Good afternoon all,
I am looking for a proven design, if only for the sake of interest, for a high-powered class D subwoofer amplifier. The general requirements would be:
- In the range of 750 to 1000 watts peak output power
- Stable with a 2 ohm load impedance ideally (4 ohms or less would be nice)
- For frequency response, a -3 dB lower knee of 10 to 15 Hz
For the purpose of this discussion, I will omit power supply requirements. Do any of you know of such a proven design? If not, would there be interest in a collaborative design of sorts? I am new to class D, and as such have no experience to offer. However, I have a background in electronics, and might be able to propose enough silly ideas to keep things interesting!
Jim
I am looking for a proven design, if only for the sake of interest, for a high-powered class D subwoofer amplifier. The general requirements would be:
- In the range of 750 to 1000 watts peak output power
- Stable with a 2 ohm load impedance ideally (4 ohms or less would be nice)
- For frequency response, a -3 dB lower knee of 10 to 15 Hz
For the purpose of this discussion, I will omit power supply requirements. Do any of you know of such a proven design? If not, would there be interest in a collaborative design of sorts? I am new to class D, and as such have no experience to offer. However, I have a background in electronics, and might be able to propose enough silly ideas to keep things interesting!
Jim
Good evening,
It would seem such a design is rare. Would there be value in working together, on an open-source design for such an application? Just for fun, here are some (newbie!) thoughts.
The design goals:
- low frequency extension (down to 15 Hz @ -3 dB would be nice)
- able to drive low impedance loads (4 ohms or less)
- high power output (750 W peak or greater into 4 ohms within usable BW)
- modest parts count if possible
- reasonable cost
- reliable
Compromises:
- THD (within reason, distortion content is not as important)
- physical size (groups of output devices, for example)
- efficiency if need be (current sharing resistors added if needed, for example)
***
Just to get the ball rolling, here is a rough draft of the building blocks.
IRS2092 looks like a nice front end chip. My first thought is to mate it with a gate driver IC, namely IRS2011. IRS2092 appears (from the reference design) to be intended for half-bridge use. This brings us to the first of my newbie questions: could I use it in a full-bridge design, using multiple gate drivers? Also, if I wanted to use additional output devices, can I simply add gate drivers? I would imagine PCB trace length may become an issue in order to maintain synchronous switching, but that is getting ahead of things.
More to come, it's late for now!
Jim
It would seem such a design is rare. Would there be value in working together, on an open-source design for such an application? Just for fun, here are some (newbie!) thoughts.
The design goals:
- low frequency extension (down to 15 Hz @ -3 dB would be nice)
- able to drive low impedance loads (4 ohms or less)
- high power output (750 W peak or greater into 4 ohms within usable BW)
- modest parts count if possible
- reasonable cost
- reliable
Compromises:
- THD (within reason, distortion content is not as important)
- physical size (groups of output devices, for example)
- efficiency if need be (current sharing resistors added if needed, for example)
***
Just to get the ball rolling, here is a rough draft of the building blocks.
IRS2092 looks like a nice front end chip. My first thought is to mate it with a gate driver IC, namely IRS2011. IRS2092 appears (from the reference design) to be intended for half-bridge use. This brings us to the first of my newbie questions: could I use it in a full-bridge design, using multiple gate drivers? Also, if I wanted to use additional output devices, can I simply add gate drivers? I would imagine PCB trace length may become an issue in order to maintain synchronous switching, but that is getting ahead of things.
More to come, it's late for now!
Jim
Last edited:
Good evening,
It would seem such a design is rare. Would there be value in working together, on an open-source design for such an application? Just for fun, here are some (newbie!) thoughts.
The design goals:
- low frequency extension (down to 15 Hz @ -3 dB would be nice)
- able to drive low impedance loads (4 ohms or less)
- high power output (750 W peak or greater into 4 ohms within usable BW)
- modest parts count if possible
- reasonable cost
- reliable
Compromises:
- THD (within reason, distortion content is not as important)
- physical size (groups of output devices, for example)
- efficiency if need be (current sharing resistors added if needed, for example)
***
Just to get the ball rolling, here is a rough draft of the building blocks.
