of the discrete version, we should use one with some delay to use to our advantage, (50ns) or have to add some hysteresis and that's ugly.
These little 10nS comparators dont' seem to do the job well.
Note the one Bruno said he used, 45ns...it's that ideal component argument.
-The other is perhaps an argument for going with a non inverting type (mixing the feedback with the input signal before the comparator input). These comparators dont' seem to have a very high common mode voltage range? I can't even find it in the datasheet for the Lt1116 I tried......and it is 5am so I could be all wrong
@Subwo1
I'm not in love with it yet, but I dont' hate it either
It's got some real nice features that's for sure.
Time for a break, went braindead hours ago.
Regards,
Chris
I was just fighting with that myself, if I understand your french
(It's 5am), we came to the same conclusion, I wasted all night.Also not exactly profficient on Ltspice yet "singular matrix" errors....that's new.
Here's a few thoughts I had maybe you can confirm them unless you already have.
-If we use a comparator IC instead of the discrete version, we should use one with some delay to use to our advantage, (50ns) or have to add some hysteresis and that's ugly.
These little 10nS comparators dont' seem to do the job well.
Note the one Bruno said he used, 45ns...it's that ideal component argument.
-The other is perhaps an argument for going with a non inverting type (mixing the feedback with the input signal before the comparator input). These comparators dont' seem to have a very high common mode voltage range? I can't even find it in the datasheet for the Lt1116 I tried......and it is 5am so I could be all wrong
@Subwo1
I'm not in love with it yet, but I dont' hate it either
It's got some real nice features that's for sure.
Time for a break, went braindead hours ago.
Regards,
Chris
classd4sure said:
[snip]These little 10nS comparators dont' seem to do the job well.
Note the one Bruno said he used, 45ns...it's that ideal component argument.
-The other is perhaps an argument for going with a non inverting type (mixing the feedback with the input signal before the comparator input). These comparators dont' seem to have a very high common mode voltage range? I can't even find it in the datasheet for the Lt1116 I tried......and it is 5am so I could be all wrong
@Subwo1
I'm not in love with it yet, but I dont' hate it either
It's got some real nice features that's for sure.
Time for a break, went braindead hours ago.
Regards,
Chris
phase_accurate said:
Hmm, I have to give the idea of some comparators seeming to be too fast additional thought. Definitely, some can be too slow. Bruno said something like there being a possibility of including an additional RC time constant somewhere before the mosfet drive.
Chris,
If it did not have at least somewhat of an intuitive user interface, it would not have a chance with me.
requirements followup
Let me sum up what I've gathered so far from the occasional requirements debate.
* All-in-one Class-D chips are not interesting.
* Single-ended output with dual supplies should be OK. Although H-bridges has distinct advantages, they come at a cost people seem unwilling to take for the "starting point" circuit.
* As for power levels, limiting it is a good idea for beginners, although some expressed that there's no point with class D if power ability is below some level. Break-point seems to lie around 100W (?).
* Double-sided PCB is not a requirement per se, but a must for performance! Try to keep upper-side for ground plane mostly.
* Try to use only through-hole components (?). SMD has size/space advantages, and most likely performance ones too, but save that for the "ultimate" circuit?
* Separate input+control loop from output stage, for easy replacement/refinement of each.
Did I forget something important? Or misrepresent something/someone?
To comment on the power requirement debate, I believe that we can come up with something that does handle up to 100-200W, although it *will* require much more care to wire things up and get it going. (EMC 'niceness' may be lost somewhere on the way up there
That doesn't preclude a beginner to start with +/- 15, or +/-25 volts and get the circuit going without a hitch. So it's more of a documentation/positioning thing...
Regards / Johan
Let me sum up what I've gathered so far from the occasional requirements debate.
* All-in-one Class-D chips are not interesting.
* Single-ended output with dual supplies should be OK. Although H-bridges has distinct advantages, they come at a cost people seem unwilling to take for the "starting point" circuit.
* As for power levels, limiting it is a good idea for beginners, although some expressed that there's no point with class D if power ability is below some level. Break-point seems to lie around 100W (?).
* Double-sided PCB is not a requirement per se, but a must for performance! Try to keep upper-side for ground plane mostly.
* Try to use only through-hole components (?). SMD has size/space advantages, and most likely performance ones too, but save that for the "ultimate" circuit
?* Separate input+control loop from output stage, for easy replacement/refinement of each.
