| lumanauw |
I found 2 schematic, that may be working, but it seems wrong somewhere.
In this one, the ground is attached to -56V? And the SG3524 oscilator section (pin 6 and pin 7) is not attached to ground? |
|
|
| lumanauw |
And this one, is this idea can really work?
I assume the 1:3 output transformer is the car smps trafo made of ferrite toroidal core? Or should be low frequency metal OT transformer? |
|
|
| Pierre |
I think that the simplicity of the first sch, the Class-D design, should encourage us to try to complete it, at least to give it a try (it is obvious that the circuit is not well drawn, but some things can be fixed easily).
The oscillator cap and res (pin6 6 and 7) must go to the GND reference of the SG3524 chip. The thing is, did the author want to reference it to -56V rail? If so, the only level shift required for the PWM signal is for the high side. But in that case, a capacitive coupling of the input is mandatory, and I can't see it, so I think that its reference is GND, so there is an error.
The driver stage has NPN/PNP drivers. That's good, but the lower one should be fed at +12V referenced to the negative rail, which is not clear in the sch. And the high side one has a bootstrap diode+cap, but the diode anode should go to that same +12V supply, not to the +56V rail, unless there is a 12V zener in parallel with the bootstrap cap to prevent overvoltage.
The level shifting part is not clear. If the input stage is referenced to GND (it should be), level shifting is required for both parts (HS and LS).
That's easy to fix, we can use a PNP for the LS drive, but that's a bit more difficult for the HS part.
Another thing that worries me is the capability of SG3524 chip to generate a 0 to almost 100% PWM (from what I know it goes from 0 to 45% at each output, although this is wired a bit different in this sch...)
Where did you find this. Do you have any evidence that a variation of this circuit is working anywhere (it is obvious that it can't as it is drawn) |
|
|
| Pierre |
About the second circuit, the PWM modulator is quite straightforward, and the power conversion circuit is very similar to a DC/DC push-pull inverter, with buffered gate drivers.
The "secondary" part uses the transformer inductance in combination with the .1uF capacitor as the output filter.
However, due to the imperfections in the transformer (leakage inductance producing spikes, and its output going int o discontinuous mode, etc, I doubt this can produce an accurate representation of the input signal... unless someone demonstrates the opposite, of course! |
|
|
| lumanauw |
Hi, Pierre,
I found this on a member's website, Dan Fraser.
Do you think the first one (SG3524) based is more likely to work? I also think the same thing, because in the second one, there will be not-balanced magnetic flux on the transformer, will make it hot or saturated.
I wanted to build a lo-fi subwoofer amp, no need to be hi-fi, just be loud enough and runs cool :D And because it is for lo-fi subwoofer, maybe the oscilating frequency can be below 100khz or even below 50khz? (to fit the IC's purpose, usually SMPS IC's works below 100khz)
If the T-on is not close enough to 50% (just 45% like you said), I think it's OK for my application, it wont be a hi end piece.
About the T-on. Is it better to use TL494 to get T-on closer to 50%?
But....will it work AT ALL? |
|
|
| Pierre |
Lumanauw,
You may definitely give a try to the first circuit, but don't waste your time with the second one. But even that needs some modifications to start its basic operation.
I have to think more about it (there are people much more capable in this forum anyway), but the problem with SG3524 is not that it doesn't reach 50% (the only problem with that is that it won't swing to the supply rails, but that has its advantages also, as it won't never stop oscillation at clipping). What really worries me is the SG3524 output arrangement: I need to draw the waveforms in order to see if that resembles the required PWM format. Perhaps.
Another point to take care of are the level shiftings. I would modify that so the SG3524's PWM output is ground referenced. Then level-shift it so it swings between -56V and 12V above that (-44V). Then you can need to have it level shifted for the HS driver:
The second level shifter could be something like that:
Use a NPN transistor (rated at 150V at least). PWM (after first level shift) goes to base. Emitter goes to -56V with a 1k resistor. Collector goes to bootstrap voltage via a 1k resistor. Output of the level shifter is taken from collector, and then to the NPN/PNP HS driver.
Note that this arrangement is inverting, so you don't need additional inverters. The first level shift also inverts, so in fact you don't need any inverter.
The bootstrap voltage can be easily obtained: 10uF capacitor negative terminal goes to mosfets midpoint (bridge output). Positive terminal is fed from +12V (referenced to GND) via a schottky diode. The join point supplies 12V referred to HS mosfet source and that's all.
I have been thinking and writing at the same time, so please post your comments or questions if something is not clear or wrong. |
|
|
| lumanauw |
Hi, Pierre,
I cannot catch your drawing :(
Do you have your drawing of your idea above? |
|
|
| Pierre |
| I will try to draw a small draft. Perhaps it won't work but I am sure we can make it work with the help of the very valuable knowledge of some people in this forum. |
|
|
| Workhorse |
Hi Lumanauw,
Use SG3524 + IR2111 + 2 X IRFP250N
and you have very simple but an effective good quality Class-D amp, because i have tried it and made it for a subwoofer...
Set the Oscillator frequency around 40KHZ, wire the open collector outputs of PWM IC in "OR" configuration i.e. join the collectors together and add a pull up resistor of 1K to +15V supply, configure the voltage error opamp in that IC for gain of 10X , Ground the current error opamp's pin , fed the output to the IR 2111 gate driver [inbuilt 650nS deadtime]and finally to the mosfets....filter it with inductor and boom the subwoofer....
regards,
K a n w a r |
|
|
| Pierre |
Kanwar,
Did you use the SG3524 referenced to GND and the IR2111 to -Vrail? In that case, you need level shifters.
If not, you will need input AC couple.
Did you use feedback from the output of the mosfets or output filter, or simply locally, at the SG3524 opamp?
Please, could you send an schematic of what you did (specially input section and level shifters, if present). With that we can try to optimize the design further. Just as a curiosity and at the risk of become scared :) what power did you produce this way?
If you configure the SG3524 chip to 40KHz and wire it in "or" output, you will get an effective PWM at 80KHz, is that right?
In theory, you can use that chip at 100KHz ( effective 200KHz ), so you can start thinking on a full-range amplifier.
A suggestion: if you need some more fidelity, I would use another similar driver (IR2110 or IR2113) so you can set-up dead-times more tightly. However, for subwoofer, you need a lot of power and reliability, so using BIG mosfets and BIG dead-time will do better.
Keep on with this interesting thread!!
Best regards. |
|
|
| Pierre |
Let's see...
This is what I was trying to describe. Please take the general idea. Input is PWM from modulator (SG3524 or whatever).
Then it is level-shifted and inverted at once, so it attacks directly the LS driver. Then it is level-shifted again, with another inversion, and that attacks the HS driver, which has a bootstrap supply.
It needs +12V supply to GND for modulator, and +12V referrenced to -56V for the driver (as usual).
If your modulator is running referrenced to GND, you will always need level shifting as any half-bridge driver will be referrenced to -56V. (unless you use fast and expensive optocouplers ;)
That's what I am curious about Kanwar's design, because if he doesn't use level shifters then his modulator is running referrenced to -56V, and that is tricky.
Please post your comments on this. |
|
|
| Workhorse |
| quote: | Originally posted by Pierre
Kanwar,
Did you use the SG3524 referenced to GND and the IR2111 to -Vrail? In that case, you need level shifters.
If not, you will need input AC couple.
Did you use feedback from the output of the mosfets or output filter, or simply locally, at the SG3524 opamp?
Please, could you send an schematic of what you did (specially input section and level shifters, if present). With that we can try to optimize the design further. Just as a curiosity and at the risk of become scared :) what power did you produce this way?
If you configure the SG3524 chip to 40KHz and wire it in "or" output, you will get an effective PWM at 80KHz, is that right?
In theory, you can use that chip at 100KHz ( effective 200KHz ), so you can start thinking on a full-range amplifier.
A suggestion: if you need some more fidelity, I would use another similar driver (IR2110 or IR2113) so you can set-up dead-times more tightly. However, for subwoofer, you need a lot of power and reliability, so using BIG mosfets and BIG dead-time will do better.
Keep on with this interesting thread!!