IRS2092 looks like a nice front end chip. My first thought is to mate it with a gate driver IC, namely IRS2011. IRS2092 appears (from the reference design) to be intended for half-bridge use. This brings us to the first of my newbie questions: could I use it in a full-bridge design, using multiple gate drivers? Also, if I wanted to use additional output devices, can I simply add gate drivers? I would imagine PCB trace length may become an issue in order to maintain synchronous switching, but that is getting ahead of things.
More to come, it's late for now!
Jim
Excellent works!I think self osc topic is not good choose,we have developed one module use discrete parts like JBL EON518S or EON515 active modules,(JBL :: Product)
but we not use it at 10Hz low frequency,we cut off 40Hz low frequency due to woofer can't works and out of hearing.fixed frequency class D should good choose and usually use full bridge and low voltage added when 2ohm load,
If you like you can connect us ,our mail:authlxl@gmail.com
Try to look here first: http://www.diyaudio.com/forums/class-d/166214-ucd-25-watts-1200-watts-using-2-mosfets.html
Thank you for your comments. A commercial product may be the best solution here, but for interest's sake I would like to consider an alternate DIY design. Thank you for the email, however - I will keep that on hand!
Jim
Try to look here first: http://www.diyaudio.com/forums/class-d/166214-ucd-25-watts-1200-watts-using-2-mosfets.html
Good afternoon,
Thank you for your comments. Hmm, that's interesting: so that circuit has a descrete PWM front end and a high-voltage gate-driver IC? An interesting design, but for heavy-duty, long-term use I wonder how such a few number of output devices would fare.
Hey again all,
Here is a quick drawing I made to go along with some questions, concerning the IRS2092 IC. Could you do something like the following to use the IRS2092 in a full-bridge design? If VB and VCC were supplied with low (TTL) level voltages, could HO and LO be used to drive dedicated gate driver IC's in this way? My goal here is to drive multiple pairs of output fets, with each pair having its own driver. My guess is that the bootstrap circuitry will not be manipulated like this. Any thoughts? (be gentle, I'm new!)
edit: this drawing is intentionally simplified for the purposes of this discussion!
Jim
Here is a quick drawing I made to go along with some questions, concerning the IRS2092 IC. Could you do something like the following to use the IRS2092 in a full-bridge design? If VB and VCC were supplied with low (TTL) level voltages, could HO and LO be used to drive dedicated gate driver IC's in this way? My goal here is to drive multiple pairs of output fets, with each pair having its own driver. My guess is that the bootstrap circuitry will not be manipulated like this. Any thoughts? (be gentle, I'm new!)
edit: this drawing is intentionally simplified for the purposes of this discussion!
Jim
Last edited:
Hey again all,
I may have killed the conversation by considering too many things at once. Here is a question then, that perhaps someone with experience might be able to help me with:
Assuming I'd like to go with a full-bridge design, is the IRS2092 a good choice as the front-end?
While considering this question, I have been thinking a bit about the basic circuit structure of such a design. One method of implementing a full-bridge makes use of 2 IRS2092 IC's, driven with a differential signal, generated by inverting the incoming audio signal to 1 of the IC's, with each IC running in half-bridge mode. Is this the preferred method? Would it be possible to use a single IRS2092, and a gate-drive-buffer section to drive all 4 fet groups?
Any and all comments are appreciated. This past weekend I managed to get a free schematic layout / PCB package up and running at home, so I will start to post some schematics as things progress.
Jim
I may have killed the conversation by considering too many things at once. Here is a question then, that perhaps someone with experience might be able to help me with:
Assuming I'd like to go with a full-bridge design, is the IRS2092 a good choice as the front-end?
While considering this question, I have been thinking a bit about the basic circuit structure of such a design. One method of implementing a full-bridge makes use of 2 IRS2092 IC's, driven with a differential signal, generated by inverting the incoming audio signal to 1 of the IC's, with each IC running in half-bridge mode. Is this the preferred method? Would it be possible to use a single IRS2092, and a gate-drive-buffer section to drive all 4 fet groups?
