I am just finishing the SMPS for a class T amplifier of 2x1500W with 2x1500W/4R output power. The power rails voltage is +-60V.
The problem is the stabilization of the negative rail.
Although it seems to be no problem because amplifiers are bridged, the current drawn from both rails would be the same, but I am afraid to put my TA3020s on it, since the maximum supply voltage for the chip is +-70V and an regulated negative rail would easily overshoot above that voltage.
Of course the positive rail stabilization works perfect.
It is a full-bridge SMPS with SG3525.
I had one idea to make some kind of switch, that would switch the feedback to SG3525. As soon as it flips so that the output of the transformer will be negative and the negative rail will begin to charge, it would switch the feedback from positive to negative rail so the IC will now see the negative rail status and will regulate it to the proper voltage. Then it would switch back the feedback signal to monitor the positive rail. I think it is kinda crazy but this is my only idea, (except the one to regulate the output voltage to 120V alltogether, but I think this is not a good idea because one rail can still overshoot).
All comments or other ideas of solution on this problem are very appredicated.
The problem is the stabilization of the negative rail.
Although it seems to be no problem because amplifiers are bridged, the current drawn from both rails would be the same, but I am afraid to put my TA3020s on it, since the maximum supply voltage for the chip is +-70V and an regulated negative rail would easily overshoot above that voltage.
Of course the positive rail stabilization works perfect.
It is a full-bridge SMPS with SG3525.
I had one idea to make some kind of switch, that would switch the feedback to SG3525. As soon as it flips so that the output of the transformer will be negative and the negative rail will begin to charge, it would switch the feedback from positive to negative rail so the IC will now see the negative rail status and will regulate it to the proper voltage. Then it would switch back the feedback signal to monitor the positive rail. I think it is kinda crazy but this is my only idea, (except the one to regulate the output voltage to 120V alltogether, but I think this is not a good idea because one rail can still overshoot).
All comments or other ideas of solution on this problem are very appredicated.
Since you are bridging the amps thus little current will be flowing in the 0V rail, can't you just make the secondary a single winding and derive the 0V rail from a rail splitting potential divider? Then you could just regulate the rail as a single entity.
In fact, couldn't you just do this anyway even with a centre-tapped or even dual secondary design, stabilising acros both rails rather than ground? As the midpoint of the windings will never really shift.
In fact, couldn't you just do this anyway even with a centre-tapped or even dual secondary design, stabilising acros both rails rather than ground? As the midpoint of the windings will never really shift.
the simple fix is to use a pair of + and - linear post-regulators -- this will also remove some of the switching noise -- you are beyond the limits of most off the shelf regulators so consider using a couple zeners to drive the base of a pass transistor, just adjust the duty cycle of the sg35XX driver to compensate for the loss you will experience in the pass transistor.
@jackinnj:
That is a very simple solution, but linearr stabilization at the SMPS output would rapidly decrease its overall efficiency and the maximum output power respectively. It would also generate a lot of heat, and I can not afford this in my chassis.
@richie00boy:
You are right that with bridged amplifier the 0V could never shift too much, but who knows... I just do not want any risk of damaging the processor with overvoltage.
I will definitely try the solution with splitting potential divider, because it is a great idea. But for this time, I am going to do it another way. I am going to use an extra optocoupler that will have 60V zener diode in series with the LED, which will be connected to -60V rail. The other end will drive the shutdown pin of SG3525... so when the -60V rail will be over approx. -61V, the control pwm will shutdown. But I think that your solution is simpler and more reliable.
Thank you very much for your interest in my problem.
Disney_SK
That is a very simple solution, but linearr stabilization at the SMPS output would rapidly decrease its overall efficiency and the maximum output power respectively. It would also generate a lot of heat, and I can not afford this in my chassis.
@richie00boy:
You are right that with bridged amplifier the 0V could never shift too much, but who knows... I just do not want any risk of damaging the processor with overvoltage.
I will definitely try the solution with splitting potential divider, because it is a great idea. But for this time, I am going to do it another way. I am going to use an extra optocoupler that will have 60V zener diode in series with the LED, which will be connected to -60V rail. The other end will drive the shutdown pin of SG3525... so when the -60V rail will be over approx. -61V, the control pwm will shutdown. But I think that your solution is simpler and more reliable.