Best regards. |
Hi Pierre,
Thanks for the interest!,
I donot use level shifters, no feedback just an openloop design.....
The Oscillator frequency with outputs wired as OR ed remains same and it only get halfed when the outputs are operated seperatedly....
The GND of PWM IC was Tied to COM of Gate Driver IC that is -V rail , the input was AC coupled...
Frequency Response = 10Hz to 1KHZ Flat....
Power = 1KW [bridged]
I use 2 PWM IC ....One master and other as a slave to get 3 Level PWM at bridged outputs.....
Since the project was just for experimentation as well as for Subwoofer application so I never used it for Hi-Fi purposes....
The Bass output was just Tight and had a amazing Depth,
I will post the schematics , but first i have to prepare them....
regards,
K a n w a r |
|
|
| Pierre |
You are right about there is no frecuency doubling.
You are right also in that AC coupling is needed. Don't you have startup problems due to the input being biased to a very negative voltage (input cap charging, etc)?
Your rise/fall times should be very slow, due to the low current sink/source capability of the IR2111 and how big the mosfets are. In the order of 200ns or so, right?
I once tried something similar but with feedback. The only thing is that it must be AC coupled as well, but then you have some DC offset inestabilities at the output. I am sure it can be solved with some care, perhaps I will give it another try.
Input configuration of biasing of SG352X chips is tricky. An sch would help.
About freq. response, if you increased freq. to, say, 200KHz, you could get good results in full range, and also smaller carrier residue. At the cost of reduced maximum duty cycle, say 85-90%.
I am surprised you "only" went to 1kW bridged. With that mosfets you could go higher if you wanted to! |
|
|
| Workhorse |
| quote: | Originally posted by Pierre
You are right about there is no frecuency doubling.
You are right also in that AC coupling is needed. Don't you have startup problems due to the input being biased to a very negative voltage (input cap charging, etc)?
Your rise/fall times should be very slow, due to the low current sink/source capability of the IR2111 and how big the mosfets are. In the order of 200ns or so, right?
I once tried something similar but with feedback. The only thing is that it must be AC coupled as well, but then you have some DC offset inestabilities at the output. I am sure it can be solved with some care, perhaps I will give it another try.
Input configuration of biasing of SG352X chips is tricky. An sch would help.
About freq. response, if you increased freq. to, say, 200KHz, you could get good results in full range, and also smaller carrier residue. At the cost of reduced maximum duty cycle, say 85-90%.
I am surprised you "only" went to 1kW bridged. With that mosfets you could go higher if you wanted to! |
Hi Pierre,
I think your interest in my circuit is just increasing!
Since I use Class-BD mode 3-level PWM and ofcourse H- bridge , therefore the startup problem wasn't encounter because both sides of bridge startup simuntaneously and the difference voltage at output was just between 50mV.....
The IR2111 just Drives the Mosfets well because the switching frequency is only 40KHZ which is not a very High as compared to 250KHZ
The Mosfets were IRFP250N , However with more big rails one can easily increase the power output to higher Kilowatt levels +2KW easily....
The input error amp is biased to 1/ 2 of VREF of PWM IC, hence there is no problem, one can also adjust the DC offset by varying the bias....
SG3524 is good upto 100KHZ only, BUT LM3524 from National Semiconductor Features dedicated HF Oscillator and 48% X 2 Modulation Factor at HF....
One can also use Higher power Mosfets to produce Very Large Power Bass Outputs, but its of his own interest only.....
regards,
K a n w a r;) |
|
|
| phase_accurate |
You could as well reference the controller to the negative rail and shift the audio signal instead of the control signals. You could use an LTP for this purpose. This would be cheap and elegant and it has already been done by Ivan (IVX) within his hysteresis modulator based amp (and others maybe as well).
Regards
Charles |
|
|
| lumanauw |
Hi, Pierre, Kanwar,
Thanks for the ideas :D
Do you have the Full Bridge one? To advoid bus pumping when using half-bridge+heavy work on subwoofer?
Pierre,
Your gate driver proposal is clever. I like it, because the purpose of this project to me is low cost, lo-fi subwoofer amp, so this is a good solution, compared if I have to buy those expensive Hewlett Packard power optocouplers :D. Is it free from crossconduction?
Hi, Kanwar,
Waiting forward for your schematic of this lo-fi classD :D
As Pierre said, do you also have the same lo-cost, lo-fi subwoofer classD that based on TL494 (not SG3524)? |
|
|
| Workhorse |
| quote: | Originally posted by lumanauw
Hi, Pierre, Kanwar,
Thanks for the ideas :D
Do you have the Full Bridge one? To advoid bus pumping when using half-bridge+heavy work on subwoofer?
Pierre,
Your gate driver proposal is clever. I like it, because the purpose of this project to me is low cost, lo-fi subwoofer amp, so this is a good solution, compared if I have to buy those expensive Hewlett Packard power optocouplers :D. Is it free from crossconduction?
Hi, Kanwar,
Waiting forward for your schematic of this lo-fi classD :D
As Pierre said, do you also have the same lo-cost, lo-fi subwoofer classD that based on TL494 (not SG3524)? |
Hi LumanauW,
TL494 couldnot be used because the output of error amp and the input of its comparator, has a diode in the series in between them....
Dear Charles,
The LTP is a good choice,as it eliminate seperate supplies when you simply reference the LM360+IR2113/2110 to Negative rail of Mosfets...
Is there is a need of error amp , certainly not because i think LTP acts as an error amp itself....
correct me if i am wrong
regards,
K a n w a r |
|
|
| Pierre |
About the proposed transistor circuit, it is only an idea and should be polished and experimented for cross-conduction, etc, although that can be somewhat controlled. with resistor+diode in parallel at the gates.
Is it possible to have the SG3524 referenced to -VSS with a half-bridge, and AC couple the input? If so, what are the possible problems? And how to implement feedback correctly then? I once did that simply by AC coupling the feedback also, but that way DC was not corrected properly. I am sure there is another simple but better way to do it.
Let's try to keep the design as simple as possible! |
|
|
| lumanauw |
Hi, Pierre,
I think it is the best that SG3524 is referenced to absolute 0V (gnd), like you said, to get best 0 DC offset. We can use of your level shifter, and make 12V+ relative to -56V, no need to make floating power supply, just make it with 1 zener, 1 transistor, 1 resistor (this transistor can be mosfet, runs quite hot, but making everything simpler, providing all mA needed by the gate drivers). |
|
|
| Pierre |
Yes, making the 12V supply referred to -56V is not a problem at all. If you see my schematics in the "Turn on/off transients" thread, you will see how I solved it: just a resistor, 12V regulator and decoupling cap, and power dissipation is quite moderate.
The first level shifter is well proven also in my design, but I haven't implemented the second in any of my prototypes, as well as the discrete drivers. I guess they shouldn't produce any problems.
I agree in that it is better to have the modulator referrenced at GND.
Best regards,
Pierre |
|
|
| Workhorse |
Hi Pierre,
The Feedback could be very easily referenced to Midpoint by using a level shifter at the output of SG3524 , GND reference the IC and then Level shift its output to -Vrail and connect it to the input of IR2111 and drive the mosfets and then take the feedback and apply it to the inverting pin of error amp of PWM IC.....
regards,
K a n w a r |
|
|
| Pierre |
Thanks, Kanwar, but I think that was clear from the start (that's what I did in my amp in fact, although with double level-shifter and a IR2113 chip :D)
What I was trying to figure out is how to simplify it to the extreme by referencing the modulator chip to -Vrail, so no level shifter is needed, but then you have to AC couple the input signal, and feedback cannot be connected directly unless you AC couple it, producing DC offset error problems.
Anyway, I think that's a dead-end, because in the best of cases, you may need some kind of level-shift in the feedback signal, voiding the advantage of simplicity we were pursuing.
IMHO we could keep on defining an sch based on SG3524, as Kanwar's amp. However, for the sake of simplicity, I would like to work a little bit more on the concept of discrete driving instead of using a IR chip. I proposed a partial schematic and we are waiting for suggestions, critics or improvements.