Did I forget something important? Or misrepresent something/someone?
To comment on the power requirement debate, I believe that we can come up with something that does handle up to 100-200W, although it *will* require much more care to wire things up and get it going. (EMC 'niceness' may be lost somewhere on the way up there
That doesn't preclude a beginner to start with +/- 15, or +/-25 volts and get the circuit going without a hitch. So it's more of a documentation/positioning thing...
Regards / Johan
I guess there's a time and place for everything - could someone provide some insight about when to use low-ESR caps, and when to avoid them?phase_accurate said:
(Or on second thought - don't; there's a another forum for that
.Perhaps someone has seen a good thread about it, or wants to start one?)
Re: requirements followup
Thanks for this review. I think another advantage of separating the power stage from the or input or signal processing stage is the ability to provide physical separation for reducing interference. The signal may be able to reach the output stage via a shielded cable, for example. Each may benefit from a faraday cage which may also be ferrous to shunt magnetic fields as well.johanps said:Let me sum up what I've gathered so far from the occasional requirements debate.
* All-in-one Class-D chips are not interesting.
* Single-ended output with dual supplies should be OK. Although H-bridges has distinct advantages, they come at a cost people seem unwilling to take for the "starting point" circuit.
* As for power levels, limiting it is a good idea for beginners, although some expressed that there's no point with class D if power ability is below some level. Break-point seems to lie around 100W (?).
* Double-sided PCB is not a requirement per se, but a must for performance! Try to keep upper-side for ground plane mostly.
* Try to use only through-hole components (?). SMD has size/space advantages, and most likely performance ones too, but save that for the "ultimate" circuit
* Separate input+control loop from output stage, for easy replacement/refinement of each.
Did I forget something important? Or misrepresent something/someone?
To comment on the power requirement debate, I believe that we can come up with something that does handle up to 100-200W, although it *will* require much more care to wire things up and get it going. (EMC 'niceness' may be lost somewhere on the way up there
That doesn't preclude a beginner to start with +/- 15, or +/-25 volts and get the circuit going without a hitch. So it's more of a documentation/positioning thing...
Regards / Johan
Hi,
Intuitive is almost an understatement, keep using LTspice for any length of time and it will teach you spice code if you like it or not! I really liked how that was incorporated with it, very impressive.
I wonder what kind of a delay the length of wires connecting the two circuits together would introduce? Coax cable would probably make an excellent connection, I think it will have to be kept as close as possible to the control circuit though, trivial detail anyway, we'll find out won't we.
Say we go with a two layer board, all through hole. We would have our ground plane but those connections won't be very strong if it's a home made board without plated vias..reduces the odds of the traces surviving component changes as well.
Could be avoided by going all SMD, and could still make our own dual layer boards with ground plane?
Anyway, yeah I'm in total agreement with those requirements how about everyone else? Personally 100W isnt' worth building for me, but if it's 100W into 8 and I can get near 200 into 4 plus the ability to go full bridge well... I might have no choice but to reconsider.
Regards,
Chris
Intuitive is almost an understatement, keep using LTspice for any length of time and it will teach you spice code if you like it or not! I really liked how that was incorporated with it, very impressive.
I wonder what kind of a delay the length of wires connecting the two circuits together would introduce?
Coax cable would probably make an excellent connection, I think it will have to be kept as close as possible to the control circuit though, trivial detail anyway, we'll find out won't we.Say we go with a two layer board, all through hole. We would have our ground plane but those connections won't be very strong if it's a home made board without plated vias..reduces the odds of the traces surviving component changes as well.
Could be avoided by going all SMD, and could still make our own dual layer boards with ground plane?
Anyway, yeah I'm in total agreement with those requirements how about everyone else? Personally 100W isnt' worth building for me, but if it's 100W into 8 and I can get near 200 into 4 plus the ability to go full bridge well... I might have no choice but to reconsider.
Regards,
Chris
.
It helps in that regard when spice directives are shown directly on the schematic page.
I wonder how many folks can solder SMDs though. Personally, I tolerate soldering ones as small as SO-8.
Also, I forgot to mention that the feedback signal would need to be connected with something like a short length of coax. The added delay from such cables would be very small, I think. I believe the cable type makes a difference, but I could be wrong there.