Thank you very much for your interest in my problem.
Disney_SK
@sss:
Can you imagine an unregulated SMPS at this power level? Sorry but I can't. Also today when I was enjoying the sound of an amplifier, I found that my regulation is insufficient even with low power levels. It fallen from 49V to about 45V with 2W output, it is designed for 4kW output... so something is wrong.
Btw does anybody know why? I think that it is a problem with SG3525's error amplifier compensation. What capacity and resistance between -IN and CMPEN pins would you use?
Can you imagine an unregulated SMPS at this power level? Sorry but I can't. Also today when I was enjoying the sound of an amplifier, I found that my regulation is insufficient even with low power levels. It fallen from 49V to about 45V with 2W output, it is designed for 4kW output... so something is wrong.
Btw does anybody know why? I think that it is a problem with SG3525's error amplifier compensation. What capacity and resistance between -IN and CMPEN pins would you use?
Disney_SK said:
I really think that it is unwise to use "voltage control" in an SMPS delivering this much power. Current mode control allows you to control the current on the primary side of the transformer and output voltage on the secondary side.
The compensation scheme really requires a network analyzer to get it right -- and if you don't get it right you await disastrous instability. A network analyzer can be as simple as a signal generator and a two channel scope. Essentially you "break" the control loop of the error amplifier, inserting a transformer (with a load which matches the impedance of the analyzer, i.e. 50R) inject a swept signal into the break, measuring gain and phase. Too long for a paragraph here -- but there are great application notes on every chip manufacturer's website.
@jackinnj
Hmmm... looks too complicated for me.
First, I do not understand the current mode control. Do you sense the secondary current and adjust switching duty cycle to get higher primary current? I wound not understand this, because it would overshoot easily...
Yesterday I have measured the negative rail current, who was unstable. Today I have measured the positive rail current and that one is much stiffer... It holds 49V all the time. I was thinking that bridged amplifier would take same (absolute) positive and negative rail current.. seems it is not. Today I am going to measure zero rail current, perhaps it will tell me something
interesting.
In SG3525's schematic, there is only that simple voltage mode regulation that I use for the positive rail. Also did not found anything more about this chip on ST's site.
Hmmm... looks too complicated for me.
First, I do not understand the current mode control. Do you sense the secondary current and adjust switching duty cycle to get higher primary current? I wound not understand this, because it would overshoot easily...
Yesterday I have measured the negative rail current, who was unstable. Today I have measured the positive rail current and that one is much stiffer... It holds 49V all the time. I was thinking that bridged amplifier would take same (absolute) positive and negative rail current.. seems it is not. Today I am going to measure zero rail current, perhaps it will tell me something
interesting.
In SG3525's schematic, there is only that simple voltage mode regulation that I use for the positive rail. Also did not found anything more about this chip on ST's site.
Look at d-amp.com
I use an unregulated SMPS for a 4Kw amplifier with the SG3525, if your tranformer is thight, good switching mosfet with direct drive (no transformer drive) and a good output capacitor, you will get all the power and regulation you need. Try it! www.d-amp.com
I use an unregulated SMPS for a 4Kw amplifier with the SG3525, if your tranformer is thight, good switching mosfet with direct drive (no transformer drive) and a good output capacitor, you will get all the power and regulation you need. Try it! www.d-amp.com
My transformer is tight enough, but I use little output capacitance because of high switching frequency, so unregulated SMPS won't be the good choice here. I use IGBTs at primary and because of the regulation on the secondary side, I use pulse transformers to drive them. Also, this kind of class-T amplifier has gain very dependent on supply voltage, so unregulated SMPS could lead into unpleasant signal compression at high power...
I have been looking at your site, but hell I cant believe that you have 97,5% overall efficiency at full power. Also, how do you use SG3525 as a resonant supply controller, when it has fixed-frequency oscillator and variable duty cycle?? This it what I am really interested in.