Best regards,
Pierre |
|
|
| lumanauw |
Hi, Pierre,
I really wanted to help with reviewing your idea of level shifter, but since I'm not familiar with classD design, nor I have other schematic to compare (the only SG3524 schematic that I have is on the first post, the defected one), I cannot help much here.
Maybe Kanwar can help with showing what he does with lo-fi SG3524 based classD? How about it, Kanwar? I think your commercial business wont be disturbed by showing this lo-fi design of yours. But it can teach us something ;) |
|
|
| Pierre |
I have detected an error in the sch I posted. The second level shifter collector resistor (R2) should go to the bootstrap voltage, not to +12V.
Anyway, I will try to draw an sch of my amplifier, removing my PWM modulator and exchanging it for a SG3524. That's pretty close to what Kanwar suggests.
Best regards,
Pierre |
|
|
| lumanauw |
Hi, Pierre,
Without additional opamp, just using the comparator inherent in SG3524? |
|
|
| Pierre |
Yes, I think it can be done.
I use a discrete modulator with input stage, triangle generator, comparator, etc, but everything is in the SG3524 (although you won't get the same level of performance ;-) |
|
|
| lumanauw |
Hi, Pierre,
Looking forward for the sch :)
It seems the Triangle generator is more like Sawtooth generator, is it a problem? |
|
|
| Pierre |
It shouldn't be a problem with the sawtooth generator. With the outputs wired in OR configuration, you still get 0 to around 90% duty-cycle.
I will try to draw it today so I can show it to you at the end of the day and comments can be received.
Only a detail: presumably, soft-start of the modulator should occur before the output drivers are activated, as 0% duty cycle will correspond to near -Vrail output. It should be stablished at 50% before output starts switching.
Best regards. |
|
|
| Workhorse |
| quote: | Originally posted by lumanauw
Maybe Kanwar can help with showing what he does with lo-fi SG3524 based classD? How about it, Kanwar? I think your commercial business wont be disturbed by showing this lo-fi design of yours. But it can teach us something ;) |
Hi David,
There's no problem on the schematics, but i have to draw it , then i would be able to post it......thats not my commercial project, I would certainly like to share it with you guys... ;)
K a n w a r |
|
|
| Kenshin |
| Is the high side gate driver of the first circuit useable? I saw some ringing in simulation of something like this. |
|
|
| Pierre |
Here is a proposal of a Class-D amp based on a SG3525 and IR mosfet driver chip, with only a level shifter.
The circuit is only a proposal and must be reviewed and perfectionated.
Some hints...
- SG3525 is very similar to SG3524. I am more used to it, but SG3524 has open collector output stages that make the diodes unnecessary.
- Switching freq. is around 80KHz here, but can be modified by means of CT1 and RT1.
- There is only a level shifting, it produces an inverted version of the PWM, that hence must attack directly to the low side driver input. The inverted version can attack the high side driver input.
- Using this chip this way doesn't provide shutdown. To implement it, either use an IC that provides it, or simply add an AND gate before each input, so the shutdown signal enters both gates in order to make both inputs 0 simultaneously when it is "0".
- Dead-time is only taken care of at the mosfet gates. For more controlled dead-time, either use a chip with dead-time control (such as LM5104, but that doesn't have SD), or implement dead-time with a RCD circuit before each input.
- The input arrangement must be reviewed. I have biased the +IN to Vref/2, and AC coupled the input, but I don't know if that's totally ok.
- Compensation is not added. Perhaps it needs to be done.
- I haven't added the necessary supply bypassing caps near the mosfets. Those are taken for granted!
I hope this provides a good point to start working on for a mid-fi and very low cost Class-D amplifier, at least for subwoofer use.
Best regards,
Pierre |
|
|
| lumanauw |
Hi, Pierre,
Very nice, thank you :D
SG3525 has totem pole output inherent in the chip, compared to SG3524, is this contributing something that makes you choose SG3525?
It consist only of 3 vital components (or 4, +4049 IC), that's good for this project :D
To make it even simpler, maybe we can use IR2111, low cost, and can omit the 4049 IC? I can buy IR2111 here, quite cheap. For others, I should import it, making them will be expensive.
Oh yes, Pierre. I always wonder how to make classD a full bridge (like HIP4080 based). It won't have rail pumping due to heavy subwoofer work (when headed to half bridge topology).
Is it possible to make this lo-fi, lo-cost, SG3524 based classD a full bridge one?
Another one. How to calculate the compensation(s) / feedback networking for any classD? Does it needs Simulator, or can be done with calculator? |
|
|
| Pierre |
IR2111 can be ok, but it is quite slow (only about 400mA drive) and it doesn't have SD nor dead-time control. You really need shutdown in order to control start-up and add protections such as overcurrent.
I have chosen SG3525 only because I had the symbol created for Protel ;-) You can freely use SG3524 and that should be ok too. Connection is very similar.
About full-bridge: Yes, it can be done very easily. You only need two mosfet drivers (one for each "side"), but with the inputs inverted. The filter changes, too, and so does the feedback.
IMHO, it is better and easier to build two half-bridge amps and then connect them in bridge-mode. The only added component is another SG3524/5 and an opamp for signal inversion to one of them.
About feedback, if you use this for subwoofer where ultra-low distortion and frequency response doesn't need to be load independent, feedback can be taken before filter (ZAP pulse does that, for example), and compensation is very easy or perhaps simply solved by adding a cap between inverting input and COMP terminal, as shown.
That needs experimentation, however.
Hope you get the time to build one and tell us how it works! |
|
|
| Pierre |
I have been doing some tests with SG3524 trying to generate a 0% to 90% PWM. If I haven't done something wrong, the arrangement shown in the initial figure posted by lumanauw (both output transistors inside the chip connected in parallel, with emitters to GND and collectors with a pullup to Vcc, output at collectors), won't work: as both outputs won't never be active at the same time, at any moment one of the transistors is saturated and pulling down, so output will be always 0V.
I think the correct arrangement for OR configuration (output is high whenever one of the outputs is high), transistors must still be connected in parallel, but collectors must go to VCC, and emitters to output, with a 620ohm or so pulldown to GND.
It still remains to see if this arrangement inverts, in which case polarity of the PWM should be inverted (i.e, output of the level shifter should go to HS driver and after inverting it, to LS driver).
Just some ideas... |
|
|
| Workhorse |
| quote: | Originally posted by Pierre
I have been doing some tests with SG3524 trying to generate a 0% to 90% PWM. If I haven't done something wrong, the arrangement shown in the initial figure posted by lumanauw (both output transistors inside the chip connected in parallel, with emitters to GND and collectors with a pullup to Vcc, output at collectors), won't work: as both outputs won't never be active at the same time, at any moment one of the transistors is saturated and pulling down, so output will be always 0V.
I think the correct arrangement for OR configuration (output is high whenever one of the outputs is high), transistors must still be connected in parallel, but collectors must go to VCC, and emitters to output, with a 620ohm or so pulldown to GND.
It still remains to see if this arrangement inverts, in which case polarity of the PWM should be inverted (i.e, output of the level shifter should go to HS driver and after inverting it, to LS driver).
Just some ideas... |
Oh Pierre,
I am afraid, you are wrong buddy, whether you take outputs from collector or from emitters the only difference is phase inversion...
K a n w a r |
|
|
| phase_accurate |
Apart from the signal inversion (which is only a minor problem BTW) connecting the collectors is a more elegant solution than connecting the emitters. Such an arrangement is called a "wired OR" by the digital guys BTW.
Regards
Charles |
|
|
| Pierre |
Kanwar,
I suppose you are right, because I have also seen that arrangement (connecting the emitters to GND and the collectors together produce the output with a pullup to VCC), but the thing is that I must be missing something, because when one of the outputs goes low, its transistor is activated, pulling the signal down to 0, so theoretically (I repeat, I must be wrong), output should be always 0, as both outputs are never 1 at the same time ¿¿¿???
Anyway. I have tried with the output taken from the emitters and it seems to work well (0 to 93% at 125 KHz).
I have also started a simulation in order to understand feedback and biasing better. In the schematics shown below, the first opamp is SG3524 internal error-amp. The other simulates the gain of the modulator and power stage.