Any and all comments are appreciated. This past weekend I managed to get a free schematic layout / PCB package up and running at home, so I will start to post some schematics as things progress.
Jim
Hi J. R Freeman
If this is your first desing, I would go for a half bridge first, to get to know how to make class-d work.
If you are going to drive FETs with high Ciss use a buffer stage after the IRS2092 in stead of using a IRS2011 (Zetex makes appropiate transistors for this).
Don't go for groups of FETs .... use as few as possible to minimize problems with layout etc. ..... you can get fast TO220 FETs that can handle more than 50A!!!!
Go for something smaller in your first go .... say 200W until you have a good working design .... class-d is nothing like a "normal" analog amp
If this is your first desing, I would go for a half bridge first, to get to know how to make class-d work.
If you are going to drive FETs with high Ciss use a buffer stage after the IRS2092 in stead of using a IRS2011 (Zetex makes appropiate transistors for this).
Don't go for groups of FETs .... use as few as possible to minimize problems with layout etc. ..... you can get fast TO220 FETs that can handle more than 50A!!!!
Go for something smaller in your first go .... say 200W until you have a good working design .... class-d is nothing like a "normal" analog amp
Hi all,
Thank you for your comments. For the moment, this is a day-dream project. If it looks like it is going places, however, I do have layout experience. For now, this is just for fun.
Having said that, over the past while I've read up a little on the IRS2092. I think I now have a better idea of how the high-side boot-strap supply works. Here's a corner-stone problem in the quest for scalable numbers of output devices: how do you drive them without over-loading the gate drive outputs of the IRS2092?
Here is a new schematic. It's based on the IRaudamp7s design from IR. It is a stripped-down schematic of the high-side drive section, with the boot-strap supply.
HighDrive.pdf
What I have done here, to try and make things easier to read, is I've separated this part of the system into the high-side boot-strap section, and the output section. Please just imagine that the low-side is all connected up: I realize the boot-strap supply won't work without it, it is just not shown for clarity's sake.
Alright, so my first problem is scaling the boot-strap's available power. My feeling is that this is a simple case of using a large (or multiple) voltage follow transistors, and multiple blocking diodes. Also, I would likely want C1 to be large, would I not? What is the limit to the size of C1: transient current through the rest of the circuit?
Next, and more of a puzzle to me, is how to drive multiple output stages. Is the stage shown here a reasonable idea? I've just used an emitter-follower to buffer the gate drive output of the IRS2092. Is that correct? How do I determine the value of R1? I have a feeling it will mostly effect the turn-off time off time of the mosfet, as it is draining the gate charge away.
So does this attempt at a modular, expandable output section seem reasonable? Any comments are welcomed!
Jim
Thank you for your comments. For the moment, this is a day-dream project. If it looks like it is going places, however, I do have layout experience. For now, this is just for fun.
Having said that, over the past while I've read up a little on the IRS2092. I think I now have a better idea of how the high-side boot-strap supply works. Here's a corner-stone problem in the quest for scalable numbers of output devices: how do you drive them without over-loading the gate drive outputs of the IRS2092?
Here is a new schematic. It's based on the IRaudamp7s design from IR. It is a stripped-down schematic of the high-side drive section, with the boot-strap supply.
HighDrive.pdf
What I have done here, to try and make things easier to read, is I've separated this part of the system into the high-side boot-strap section, and the output section. Please just imagine that the low-side is all connected up: I realize the boot-strap supply won't work without it, it is just not shown for clarity's sake.
Alright, so my first problem is scaling the boot-strap's available power. My feeling is that this is a simple case of using a large (or multiple) voltage follow transistors, and multiple blocking diodes. Also, I would likely want C1 to be large, would I not? What is the limit to the size of C1: transient current through the rest of the circuit?
Next, and more of a puzzle to me, is how to drive multiple output stages. Is the stage shown here a reasonable idea? I've just used an emitter-follower to buffer the gate drive output of the IRS2092. Is that correct? How do I determine the value of R1? I have a feeling it will mostly effect the turn-off time off time of the mosfet, as it is draining the gate charge away.