Best regards
Matej
I have been looking at your site, but hell I cant believe that you have 97,5% overall efficiency at full power. Also, how do you use SG3525 as a resonant supply controller, when it has fixed-frequency oscillator and variable duty cycle?? This it what I am really interested in.
Best regards
Matej
Disney_SK, that's an interesing objective to build power supply for 3kW amplifier. I'm also working now on 3kW amplifier, single channel though. And, that will be unregulated power converter 🙂
Well, if you still want to have regulated SMPS, then you better put SG3525 on primary side and drive gates with special chips (like IR2110 or anything else). Then take output voltage sensing from both rails, like richie00boy suggests. There's completely no point to control rails separately in full-bridge amp.
Also, you have to keep output capacitors as large as your free space permits, no matter the swithing frequency - keep in mind those tens of ampers of current pulses flowing. Anyway some minimum for such power is 10,000 uF per rail.
Again, your +/-45V are not sufficient for 3kW/4ohm RMS, I think neither +/-60V, the right will be +/-70V - which is limit as you told. But maybe you will be sutisfied with 1200W 🙂
The voltage drop at light load may be caused by output DC inductor discontinuous current mode if regulation is designed not good. In other words, DC choke 'likes' to work at currents higher than minimum but lower than maximum accepted, both these parameters easily calculated.
Check out picture for voltage sensing that I've used many times. This type of stabilization is quite accurate, more than anyone needs for amp supply. There's a little temperature deviation, depending mostly on zener Tc.
Well, if you still want to have regulated SMPS, then you better put SG3525 on primary side and drive gates with special chips (like IR2110 or anything else). Then take output voltage sensing from both rails, like richie00boy suggests. There's completely no point to control rails separately in full-bridge amp.
Also, you have to keep output capacitors as large as your free space permits, no matter the swithing frequency - keep in mind those tens of ampers of current pulses flowing. Anyway some minimum for such power is 10,000 uF per rail.
Again, your +/-45V are not sufficient for 3kW/4ohm RMS, I think neither +/-60V, the right will be +/-70V - which is limit as you told. But maybe you will be sutisfied with 1200W 🙂
The voltage drop at light load may be caused by output DC inductor discontinuous current mode if regulation is designed not good. In other words, DC choke 'likes' to work at currents higher than minimum but lower than maximum accepted, both these parameters easily calculated.
Check out picture for voltage sensing that I've used many times. This type of stabilization is quite accurate, more than anyone needs for amp supply. There's a little temperature deviation, depending mostly on zener Tc.
Attachments
To fredos: I also reviewed your web-site. That's almost unbeleivable about "more than 97,5%" total efficiency for senior model - the only case that I've known of about 96% amp efficiency was 5kW D-class amplifier, supplied from external battery DC rails (!) and combined high-voltage DC converter and amp - all together all-in-one. Again, the best efficiency I've ever seen for AC mains SMPS was some 4kW unit with 96%. Now, taking 0.96x0.96=0.92 ....... 
Then why you need there 4 fans inside even in junior 1200W model, as total power losses would be laughable (1200/0.96)-1200=50 watt ? 😱 50W is not much for aluminium heatsink that unit looks to have.
And, can you please explain what drawback implies transformer drive for IGBT or FET gates.
I tried transformer drive for unregulated converter (half-bridge type) and didn't notice major problems with it.
Your explanations appreciated with no technology secrets giving away 😉
Then why you need there 4 fans inside even in junior 1200W model, as total power losses would be laughable (1200/0.96)-1200=50 watt ? 😱 50W is not much for aluminium heatsink that unit looks to have.
And, can you please explain what drawback implies transformer drive for IGBT or FET gates.
I tried transformer drive for unregulated converter (half-bridge type) and didn't notice major problems with it.
Your explanations appreciated with no technology secrets giving away 😉
I'm interested if anyone was able to get good sqare impulses on oscilloscope using SG3525 PWM regulation?