Note that SG3524 produces 93% PWM when COMP is at around 3.5V, and 0% when it is below, say, 0.6V. So I have arranged the rest so that I can simulate that behaviour and get positive and negative output. Of course I have used any opamp I have found (TL082 can't be fed at +/-60V !!!) only to get the behavior.
The simulation works nice, you get the expected gain (RFB/R4) but there is always an offset that is only removed if I add a capacitor in series with the feedback resistor, that I don't like (it will produce DC droop, error, etc).
There must be a way to DC couple the feedback signal in unipolar-fed error amplifiers. Any ideas? |
|
|
| lumanauw |
Today I've bought 2pcs of IR2111 and 2pcs SG3524. I'm ready to build the full bridge one. Ohhh...wait a minute. I don't have the schematic yet :D
Is it possible to build the full bridge one with just 1 SG3524? |
|
|
| lumanauw |
| I also have the drum ferrite core, ready for the output filter.:D |
|
|
| Pierre |
Yes, it can be possible to build a full-bridge with a single SG3524, but you will need an additional opamp for the differential feedback, and it can become much more difficult.
I would recommend that you started with a half-bridge and then connect two of them in bridge mode.
I posted an sch that is being commented right now, when it is more clear and definitive you can start experimenting.
About the IR2111 driver, be prepared to have quite high switching losses, due to its small drive capability that will lead to slow rise/fall times unless you use small or low-capacitance mosfets. IRF640N is a good start. The only advantage of that chip is simplicity.
Be prepared also for turn on/off transients due to the lack of SD pin.
The coil is a drum core, isn't it? I started also with one of them from Wilco, with good results, although a iron-powder toroid core is better.
It only rests to fix the input/feedback coupling to remove the offset. The rest is quite straightforward, at least for a prototype.
Kanwar, can we have a look at your sch also? |
|
|
| Pierre |
I have solved the offset issue. At least in the simulations, it works perfect. You simply have to add a resistor from - INPUT to VREF, with the same value as the feedback resistor (100K). This way, the feedback signal has an added 2.5V so the output is now centered at 0V (excluding opamp offset errors as usual).
Output DC offset can be finely adjusted by making that resistor adjustable. I attach an updated proposal of the complete Class-D amp.
Gain is 20V/V (26dB). With the components used in the simulation, freq. response at -3dB is from 14Hz to 34KHz, but the simulation doesn't include the output filter nor accurate modelling of the error amplifier so there results should be verified experimentally.
The compensation network may need to be adjusted, but that won't be difficult if NFB is taken before filter.
With the proposed Rt and Ct values, switching frequency is around 175KHz (using a SG3524). |
|
|
| phase_accurate |
I don't think that this one will work ! The diodes at the SG... control outputs are connected the wrong way around (and maybe a resistor between the base of Q1 and +12 V would be a good idea).
Otherwise it is a simple and sexy solution.
Regards
Charles |
|
|
| Pierre |
Charles.
The sch is for a SG3525 that has totem-pole outputs.
IMHO the diodes are ok; this is how I see it: Each totem-pole output stage of the SG3525 will drive 12V when active, and pull to 0V when inactive, so the diodes are there to avoid that any transistor pulls down to 0V, blanking the output at any time (as two outputs are never active simultaneously).
BUT: you are right in that it won't work: a pull-down resistor is needed from base of Q1 to GND.
Is this ok? (I may be wrong, I am writing and drawing as the ideas come to my mind, so plese help debuggin it) :D |
|
|
| phase_accurate |
If you take the emitters then it is right that way. If you had used the collectors then the diodes would have to be turned and a pull-UP resistor would have to be used (and the drive inverted).
Regards
Charles |
|
|
| Pierre |
Then it is ok as I have drawn it, right? (with the pull down, of course)
I attach a drawing of the internal output section and how it is wired. This way, PWM is not inverted. So it attacks the level shifter, which inverts and hence must go to the LS driver.
But I am still not sure about how it should be done with a SG3524 for not-inverted PWM output ¿¿¿??? |
|
|
| phase_accurate |
Is the partial schematic shown from a 3524 ? If yes you can ground the middle of both totem poles and tie the collectors of the upper transistors together. Said collectors would then be connected to the base of the voltage-translator transistor and a pull-up resistor to +12V should also be in place. But phase would be inverted in this case which can easily be solved by changeing the inputs of the mosfet driver IC.
Regards
Charles
Edit: didn't notice that the collectors are already connected together ! |
|
|
| Pierre |
No, Charles. That's for a SG3525.
SG3524 has "uncommitted" transistors (only one per output as opposed to the totem-pole output of the SG3525).
Anyway, I think that this is almost solved. This afternoon I will try to connect a couple of diodes to a SG3525 test board I made some time ago, and let's see how the PWM looks like.
As for the NFB network: I think it should be almost ok this way, provided that it is taken before filter, what do you think, Charles?
It may work up to 150-200KHz, that should be ok for decent full range quality.
Thanks |
|
|
| lumanauw |
| quote: | | I would recommend that you started with a half-bridge and then connect two of them in bridge mode. |
Yes, I can do that, if the full bridge mode is too difficult for real amp. Chris also said that making full bridge is difficult, making 2 halfbridge in bridge mode is OK for me, as long as the rail bus pumping will not existed.
| quote: | | About the IR2111 driver, be prepared to have quite high switching losses, due to its small drive capability that will lead to slow rise/fall times unless you use small or low-capacitance mosfets. IRF640N is a good start. The only advantage of that chip is simplicity. |
If the full bridge (or 2 bridge half ones) can be done, I will do it with only +/-30V rail. That is the same with +/-60V half bridge, isn't it?
I planned to use IRF540 to keep the cost down. It is not OK to use IRF540?
| quote: | | The coil is a drum core, isn't it? I started also with one of them from Wilco, with good results, although a iron-powder toroid core is better. |
How about air core inductor? Is it OK to use this too? Like the ones for passive Xover? |
|
|
| Pierre |
In a 2 x half-bridge in bridge mode you won't have bus-pumping.
In fact, if you use +/-30V rails, provided that you won't get 100% modulation index with the SG3524/5, you will get around 54Vp-p or double of that (108Vpp) in bridge mode. With a 8ohm load, that's about 180W. But you can go to a higher voltage, say, +/-40V, and you can get around 325W into 8 ohm.
Of course, you can go to 4 ohm, but in that case each stage "sees" 2 ohm, so for +/-30V you will get a peak current of about 15A, and a total power in bridge mode of around 360W/4ohm (say, 450W with +/-40V rails).
The IRF540 (better, IRF540N) mosfets seem ok for your application as they have not very high gate charge, low Rds(on) and high dV/dt rating.
About the inductor: the problems with air core inductors are: higher conductivity losses (you need a lot of turns and hence high resistivity for a given inductance), and radiated EMI. You should use a drum core or, better, a toroidal one.
Best regards,
Pierre |
|
|
| Pierre |
After doing some testing with SG3524/25, I have found them to be a bit too slow in terms of operating frequency, minimum dead-time and rise/fall times.
I think that a great improvement would be to use UC3825 chip, that is very similar in operation, but much faster. Even the error amplifier is faster (6-12MHz vs. 1MHz GPBW, 6V/us vs 0.5v/us SR). This would allow full-range operation with very good quality in my opinion.
Cost is higher, of course, but not so much ;-)
BTW: Please have a look at this reference: it details some alternatives for the high side driver, all of them discrete. The low side one is straightforward, so if we could do it without the IR/HIP/National chips, that would be nice.
The third alternative seems better, although it may need some tweaking:
http://www.innovatia.com/Design_Cen...e%20Drivers.htm |
|
|
| ledmania |
| quote: | Originally posted by Pierre
Charles.
The sch is for a SG3525 that has totem-pole outputs.
IMHO the diodes are ok; this is how I see it: Each totem-pole output stage of the SG3525 will drive 12V when active, and pull to 0V when inactive, so the diodes are there to avoid that any transistor pulls down to 0V, blanking the output at any time (as two outputs are never active simultaneously).
BUT: you are right in that it won't work: a pull-down resistor is needed from base of Q1 to GND.