Another thing I like about separate stages is that fewer variables need to be tampered with at a time. Bruno alluded to a similar notion. I would like an output stage with 500w peak capability. I could substitute it for a lower power one and look for overall performance changes.
It helps in that regard when spice directives are shown directly on the schematic page.
I wonder how many folks can solder SMDs though. Personally, I tolerate soldering ones as small as SO-8.
Also, I forgot to mention that the feedback signal would need to be connected with something like a short length of coax. The added delay from such cables would be very small, I think. I believe the cable type makes a difference, but I could be wrong there.
Another thing I like about separate stages is that fewer variables need to be tampered with at a time. Bruno alluded to a similar notion. I would like an output stage with 500w peak capability. I could substitute it for a lower power one and look for overall performance changes.
Hi,
When editing a model, mosfet, you see the table of values and the .model statements for it as you fill them in, same with the sources and everything...I was very impressed with that. It's also spelled out nicely in the help files for every dot command and everything. Something no other program I've seen features. I honestly think others try to bury the spice aspect of their capture programs as a microsoft windows like attempt to keep you dunce about it, and dependant on them. Sorry that's another forum lol.
Hah, I have no idea to be honest I've never had the pleasure to work with them. I understand with a good set of tweezers and perhaps some good magnification if your eyes aren't the best is all you need ? Steady hand goes without saying, but a comfortable work position goes a long way there. Build a soldering station with arm rests
Yep, so did I. I'm sure coax has to be the way to go for those connections, it has the shielding already, and bandwidth. I wonder why skin effect doesn't appear to be a concern with coax as it's normally just a solide strand.
500W rms is about my goal, I really need 400RMS. What you mentioned is an excellent approach and hopefully many will use this for exactly that! I can't see that task being an impossible one if we use a discrete solution as a driver stage and I do believe we should for that reason alone if nothing else.
Regards,
Chris
When editing a model, mosfet, you see the table of values and the .model statements for it as you fill them in, same with the sources and everything...I was very impressed with that. It's also spelled out nicely in the help files for every dot command and everything. Something no other program I've seen features. I honestly think others try to bury the spice aspect of their capture programs as a microsoft windows like attempt to keep you dunce about it, and dependant on them. Sorry that's another forum lol.
Hah, I have no idea to be honest I've never had the pleasure to work with them. I understand with a good set of tweezers and perhaps some good magnification if your eyes aren't the best is all you need ? Steady hand goes without saying, but a comfortable work position goes a long way there. Build a soldering station with arm rests
Yep, so did I. I'm sure coax has to be the way to go for those connections, it has the shielding already, and bandwidth. I wonder why skin effect doesn't appear to be a concern with coax as it's normally just a solide strand.
500W rms is about my goal, I really need 400RMS. What you mentioned is an excellent approach and hopefully many will use this for exactly that! I can't see that task being an impossible one if we use a discrete solution as a driver stage and I do believe we should for that reason alone if nothing else.
Regards,
Chris
Hmm, discrete driver stage and high power are somewhat at odds with each other, I think. It may become hard to find fast transistors that can pass high currents with enough gain. Then, there is the problem of coupling the signal to them. Their input impedance may not be low enough, meaning power dissipation problems in the current mirror supplying that signal. These are my observations from several months ago when I was able to simulate the discrete UcD circuit, though I did not get the important Baker clamps into the circuit correctly.
I guess I will throw in an aside comment here since I have already mentioned that I did some simulations on the circuit. I never was good with all that distortion and FFT stuff though. Well, I need bridging, and I could not get stable operation that way. I seemingly had a problem with the resonance of the output filter making an untidy appearance and robbing power and fidelity from the bridge configuration.
Chris,
I am pleasantly impacted by your experiences with LTspice. I am glad you gave it try. Another great thing about it is that it has no circuit size limitations unlike other free simulators.
I wouldn't exactly pair SMD with pleasure. It is rather hell on my eyes even with 275 diopter reading glasses. I could use the set-up where I could place a magnifying loupe over the right lens.
I am pleasantly impacted by your experiences with LTspice. I am glad you gave it try. Another great thing about it is that it has no circuit size limitations unlike other free simulators.
I wouldn't exactly pair SMD with pleasure. It is rather hell on my eyes even with 275 diopter reading glasses. I could use the set-up where I could place a magnifying loupe over the right lens.