So does this attempt at a modular, expandable output section seem reasonable? Any comments are welcomed!
Jim
Hi Jim!
All of your idea can be done, but I think the benefit of using 2092 is the built-in overcurrent protection, and very low component number. To enjoy this benefit you need to use it alone, or two of them in two complete amplifiers and then connect them to bridge (not 2092 + gate driver).
2092 is fairly expensive and in some countries hard to buy. If you want to make it usable worldwide, I suggest you to use some more common gate-driver.
Full bridge is good for very low freq response, but is it really neccessary?
You don't neccessarily need groups of FETs: with proper heat transfer and current sensing on Rds two pcs of IRFP4227 can deliver 800 W on 2 ohms or 4 ohms.
If you choose full-bridge, a simplification is possible: you don't have to make so many power supply. Only a power rail of 80...90 V and a low side supply (12V 150mA) is needed. In this case IRS2092 is cannot be used, but it also not needed either. Two simple IR2010S (or the cheaper 2110) can drive the gates. One modulator can generate PWM for both legs, and differential feedback can be done. Somewhere in this forum there were a successfully built design this way.
I prefer 3 level modulation in bridge for LF application, this lets the output filter be small, but this kind of modulator is significantly more complicated then a phase-shift self oscillating one.
All of your idea can be done, but I think the benefit of using 2092 is the built-in overcurrent protection, and very low component number. To enjoy this benefit you need to use it alone, or two of them in two complete amplifiers and then connect them to bridge (not 2092 + gate driver).
2092 is fairly expensive and in some countries hard to buy. If you want to make it usable worldwide, I suggest you to use some more common gate-driver.
Full bridge is good for very low freq response, but is it really neccessary?
You don't neccessarily need groups of FETs: with proper heat transfer and current sensing on Rds two pcs of IRFP4227 can deliver 800 W on 2 ohms or 4 ohms.
If you choose full-bridge, a simplification is possible: you don't have to make so many power supply. Only a power rail of 80...90 V and a low side supply (12V 150mA) is needed. In this case IRS2092 is cannot be used, but it also not needed either. Two simple IR2010S (or the cheaper 2110) can drive the gates. One modulator can generate PWM for both legs, and differential feedback can be done. Somewhere in this forum there were a successfully built design this way.
I prefer 3 level modulation in bridge for LF application, this lets the output filter be small, but this kind of modulator is significantly more complicated then a phase-shift self oscillating one.
Hi there Pafi!
Thank you very much for the time you put into your response. I appreciate the chance to learn from someone with more experience.
Hmm, I see what you mean about the application of IRS2092. I suppose if it has a gate driver section built-in, it was intended to be used with a minimum number of external parts. It is a very attractive IC, however, because it takes care of so many things at once.
Regarding multiple output devices, I was impressed to hear the power potential of 2 IRFP4227's. I will have to keep that part number in mind! However, wouldn't groups of output devices be more reliable? I would rather have a group of MOSFET's working at 50% of their rated load vs. fewer devices approaching 90%. But I understand that multiple output devices complicates the design and introduces parasitic losses and timing issues (though I must admit I don't understand the extent of these gotcha's ).
IRS2010S looks very interesting. IR seems to have a few high-powered gate drivers available. That is nice, because as you said, a single PWM front-end could control a number of gate drivers. Hmm, I will be day-dreaming of this instead of working this afternoon, likely
That being said, here is a question then for you: do you know of an IC that takes car of the PWM - that has only a PWM output? I know that is a shameful question, but the idea of a discrete PWM front end with dead time and all is a scary thought to a beginner such as myself. For a single channel subwoofer work-horse amp, I would gladly sacrifice some fidelity for simplicity (another sacrilegious thought!).
Thanks again for your help. I'm already learning! Gate drive is more complex than I had thought originally - I suppose it's not just about charging the gate quickly, but also discharging as well.
Jim
Thank you very much for the time you put into your response. I appreciate the chance to learn from someone with more experience.
Hmm, I see what you mean about the application of IRS2092. I suppose if it has a gate driver section built-in, it was intended to be used with a minimum number of external parts. It is a very attractive IC, however, because it takes care of so many things at once.