Is this ok? (I may be wrong, I am writing and drawing as the ideas come to my mind, so plese help debuggin it) :D |
Hi Pierre,
I"ve been toying this kind of circuit for about 7 months now using sg3525 mounted on solderless breadboard and it sounds good with nice bass and treble when driven by my son"s Sony discman via a coupling cap on pin 9 directly, but when I use its own comparators input pin 1 or 2 given with correct bias network, the bass and treble disapear. the reason is just because of the compensating .01uf cap connected to comparators o/p pin 9(as per sugested by chip maker). Removing this cap creates erratic oscillation in the o/p in which my scope cant follow the waveform.
BTW the 2 diodes in your o/p (pin 14 and 11) is correct followed by a pulldown resistor. My prototype uses 220ohms 2watts on that resistor. the purpose of this 2 diodes is to "source" during logic 1 of either o/p then the 220ohms resistor will handle the "sink" because diodes dont have sink capability:rolleyes:
BTW, I use my big Pioneer head phone in listening b'coz I dont have any power Fet lying around yet.:angel:
cheers
ledmania |
|
|
| ledmania |
| quote: | Originally posted by Pierre
After doing some testing with SG3524/25, I have found them to be a bit too slow in terms of operating frequency, minimum dead-time and rise/fall times.
I think that a great improvement would be to use UC3825 chip, that is very similar in operation, but much faster. Even the error amplifier is faster (6-12MHz vs. 1MHz GPBW, 6V/us vs 0.5v/us SR). This would allow full-range operation with very good quality in my opinion.
Cost is higher, of course, but not so much ;-)
htm[/url] |
Hi Pierre,
You are correct, sg3525 is slow on its rise/fall times but comparing its response time to other discrete devices, this chip is still more robust than some discretes counter part. I can say this b'coz of its interesting features inside the package like soft start which avoids turn on "thump" during start up, very stable/predictable on its target frequency(no FM modulation), very cheap and readily available. You are right about this UC3825. this is more faster and is
equipt with over current protector and can be powered up either on line or off line. I am looking forward in using this chip soon if I have enough time.
regards |
|
|
| lumanauw |
Hi, Ledmania,
So, the idea really WORKS......That's good.
How do you do it? Any schematics maybe? |
|
|
| Pierre |
Ledmania,
Thanks for the comments. About the pull-down resistors after the diode: that's true, in fact I realized and I think I corrected that in this forum.
About attacking directly the comparator via the COMP pin: well, then the error-opamp is before and it can do something to the signal, because you can't isolate it. And coupling the signal directly, you can't implement feedback, right?
Please confirm this, but I wouldn't go without feedback in any case, at least before filter.
I will do some testing with UC3825 the next days, I will tell you how it goes.
Lumanauw: it MUST work, it can cost some time and experimentation, but there is nothing that points in the opposite direction. Another issue is what the final performance will be, but I am optimistic about that, and if you need hi-fi, UC3825 should do perfectly.
Best regards,
Pierre |
|
|
| ledmania |
| quote: | Originally posted by lumanauw
Hi, Ledmania,
So, the idea really WORKS......That's good.
How do you do it? Any schematics maybe? |
| quote: | Originally posted by Pierre
Ledmania,
Thanks for the comments. About the pull-down resistors after the diode: that's true, in fact I realized and I think I corrected that in this forum.
About attacking directly the comparator via the COMP pin: well, then the error-opamp is before and it can do something to the signal, because you can't isolate it. And coupling the signal directly, you can't implement feedback, right?
Please confirm this, but I wouldn't go without feedback in any case, at least before filter.
I will do some testing with UC3825 the next days, I will tell you how it goes.
Lumanauw: it MUST work, it can cost some time and experimentation, but there is nothing that points in the opposite direction. Another issue is what the final performance will be, but I am optimistic about that, and if you need hi-fi, UC3825 should do perfectly.
Best regards,
Pierre |
Hello lumanauw and Pierre,
Yes, I have the schematic that I made since last year. I think It is much beter for you (lumanauw) to download the PDF file of SG3525 so that you can easily digest what I'm going to discuss.
Ready? get set?, here we go....Based on my drawing you will notice that ther are two coupling caps inputs on either pin 9 and pin 1. This means that you can trigger this pins with any audio source that you have(mine is sony discman). Based on my initial listening experience in using pin 1(inverting input) was not satisfactry accepted b'coz of the loss of bass and treble I got from the output. But when I totaly disregard the comparator concept and and use pin 9 via cap as the input, I was totaly surprised about the sound that I got form the output. I mean it sounds more solid than my discman. the bass and treble is much well pronounce against my discman. As I've told you on my previous post that the primary cause of this discrepancy on using this on board comparator is due to that 0.01uf cap on pin 9 as per suggested by the chip maker. Removing this will simply produce erratic oscillation in which my scope cant follow.
regards..
 |
|
|
| ledmania |
| quote: | Originally posted by Pierre
Ledmania,
Thanks for the comments. About the pull-down resistors after the diode: that's true, in fact I realized and I think I corrected that in this forum.
About attacking directly the comparator via the COMP pin: well, then the error-opamp is before and it can do something to the signal, because you can't isolate it. And coupling the signal directly, you can't implement feedback, right?
Please confirm this, but I wouldn't go without feedback in any case, at least before filter.
I will do some testing with UC3825 the next days, I will tell you how it goes.
Lumanauw: it MUST work, it can cost some time and experimentation, but there is nothing that points in the opposite direction. Another issue is what the final performance will be, but I am optimistic about that, and if you need hi-fi, UC3825 should do perfectly.
Best regards,
Pierre |
Hi Pierre,
Sorry, I forgot to mention something about your question regarding confirming the use of feedback on this circuit. My answer is "no" I never practice adding Any feedback on all of my class D prototype b'coz I'm pretty much enjoying in using PWM controller chips that include RC oscillator "built in".(I have already tested 5 chips:cool: )
The high frequency/carrier feedback that you are reffering to is only intended for "discrete self oscillating" class D just to sustain oscillation and or to stabilized it. Ditching this feedback on this type of circuit will simply stop its oscillation. However, there is a situation in which an amplifier needs a feedback but in our circuit(sg3525 et. al) you can put it if you like but in this case it must be connected after the filter(audio side). In this manner, you are controlling the gain to reduce overdriving by the source. This is slightly tricky b'coz you have to detrmine if it is inverting or non inverting.
Hope it helps
Cheers:D |
|
|
| Pierre |
Ledmania.
Let me clariry the concept of NFB (negative feedback) a little bit...
(Negative) Feedback is used to sample a portion of the output and compare it to the input so any deviation (well, not any, but to some extent determined by the loop gain, bw, etc), can be corrected by generating an "error" signal that is the real input to the PWM amplifier.
Clock based Class-D amps can implement feedback (it's not something exclusive of self-oscillating ones, for god's shake!!!), and in fact they do. I have tested about 8 different variations of PWM amplifiers and all of them use feedback. Some of them before filter, others after it. The fact is that they sound much better that in open loop. (In fact I wouldn't even think on building an open loop PWM amplifier, specially with a PWM chip!!)
Lumanauw: Feedback can (I would say MUST) be implemented with a chip like SG3525. |
|
|
| lumanauw |
Hi, Ledmania,
Thanks for showing your arrangement of this idea. Nice. :D
Is it for headphone amp without global feedback? You don't use pi-filter at output. Is this OK for the headphone, not ruining them?
Do you have idea how to make your idea for +/-30V and with additional power mosfets (IRF540)? I think about using half bridge driver IR2111 too. |
|
|
| ledmania |
Pierre,
I scrutinized your PDF schematic and I am worried that the high frequency/carrier feedback that you implemented "before the filter" may jeopardize the audio quality of the comparator input. I am not sure here but my intuition dictates that the correct location of your feedback should be on the audio side,that is "after the filter",then remove the parallel RC on that fb and put only R that is calculated based on convetional opamp gain equation. You can use SG3825 on this apps as well and I bet that it will sound more beter than SG3525 b'coz sg3525's pin 9 were bypassed with 0.01uf caps that's really makes me nuts.