Doesn't hurt to give a new/different program a fair shot every now and then, to be honest if I keep using it a bit more and find it's not very hard to add competitor models to it, the ease of..or perhaps...fluidity? of it's use will have me sold.
I just took a look at a motherboard I have laying around, WOW...they sure make em small don't they? Since you have more knowledge than I, perhaps since this isn't going to be headed for a robotic assembly line, would it be possible to limit ourselves to a certain size (the biggest available) of SMT, if so, how bad would that be to work with? I'm extremely skilled when it comes to working with my hands but I just don't see myself soldering anything the size of a breadcrumb, if I can see it or not.
I just took a look at a motherboard I have laying around, WOW...they sure make em small don't they? Since you have more knowledge than I, perhaps since this isn't going to be headed for a robotic assembly line, would it be possible to limit ourselves to a certain size (the biggest available) of SMT, if so, how bad would that be to work with? I'm extremely skilled when it comes to working with my hands but I just don't see myself soldering anything the size of a breadcrumb, if I can see it or not.
hi.
allow me to make a few comments based on a bit of real life experience in this field;
* Single-ended output with dual supplies should be OK. Although H-bridges has distinct advantages, they come at a cost people seem unwilling to take for the "starting point" circuit.
the added cost for a full-bridge design isnt very big and for higher powers no added cost really , but several advantages.
* As for power levels, limiting it is a good idea for beginners, although some expressed that there's no point with class D if power ability is below some level. Break-point seems to lie around 100W (?).
200-400W should be well within reach of a half bridge design, 500W+ for a full bridge design.
* Double-sided PCB is not a requirement per se, but a must for performance! Try to keep upper-side for ground plane mostly.
double sided will soon prove to be a must
* Try to use only through-hole components (?). SMD has size/space advantages, and most likely performance ones too, but save that for the "ultimate" circuit ?
through-hole design will soon show to be waste of time
* Separate input+control loop from output stage, for easy replacement/refinement of each.
decide if you want to have feedback before or after the output filter and decide if you want to make a clock based or a self-osc design.
rgds - karsten madsen - www.cadaudio.dk
allow me to make a few comments based on a bit of real life experience in this field;
* Single-ended output with dual supplies should be OK. Although H-bridges has distinct advantages, they come at a cost people seem unwilling to take for the "starting point" circuit.
the added cost for a full-bridge design isnt very big and for higher powers no added cost really , but several advantages.
* As for power levels, limiting it is a good idea for beginners, although some expressed that there's no point with class D if power ability is below some level. Break-point seems to lie around 100W (?).
200-400W should be well within reach of a half bridge design, 500W+ for a full bridge design.
* Double-sided PCB is not a requirement per se, but a must for performance! Try to keep upper-side for ground plane mostly.
double sided will soon prove to be a must
* Try to use only through-hole components (?). SMD has size/space advantages, and most likely performance ones too, but save that for the "ultimate" circuit ?
through-hole design will soon show to be waste of time
* Separate input+control loop from output stage, for easy replacement/refinement of each.
decide if you want to have feedback before or after the output filter and decide if you want to make a clock based or a self-osc design.
rgds - karsten madsen - www.cadaudio.dk
Hi, your input is most welcomed.
Is this with respect to the supply pumping issue?
I think we decided on a self oscillating type for it's ease simplicity and other advantages.
I think said self oscillating was decided to be a phase shift type of oscillator, as per ucd, so feedback after the filter, but not limited to? Maybe we'll wind up with both, it wouldnt' hurt to add provisions for taking feedback from either point would it, more room to experiment should anyone want to.
Regards,
Chris
Is this with respect to the supply pumping issue?
I think we decided on a self oscillating type for it's ease simplicity and other advantages.
I think said self oscillating was decided to be a phase shift type of oscillator, as per ucd, so feedback after the filter, but not limited to? Maybe we'll wind up with both, it wouldnt' hurt to add provisions for taking feedback from either point would it, more room to experiment should anyone want to.
Regards,
Chris
Hi Km,
Too, thanks for the input.
I did just today see that Philips has a new mosfet package which is basically a one-sided SO-8 package with a small heatsink tab. They said that it it is good up to 2 mhz because it does not use internal wire leads to connect its pins to the silicon substrate. That style of mosfet may have a future in lessening switching problems in power stages.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.