I just made regulated SMPS based on SG3525 chip. I used optocoupler + zeners and connected all this to pin 9. The output voltage is within 0.1 volt. But I have one problem: the signal at <50% duty cycle doesnt look like sqare wave. It's is rather like something I can't describe in words 🙂
I just made regulated SMPS based on SG3525 chip. I used optocoupler + zeners and connected all this to pin 9. The output voltage is within 0.1 volt. But I have one problem: the signal at <50% duty cycle doesnt look like sqare wave. It's is rather like something I can't describe in words 🙂
4 slash:
I have a very nice square-wave impulses, I have about an 10% overshoot at the output, and I am not using optocoupler, I use two impulse transformers. Although the transformer output is not very exactly following the input, I have still nice GTDRV signal and switching efficiency.
I have a very nice square-wave impulses, I have about an 10% overshoot at the output, and I am not using optocoupler, I use two impulse transformers. Although the transformer output is not very exactly following the input, I have still nice GTDRV signal and switching efficiency.
4 Alme
Thank you very much for your recommendations and schematic. But my SMPS is already done and working very good.
I know that the output capacitance should be as high as possible, but I have very limited space on the PCB :-( There is totally 2x3000uF filtering on the secondary. The PWM is reacting pretty fast, so I have very stiff and flat output voltage even with large powers and current transients.
I am not designing 3kW/4ohm amplifier, but two independent 1500W/4ohm channels. It is class T, so it can make the output voltage almost as high as supply voltage rails. So 2x60V = 120Vmax, that is 84,8Veff, so max power would be 1850W with THD of about 10%.
Do you also wonder about Fredos' amplifier? I think it's all VERY interesting 😉
Thank you very much for your recommendations and schematic. But my SMPS is already done and working very good.
I know that the output capacitance should be as high as possible, but I have very limited space on the PCB :-( There is totally 2x3000uF filtering on the secondary. The PWM is reacting pretty fast, so I have very stiff and flat output voltage even with large powers and current transients.
I am not designing 3kW/4ohm amplifier, but two independent 1500W/4ohm channels. It is class T, so it can make the output voltage almost as high as supply voltage rails. So 2x60V = 120Vmax, that is 84,8Veff, so max power would be 1850W with THD of about 10%.
Do you also wonder about Fredos' amplifier? I think it's all VERY interesting 😉
Slash: if you mean "optocoupler+zeners" is similar feedback type that I posted above, then you need to connect optocoupler output to error amplifier input. Connecting it to pin 9 (err amp output) will make the regulation quite poor, if it will operate at all. You can try 2 possible connections;
1) connect non-inv input (pin 2) to reference (pin 16); opto collector to main voltage (pin 15), opto emitter to inv input (pin 1); also resistor about 1k from pin 1 to common (pin 12);
2) connect inv input (pin 1) to reference (pin 16); opto emitter to common (pin 12), opto collector to non-inv input (pin 2) and resistor 1k between pin 2 and pin 15. All pin numbers referenced to SG3525.
1) connect non-inv input (pin 2) to reference (pin 16); opto collector to main voltage (pin 15), opto emitter to inv input (pin 1); also resistor about 1k from pin 1 to common (pin 12);
2) connect inv input (pin 1) to reference (pin 16); opto emitter to common (pin 12), opto collector to non-inv input (pin 2) and resistor 1k between pin 2 and pin 15. All pin numbers referenced to SG3525.
Have you all really tested your SMPS at full load for a reasonable time with a dummy load? [electric heaters are very useful]. This is a must before attempting to power any amplifier with it
Have you performed any transient response tests on your SMPS? This is also essential to optimize the frequency compensation of the regulation loop before attempting to power any amplifier with it. The thougest test is to apply full load pulses and check output voltage waveform behavior, a good SMPS will show little or no fluctuations, a bad SMPS may even blow
Also be careful with current control since it tends to go unstable at low currents or low duty cycles due to noise issues, usually requiring heavy frequency compensation that ruins all of its benefits on transient response
In order to prevent amplifier damge I recommend overvoltage protection on both rails by means of TL431 precision references acting on the shutdown pin of the control IC
Have you performed any transient response tests on your SMPS? This is also essential to optimize the frequency compensation of the regulation loop before attempting to power any amplifier with it. The thougest test is to apply full load pulses and check output voltage waveform behavior, a good SMPS will show little or no fluctuations, a bad SMPS may even blow
Also be careful with current control since it tends to go unstable at low currents or low duty cycles due to noise issues, usually requiring heavy frequency compensation that ruins all of its benefits on transient response
In order to prevent amplifier damge I recommend overvoltage protection on both rails by means of TL431 precision references acting on the shutdown pin of the control IC
Disney_SK: do you mean pulse transformers to decouple OUTPUT signals? I think Slash says about optocoupler feedback problem. Actually, using transformer for SMPS feedback is too complicated unless it is flyback type unit (SG3525-based is not the one). On the contrary, one needs very special optocouplers to transfer gate drive signals: the worst thing about usual optocouplers with phototransistor receiver is that they are terribly slow for this.