Dont feel any hesitation on your endeavor in using those PWM chips that were mentioned because they are really more robust than some discrete counterpart. Trust me, they really sound good:cool:
cheers
ledmania |
|
|
| Pierre |
The NFB signal taken before the filter is integrated in the error amp.
It can be taken after the filter, but the LC filter introduces a 180º phase that can make the amplifier oscillate if not properly compensated, and that's not easy. I have done it with the aid of some people in this forum but, if hi-fi is not needed for the sake of simplicity I would take it before the filter.
Ledmania, I recommend that you take some time and read some of the threads in this Class-D forum, you will find them very informative and educational. You will see how the vast majority of topologies implement feedback and I am sure you will have things much more clear after this little effort.
Best regards,
Pierre |
|
|
| ledmania |
| quote: | Originally posted by Pierre
Ledmania.
Let me clariry the concept of NFB (negative feedback) a little bit...
(Negative) Feedback is used to sample a portion of the output and compare it to the input so any deviation (well, not any, but to some extent determined by the loop gain, bw, etc), can be corrected by generating an "error" signal that is the real input to the PWM amplifier.
Clock based Class-D amps can implement feedback (it's not something exclusive of self-oscillating ones, for god's shake!!!), and in fact they do. I have tested about 8 different variations of PWM amplifiers and all of them use feedback. Some of them before filter, others after it. The fact is that they sound much better that in open loop. (In fact I wouldn't even think on building an open loop PWM amplifier, specially with a PWM chip!!)
Lumanauw: Feedback can (I would say MUST) be implemented with a chip like SG3525. |
Hi Pierre,
Im so sorry for offending you on my last post but I have no intetion of opposing your idea of implementing fb:) . As I have said on my last paragraph that you can put it as you like but I suggest it would be more practical to put it after the filter.:) I'd been contemplating for quite a while now in digging the advantage of using high frequency fb but my mathematical capability is not enough to cope with that concept so why to bother on that route while there is a more simpler solution that my IQ can grasp? When I'm designing any class d amp, I am just treating it as a convetional linear amp in implementing fb. I use the old concept of either DC fb or AC fb but on the audio spectrum only. ;)
As in the sg3525 I admit that I did not put any fb nor any filter on it
b'coz it is only a prototype. Remember that a "working" proto with a incomplete components but produce nice sound will double its nice factor when you complete the required components.
Cheers
ledmania |
|
|
| Pierre |
No offence at all!!
From your last sentences, I think you have things a bit more clear than you think. I was not thinking on concepts such as "high frequency" feedback, but feedback in the audio band, as you say.
In fact, feedback is usually implemented so the crossover frequency (the one where loop gain is below 0dB or 1v/v) is less than half the switching frequency. The goal is to provide feedback within the audio band, where it is useful for reducing distortion.
And yes, it is better in terms of linearity to take feedback after the filter, but more difficult and dangerous in terms of estability. Usually, taking it before filter is enough. As it is integrated in the error amplifier, the high frequency content is attenuated.
The process of designing feedback in Class-D is about the same as in any other type of amplifier, with the exceptions that you have to account for the LC filter phase shift if you take after it.
But there are some other things to solve before imlementing feedback. The amplifier must work as good as possible in open-loop, and then the way to go is trying to optimize its performance with NFB. There is no point in trying to "save" a system that is very bad performing in open-loop.
Please let's go on with the original discussion (trying to establish a starting point of Class-D design basing on PWM controller chip and keeping it as simple as possible while still performing good) |
|
|
| ledmania |
| quote: | Originally posted by Pierre
But there are some other things to solve before imlementing feedback. The amplifier must work as good as possible in open-loop, and then the way to go is trying to optimize its performance with NFB. There is no point in trying to "save" a system that is very bad performing in open-loop.
[/B] |
Perfect point!!!
This is the main reason why I am permanently nailed on using those chips ever since I tried and tested it for more than two years now. Since those chips are cheap in prices, some of its many advantages over their discrete counter part are; higly stable, built in soft start(no "thump" sound during start up), built in high quality comparator, programable oscillator up to 1mhz are common.
Mind you, I abuse mine with lots of dangling wires and jumpers everywhere but it still insist to produce nice sound with good waveform on my oscilloscope. Actually, I tested 5 chips already and majority of them are working fairly identical in sound but your finall decision will reflect on choosing additional features like over current protection and etc. etc..
One interesting point I want to emphasize on this thread is that the use of PWM controller chips which was spicificaly "intended" by the chip maker for DC to DC converter applications is now playing your favorite music inside a small aluminum box is a"very"rewarding experience that a hobbyist may ever encounter :D
| quote: | Originally posted by Pierre
Please let's go on with the original discussion (trying to establish a starting point of Class-D design basing on PWM controller chip and keeping it as simple as possible while still performing good)
[/B] |
I'll obay!:D
cheers |
|
|
| Pierre |
Ledmania.
I think that features like overcurrent protection, soft start, etc, in these chips are useless (and even harmful) when using them with mosfets to create relatively high output powers (more than 1W or so :D ). The reason is that you want the chip to be oscillating at 25% duty cycle (each output) in order to have a symmetric square bipolar output and hence 0V after the filter. If you turn the PWM off by means of the overcurrent protection, shutdown or similar, you will get 0% duty cycle in each output. Remember that you join them in order to have a single PWM signal, that then you must invert to create the LS and HS drives. Then one of the mosfets will be permanently off while the other is permanently on (ok, if the "on" one is the highside it will be turned off after a while due to bootstrap cap discharge with no oscillation). That's something you really don't like!
So: in order to build a useful Class-D amp out of these chips, you need to be sure that soft-start, overcurrent, shutdown, overvoltage, undervoltage or whatever protections or features it have, don't engage, unless you have the power stage disabled, or you will get a large DC at the output, burning your speakers in the worst case.
The only function of this chip you need is the PWM modulator and error amplifier (although ledmania doesn't like this one). So choosing a chip basing on this features is an error. The only parameters that should make a decision are speed, linearity, maximum duty cycle, and output configuration.
Best regards. |
|
|
| Workhorse |
Hi Pierre,
Why dont you use LM3524 from National semiconductor which is very much faster than SG3524 and it is a very low cost solution also....
Since the motive behind the project is simplicity and low-cost and implementing it using LM3524 + IR2110 in Sub-woofer application is a very ergonomic[Human friendly] way.....
regards,
K a n w a r |
|
|
| Pierre |
From the datasheet, I see that LM3524 is just as slow as SG3524 (typ 200/100ns rise/fall times). But it does have an advantage: its output transistors are rated to 60V, so perhaps they can do the level shifting with no additional transistors...
I am sure that any of these PWM controllers will do OK for subwoofer use, what I was trying to see if it can be improved to full-range use with hi-fi performance by choosing a different (unfortunately more expensive) PWM chip.
Best regards,
Pierre |
|
|
| Workhorse |
| quote: | Originally posted by Pierre
From the datasheet, I see that LM3524 is just as slow as SG3524 (typ 200/100ns rise/fall times). But it does have an advantage: its output transistors are rated to 60V, so perhaps they can do the level shifting with no additional transistors...
I am sure that any of these PWM controllers will do OK for subwoofer use, what I was trying to see if it can be improved to full-range use with hi-fi performance by choosing a different (unfortunately more expensive) PWM chip.
Best regards,
Pierre |
Hi Pierre,
Yes ofcourse, if you want something fast than try this combination....
erroramp=OPA627
very high speed Differential comparator = LM360
Precision Triangle generated through combination of Crystal Oscillator using CD4060 counter and high speed integerator using LM6172 opamp.....
Low Gate Charge Mosfets from APT= APT30M18LVR ~Tgc=50nC
regards,
K a n w a r |
|
|
| ledmania |
Pierre,
you are mistaking in disregarding the important features of the chip. In fact the soft start features is well advisable in all audio apps to eliminate the turn on "thump" in which most of diyaudio members are complaining about. When you try to start the chip with soft start capacitor in placed, and your oscilloscope probe is connected to the output, you will notice that the signal will gradually start from 0%duty then stop on 50% at slow incremental rate. I'd been designing linear amps for 27 years now but the oldest technic that we put to aviod thump is to use relay. :D Been there done that,all I can say is SS is as important as current protector.