Regarding multiple output devices, I was impressed to hear the power potential of 2 IRFP4227's. I will have to keep that part number in mind! However, wouldn't groups of output devices be more reliable? I would rather have a group of MOSFET's working at 50% of their rated load vs. fewer devices approaching 90%. But I understand that multiple output devices complicates the design and introduces parasitic losses and timing issues (though I must admit I don't understand the extent of these gotcha's
).IRS2010S looks very interesting. IR seems to have a few high-powered gate drivers available. That is nice, because as you said, a single PWM front-end could control a number of gate drivers. Hmm, I will be day-dreaming of this instead of working this afternoon, likely
That being said, here is a question then for you: do you know of an IC that takes car of the PWM - that has only a PWM output? I know that is a shameful question, but the idea of a discrete PWM front end with dead time and all is a scary thought to a beginner such as myself. For a single channel subwoofer work-horse amp, I would gladly sacrifice some fidelity for simplicity (another sacrilegious thought!).
Thanks again for your help. I'm already learning! Gate drive is more complex than I had thought originally - I suppose it's not just about charging the gate quickly, but also discharging as well.
Jim
I don't like the dead time management on IRS2092 and similar ICs, in practice it forces either a too high dead time or too slow switching (di/dt), a bad compromise which translates into an undesirable THD vs efficiency compromise.
My choice for previous projects has been IR2110 because it allows to control switching timing properly (with separate low and high side inputs) and because relative on/off propagation delay is quite consistent against temperature changes.
In the next 2 projects that are planned I'm probably going to try IRS2011 because it features considerably lower propagation delay, which is a source of THD in the self-oscillating scheme that I use.
At first I didn't use buffer transistors for gate drive, but when I tried them I found that MOSFET turn off could be made even faster resulting in less switching losses. I also found that by adding some extra parts I could control body diode di/dt quite precisely and have matched gate voltages on low and high sides (higher efficiency, more power, lower THD), so I consider them a must now. Another interesting finding is that using only as much gate drive voltage as required allows to optimize the circuit for even lower switching losses (driving the gate "harder" at a lower voltage and relying on source lead inductance for di/dt control). For IRFB4227 I use only 8.5 volts at the gate. Lead length (and related PCB tracks) matter a lot on class D output MOSFET.
Four IRFB4227 can do 3kw on 4 ohm in an optimized circuit.
EDIT: Cascading gate driver ICs (ie: IRS2092+2*IR2011) is not a good idea because propagation delay adds up. They are not intended to be used that way. Consider discrete level shifting instead.
My choice for previous projects has been IR2110 because it allows to control switching timing properly (with separate low and high side inputs) and because relative on/off propagation delay is quite consistent against temperature changes.
In the next 2 projects that are planned I'm probably going to try IRS2011 because it features considerably lower propagation delay, which is a source of THD in the self-oscillating scheme that I use.
At first I didn't use buffer transistors for gate drive, but when I tried them I found that MOSFET turn off could be made even faster resulting in less switching losses. I also found that by adding some extra parts I could control body diode di/dt quite precisely and have matched gate voltages on low and high sides (higher efficiency, more power, lower THD), so I consider them a must now. Another interesting finding is that using only as much gate drive voltage as required allows to optimize the circuit for even lower switching losses (driving the gate "harder" at a lower voltage and relying on source lead inductance for di/dt control). For IRFB4227 I use only 8.5 volts at the gate. Lead length (and related PCB tracks) matter a lot on class D output MOSFET.
Four IRFB4227 can do 3kw on 4 ohm in an optimized circuit.
EDIT: Cascading gate driver ICs (ie: IRS2092+2*IR2011) is not a good idea because propagation delay adds up. They are not intended to be used that way. Consider discrete level shifting instead.
Last edited:
Hi Pafi,
So it is simply a triangle wave at one input, and the audio signal at the other of a comparator? There must be some extra things to work out, like the amplitude of the triangle wave, when considering the amplitude of your maximum expected audio signal. Would you have a reference to such a front-end? Don't worry if you don't, I will do some more research.
Thanks for your comments.Jim
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.