About your amplifier: do you really think that you can use full-rail output sinewave even if rails are perfectly stabilized under load? For all I know, switching amplifiers usually leave about 15-20% of rails "unused", having worse supply usage ratio than conventional amplifiers. I don't know about Tripath though. I tested simpler fixed- and variable-frequency types and found their maximum amplitude of about 40V with stabilized rails of 50V. This problem traces to difficulty of having pulse width close to its period in both types, especially in self-oscillated circuit.
About your amplifier: do you really think that you can use full-rail output sinewave even if rails are perfectly stabilized under load? For all I know, switching amplifiers usually leave about 15-20% of rails "unused", having worse supply usage ratio than conventional amplifiers. I don't know about Tripath though. I tested simpler fixed- and variable-frequency types and found their maximum amplitude of about 40V with stabilized rails of 50V. This problem traces to difficulty of having pulse width close to its period in both types, especially in self-oscillated circuit.
Alme, I know about this stuff about maximum output swing, ald Tripath datasheet also surprised me pretty much... They really say it can deliver 1850W/4ohm at 10% THD. I have also contacted their tech support about 8ohm maximum power at +-60V and they wrote that there is no problem to reach 1000W at 10% or less THD 🙂 . I recommend studying class T a little before writing things like you wrote. And I am sorry about that pin9 thing, I though that it is a switching output pin. Of course I do not use transformer for voltage feedback 😉 Mine is on secondary, so I am routing it only thru zener and resistor divider directly to -INP.
Eva, I have tested it only under real appliacation, powering both amps at full power, into clipping, and I do not know about any problems with regulation or filtering. And it sounds really great😉
Eva, I have tested it only under real appliacation, powering both amps at full power, into clipping, and I do not know about any problems with regulation or filtering. And it sounds really great😉
Alme,
I just checked the first variant you offered. The result I got was very unsuccessfull because there was a lot of noise at the output of the trafo. The noise is produced by immediate turn offs and turn ons of sg3525. I couldn't see the sqare waveforms on my oscilloscope at all. It was the noise. The noise from the trafo can be heard. Do you see the sqare waves on your oscilloscope?
The variant of regulation I offered satisfies me so far. I found that the problem with the sqare waves was in the inductance of the trafo. It can't be completely eliminated but I hope the efficiency will not suffer from it. I will test it soon. Changing the duty cycle I can vary the output voltage within 30% of total voltage. It is enough to make regulation and there is NO NOISE at the output. The transistors are all cold under 200w resistive load... The only one requirement in my variant is to use inductors after the diode bridge. Only in this case I am able to get perfect regulation within 0.1 volt from no load to maximum load.
I just checked the first variant you offered. The result I got was very unsuccessfull because there was a lot of noise at the output of the trafo. The noise is produced by immediate turn offs and turn ons of sg3525. I couldn't see the sqare waveforms on my oscilloscope at all. It was the noise. The noise from the trafo can be heard. Do you see the sqare waves on your oscilloscope?
The variant of regulation I offered satisfies me so far. I found that the problem with the sqare waves was in the inductance of the trafo. It can't be completely eliminated but I hope the efficiency will not suffer from it. I will test it soon. Changing the duty cycle I can vary the output voltage within 30% of total voltage. It is enough to make regulation and there is NO NOISE at the output. The transistors are all cold under 200w resistive load... The only one requirement in my variant is to use inductors after the diode bridge. Only in this case I am able to get perfect regulation within 0.1 volt from no load to maximum load.
- Status
- Not open for further replies.