As with the over current protection, most high end amp is equipt with this features. Take note that not all amps are "idiot proof". this means in one way or another it may encounter accidental output short circuit like in bar b-Q party with the speaker cables are lying around :D .
Its nice to hear that you will try those chip to use as class d amps. I'll tell you...You will like their sound. But when you use them and disregard their SS and over current features then you are five steps forward but moved to six step backward. :D
BTW, the 25% duty cycle that you metioned is applicable only to the chip that is "double ended push pull" converter that we are talking about. Each two o/p should be 25% each, then mixed them with the added two diodes plus pulldown resistor to convert them into "single ended topology".This way, those 25%will become 50%.
As I mentioned to you on my first post that I already tried at least 5 chips since last 2 years and two of them are for "single ended topology" that eliminates the two diodes and resistor all together. the o/p can be adjusted from 0 to 98% and they sound nice too.
Hey Pierre, are you stil awake?:D
thanks for not yawning:D |
|
|
| phase_accurate |
On ledmania's AC coupled (!!!) headphone amp the soft start feature is indeed performing soft-start.
With a DC coupled beefy amp it would cause a real "hard-start" and has to be disabled therefore.
Regards
Charles |
|
|
| Pierre |
Let's try to clarify the confusion
.
I fully agree in that soft-start, overcurrent, etc in an amplifier are mandatory, I was only saying that the on-chip features are not suitable for a double-ended topology with push-pull, where 25+25% duty-cycle is required and hence a variation from 0% will produce a negative output progressively going up to 0% where the circuit regulates properly at 25+25%=50%.
From what I can see in your schematics, you are using ONLY one of the outputs of the SG3524 (the collector of the upper output, pin 12 is tied to GND, so you are deactivating that output and using only the lower transistor, you are really wasting a diode).
That setup won't work when trying to implement a power stage for it, as your duty cycle only varies from 0 to 45-50%, and you really need it to vary symmetrically (0 to 100% or 10 to 90% for example) in order to implement a half bridge and DC couple the output as usual.
I don't know if you understand me, Ledmania. I am not saying that the circuit doesn't work for you, but we are trying to buid a High power class-D so the mosfet stage is needed and your circuit can't be used as it is for that.
If the way you combine your double outputs is different, then what I say doesn't apply. If you could show us |
|
|
| ledmania |
| quote: | Originally posted by Pierre
Let's try to clarify the confusion
From what I can see in your schematics, you are using ONLY one of the outputs of the SG3524 (the collector of the upper output, pin 12 is tied to GND, so you are deactivating that output and using only the lower transistor, you are really wasting a diode).
That setup won't work when trying to implement a power stage for it, as your duty cycle only varies from 0 to 45-50%, and you really need it to vary symmetrically (0 to 100% or 10 to 90% for example) in order to implement a half bridge and DC couple the output as usual. |
No pierre, maybe you are looking at the wrong drawing. Take a look at it once more and you'll see that I used sg3525 double ended totempole output but not sg3524 with open emiter/collector output. as I said earlier that this sg3525 was tested by myself by loading it with a headphone.
I just show to you and to lumanauw that this is one of many chips that is possible to use as primary stage class d source.
Of coarse you have to provide additional amplification and levelshifting on this part like the use of IR2011 as the driver for power fets.I did not include the HI/LO side fet driver assembly because I know that majority of the readers of this thread already know that circuit configuration
| quote: | Originally posted by Pierre
I don't know if you understand me, Ledmania. I am not saying that the circuit doesn't work for you, but we are trying to buid a High power class-D so the mosfet stage is needed and your circuit can't be used as it is for that. |
My answer is same as above...
Regards |
|
|
| Pierre |
Sorry, Ledmania. I though it was an SG3524.
OK, if your sch. shows an SG3525, the outputs are push-pull and then they are connected OK (wired OR with diodes and pulldown).
But what I was trying to point out is that it works for you with shuwdown, slow start and all that stuff because you are AC coupling the output directly: at starts, output increases from 0 to 50%. At 50% (that you have pre-adjusted), you have VCC/2=6V at the output, but as you have a coupling cap, no problem. So the effect will be your coupling cap slowly charging and hence no turn on "pop". (without that slow start, you would get the same effect as with old Class-AB amplifiers that used unipolar supply, that kind of nasty "thump!")
So we agree so far...
...BUT in the right moment that you use this chip in conjunction with a power stage, that it is all DC coupled: steady state corresponds to 50% duty cycle, that produces 0V at the output. 0% means -VCC (-50V or whatever) at the output, and 100% means +VCC (+50V or whatever). SO if you use slow start, your output will start at -50V and then will rise to the final point of 0V (assuming no input signal). And that's NASTY, believe me!
The same applies with overcurrent: you can implement a circuit at the power stage that detects overcurrent, but it should shutdown the power stage (IR211X chip or both of its inputs), but NEVER the PWM chip, as it will go to 0% duty cycle which equals -VCC at the output, burning your speaker!
Conclusion: PWM chips CAN be used for Class-D half bridge designs, but several questions must be had in mind, such as WHERE to implement slow-start and overcurrent, as well as feedback. We will try to build a high-power one based on SG3525 (or UC3825 for better fidelity)
I hope eveything is now clear from my part and also from yours.
Best regards,
Pierre |
|
|
| ledmania |
| quote: | Originally posted by Pierre
Sorry, Ledmania. I though it was an SG3524.
OK, if your sch. shows an SG3525, the outputs are push-pull and then they are connected OK (wired OR with diodes and pulldown).
But what I was trying to point out is that it works for you with shuwdown, slow start and all that stuff because you are AC coupling the output directly: at starts, output increases from 0 to 50%. At 50% (that you have pre-adjusted), you have VCC/2=6V at the output, but as you have a coupling cap, no problem. So the effect will be your coupling cap slowly charging and hence no turn on "pop". (without that slow start, you would get the same effect as with old Class-AB amplifiers that used unipolar supply, that kind of nasty "thump!")
So we agree so far... |
It's nice that you see the correct sch. Take note that my sch was intended only to verify the validity of the chip to use as class d source thats why I put coupling caps on the o/p to isolate the dc supply through the head phone as the load.This is my old technic in verifying such audio circuit:breadboard+scope+headphone.I dont have cuircuit simulator yet.But in real world of application the coupling cap must be remove then the output of the chip must be split into two via inverter (just same as your PDF sch).
| quote: | Originally posted by Pierre
...BUT in the right moment that you use this chip in conjunction with a power stage, that it is all DC coupled: steady state corresponds to 50% duty cycle, that produces 0V at the output. 0% means -VCC (-50V or whatever) at the output, and 100% means +VCC (+50V or whatever). SO if you use slow start, your output will start at -50V and then will rise to the final point of 0V (assuming no input signal). And that's NASTY, believe me! |
The topic is becoming more interesting.:)
I understant your point Pierre, and I think I have a solution for that initial logic "0" at 0% duty. You are correct that the speaker may encounter instant "pop" even the ss is at work.
First, Lets use your PDF sch because it is the most typical to most readers and is available now.
Since you use inverter at the output of the "OR" diodes to split the signal in a complementary fashion for hi/lo side of fet driver, My solution is to change that inverter with inverter also but with "TRI state" feature. Remember that IR2xxx drivers has a safety faeture of always at logic "0" on both o/p at start up unless a signal is comutated on its input. This logic "0" on both output of IR2XXX will corespond to both "off" on o/p of the power fets.
Take note that when the o/p of the two power fets are "off" it means they are at "TRI state" also and not negative. You will measure at the o/p a very small leakage current comming from the driver because of the isolation barrier on the driver and not the power fets itself.
Now, this control pin of the tri state inverter must be connected to the o/p of UC3XXX with proper RC filter to accept only plain DC.
BTW,the inverter must be "schmit inverter" so that it will respond in a sharp "knee" of hyterises to the RC filter input.
I hope you got my point
| quote: | Originally posted by Pierre
The same applies with overcurrent: you can implement a circuit at the power stage that detects overcurrent, but it should shutdown the power stage (IR211X chip or both of its inputs), but NEVER the PWM chip, as it will go to 0% duty cycle which equals -VCC at the output, burning your speaker! |
I have a solution on this issue also but lets finish the SS issue first
If there are someone on this forum who will respond on my idea above then this will be a very healthy and informative discussion.
| quote: | Originally posted by Pierre
Conclusion: PWM chips CAN be used for Class-D half bridge designs, but several questions must be had in mind, such as WHERE to implement slow-start and overcurrent, as well as feedback. We will try to build a high-power one based on SG3525 (or UC3825 for better fidelity)
I hope eveything is now clear from my part and also from yours.
Best regards,
Pierre |
I think this is the only begining of a healthy and informative discussion regarding.............
""PWM chips CAN be used for Class-D half bridge designs, but several questions must be had in mind, such as WHERE to implement slow-start and overcurrent, as well as feedback. We will try to build a high-power one based on SG3525 (or UC3825 for better fidelity)""
CHEERS
ledmania. |
|
|
| Pierre |
For slow start, I have also though on using tri-state inverters followed by pull-downs. Once you have a control pin (the OE of that chip), you can put a RC network, so at startup the voltage at that control pin is initially 12V and it decreases until it reaches the threshold and activates the OE and hence both inputs to the IR211x at the same time.
That input can also be used for other protections. |
|
|
| ledmania |
| quote: | Originally posted by Pierre
For slow start, I have also though on using tri-state inverters followed by pull-downs. Once you have a control pin (the OE of that chip), you can put a RC network, so at startup the voltage at that control pin is initially 12V and it decreases until it reaches the threshold and activates the OE and hence both inputs to the IR211x at the same time.
That input can also be used for other protections. |
Thanks for favoring the idea.:)
At this type of function, the over current protection on the chip can also share the use of this tri-state features.
But you know Pierre, when I was eating my lunch a while ago, there is somthing that bothers my mind. This is base on the SS that we want to implement. I have noticed that if the tri-state was used, and the SS cycle has just completed maybe it will return again to its original "thump" sound because of the output is now 1/2VCC again:whazzat:
maybe the overcurrent will only benefit from this but not the SS.
But try and try until we succeed.
Cheers |
|
|
| Pierre |
Sorry, I don't understand your estatement.
We will use bipolar supplies. Assuming that the chip slow-start is fast enough, it will reach 50% duty cycle (combined output of both outputs of the chip), that corresponds to 0V output. _ After_ that, the inputs to the IR211x chip are enabled (by means of the tri-state control), and it starts oscillating. I think that won't produce noise... |
|
|
| phase_accurate |
I think the soft-start can be overridden anyway.
Regards
Charles |
|
|
| ledmania |
| quote: | Originally posted by phase_accurate
I think the soft-start can be overridden anyway.
Regards
Charles |
LOL!!!:D I want to think of that too but taxing my brain to solve this puzzle will surely make me more jubilant if we hit the bulls eye..
Cheers
ledmania |
|
|
| Workhorse |
Hi Pierre & LumanauW,
Here's the schematic attached for Lo-Fi Class-D amp....
regards,
K a n w a r |
|
|
| Pierre |
Some comments for Kanwar.
Your schematics is for only a section of a full-bridge amplifier, isn't it? If not, your output must be AC coupled, or there will be a +HRail/2 DC component at the output. (and AC coupling at the output is something we really don't like in high power amplifiers)
You don't use feedback as well, that would be very nice and I think it won't be difficult to achieve.
Best regards,
Pierre |
|
|
| lumanauw |
Hi, Kanwar,
Thanks alot :D
I will study it. Like Pierre said, how to connect the speaker output? Is it needing 2 of them to become full bridge? |
|
|
| Workhorse |
| quote: | Originally posted by Pierre
Some comments for Kanwar.
Your schematics is for only a section of a full-bridge amplifier, isn't it? If not, your output must be AC coupled, or there will be a +HRail/2 DC component at the output. (and AC coupling at the output is something we really don't like in high power amplifiers)
You don't use feedback as well, that would be very nice and I think it won't be difficult to achieve.
Best regards,
Pierre |
The output is DC coupled ..... not AC coupled because this is meant for H-Bridge topology not a Half bridge .....
Due to BD mode modulation , the feedback is not much necessary when the prime target is low frequency amplifier ... | quote: | Originally posted by lumanauw
Hi, Kanwar,
Thanks alot :D
I will study it. Like Pierre said, how to connect the speaker output? Is it needing 2 of them to become full bridge? |
2 units are required for bridging ...
For 2 Bridged Units the Oscillator must be Synchronised i.e. CT pin of 2 IC's must be tied together and shunted to GND by the Capacitor chosen for desired frequency...Master IC's pin Rt is connected to resistor to GND...but Slave IC's pin Rt must be tied to VREF...for proper Sync......
regards,
K a n w a r |
|
|
| Workhorse |
| quote: | Originally posted by lumanauw
Hi, Kanwar,
Thanks alot :D
I will study it. |
David_ The LumanauW,
What are your findings after the study of my schematics.......
K a n w a r |
|
|
| lumanauw |
Hi, Kanwar,
I haven't build it yet. I still thinking about how to use symmetrical rails (+/-35VDC), make it full bridge, and to put feedback on it. I still don't know it is better feedback before or after LC filter?
You have any idea how to implement those? |
|
|
| Workhorse |
Hi David,
In my opinion Single supply Full Bridge Class-BD mode[dont use Class-AD] without feedback would be adequate as long as your aim is for sub woofer application amp
K a n w a r |
|
|
| Workhorse |
Hi David,
Have you made your class-D amp,...any news on current projects...
K a n w a r |
|
|
| lumanauw |
Hi, Kanwar,
I haven't made it yet. I get stuck in how to make it full bridge amp. I've got IR2111, but have no idea how to make full bridge with SG3524. |
|
|
| Workhorse |
| quote: | Originally posted by lumanauw
Hi, Kanwar,
I haven't made it yet. I get stuck in how to make it full bridge amp. I've got IR2111, but have no idea how to make full bridge with SG3524. |
Its very easy ...connect another CMOS inverter to get the inverted signal from the output of SG3524 to drive another IR2111, or use 2 seperate halfbridges to form a fullbridge....and you have 3 Level PWM as free gift...
K a n w a r |
|
|
| Tekko |
Bump!
I´ll be testing this lofi d-amp thing before i give up completely. I´ll make it halfbridge tho, but i know a solution called split supply with the speaker gnd referenced to the centertap. If it doesent work i may try full bridge, but with one change. Two SG2524 should not be needed for fullbridge, just use an inverter to invert the signal to the other gate driver.
I think it should work. What do you think ?
Heres a schem with my idea: |
|
|
| Tekko |
| Trying attachment again as it dident appear to show up. |
|
|
| Workhorse |
Hi Tekko,
The schematic is right..you could now take another step towards prototyping it as well..
regards,
K a n w a r |
|
|
| Tekko |
Ok. I hope your right. Thoo bad that inverters dont come as single devices, but just six @ at time.:xeye:
Maybe a resistor and transistor approach would work, i could try it.:smash:
I want a simulator that contain these ic´s, i wanna be sure that it works before i order the parts.:smash: |
|
|
| classd4sure |
Hi Tekko,
As far as simulators go these days I'd go for LTspice, it is at least bug free as far as I can see, (in terms of stability) and while it doesn't include many models it's not that hard to add them to it.
This may help you in your research in fact, you can research what component is best for the job that suits both your needs (cost and availability) and those of your circuit, and then add the model statement.
I've used several simulators and while I find analysis features somewhat limited on LTspice, certain avide users of it like analogspiceman have demonstrated it's also a matter of creativity.
It's also extremely easy to get the hang of compared to the others I've tried, and alot of people seem to be leaning that way, so the odds of getting help with it when required are pretty good.
Also, since I am the UCD salesman around here, if you want to build a really cool circuit you should take a close look at the DIY ucd projects around here. There are no IC's required, just your average small signal transistors.. and some nice mosfets which are free samples. There's an extremely good amount of information to be learnt in that, cost you pennies to make, and in the end it's bound to sound amazing. |
|
|
|
