Hi DJK & Chris,
I have just succeded in all N-channel Mosfet 2 Tier Class-H amp..with good sonics in HF range....
It uses optocoupled gate drivers to switch the Tier Mosfets ON/OFF..This makes the design extremely simple but Opto drivers are somewhat costly....as they are from HPCL with very fast switching capability upto 100kHZ....I have sensed the voltage at the input of amp via opamp and this inturn drives the LEDs of optodrivers connected to the output of opamp in inverse parallel way..very easy setup....Thus minimises the delay and refines the Envelope...
I have to design a 4KW class-H 2000W/channel... Due to the design nature in Class-H amps they require somewhat less number of output devices as compared to Class-AB amps of same wattage......My rails are split into 2 tiers +150V..+75V..-75V...-150V.....Therefore the reactive loading is much safer in Class-H as compared to Class-AB..but I need an advise about the SOA consideration in Class-H amps..as they use less output devices...what criteria should be followed.....
best regards,
K a n wa r
I have just succeded in all N-channel Mosfet 2 Tier Class-H amp..with good sonics in HF range....
It uses optocoupled gate drivers to switch the Tier Mosfets ON/OFF..This makes the design extremely simple but Opto drivers are somewhat costly....as they are from HPCL with very fast switching capability upto 100kHZ....I have sensed the voltage at the input of amp via opamp and this inturn drives the LEDs of optodrivers connected to the output of opamp in inverse parallel way..very easy setup....Thus minimises the delay and refines the Envelope...
I have to design a 4KW class-H 2000W/channel... Due to the design nature in Class-H amps they require somewhat less number of output devices as compared to Class-AB amps of same wattage......My rails are split into 2 tiers +150V..+75V..-75V...-150V.....Therefore the reactive loading is much safer in Class-H as compared to Class-AB..but I need an advise about the SOA consideration in Class-H amps..as they use less output devices...what criteria should be followed.....
best regards,
K a n wa r
Hi Kanwar,
Design your output stage as a normal class A-B amp with the highest peak collector emitter voltages you expect to see. This will be lower than a normal amp, but higher than the low rails. Plan for commutator lock up. I might look at your low rail plus your high rail for a peak voltage and add some safety. The case temperature will be much lower in practice.
I fully expect fewer devices in the output stage. But go somewhere in the middle between minimum and normal amplifier numbers for higher reliability.
The Carvers used a very effective protection / shutdown system. Powering the amp up with a shorted output would normally not cause too much more damage, if any. I repaired many with one blown output (they got a full set to maintain reliability).
-Chris
Design your output stage as a normal class A-B amp with the highest peak collector emitter voltages you expect to see. This will be lower than a normal amp, but higher than the low rails. Plan for commutator lock up. I might look at your low rail plus your high rail for a peak voltage and add some safety. The case temperature will be much lower in practice.
I fully expect fewer devices in the output stage. But go somewhere in the middle between minimum and normal amplifier numbers for higher reliability.
The Carvers used a very effective protection / shutdown system. Powering the amp up with a shorted output would normally not cause too much more damage, if any. I repaired many with one blown output (they got a full set to maintain reliability).
-Chris
Hi Chris,
Thanks for the suggestions,
In my case the Mosfets would see...150+75=225VDC..therefore minimum of 300V devices are to be used....
QSC uses 6 parallel devices [2SC5200/2SA1943]with +-144 Hi rails for its 4000W amp...
I will use APT30M85BVR..300V 40A 300W probably about 4 in parallel, because there SOA is 3 times of Toshiba Bipolars.....
regards,
K a n wa r
Thanks for the suggestions,
In my case the Mosfets would see...150+75=225VDC..therefore minimum of 300V devices are to be used....
QSC uses 6 parallel devices [2SC5200/2SA1943]with +-144 Hi rails for its 4000W amp...
I will use APT30M85BVR..300V 40A 300W probably about 4 in parallel, because there SOA is 3 times of Toshiba Bipolars.....
regards,
K a n wa r
Hi CHRIS,
The current sharing wouldn't be problem as the mosfets are well matched and employ source degeneration resistors.....regarding voltage rating i would change the device with 400V devices...or maybe 600V igbt's as well....
BTW...I have concluded that with real world music signals these Class-H amps were very much efficient than Class-AB ones...
regards,
K a n w a r
The current sharing wouldn't be problem as the mosfets are well matched and employ source degeneration resistors.....regarding voltage rating i would change the device with 400V devices...or maybe 600V igbt's as well....
BTW...I have concluded that with real world music signals these Class-H amps were very much efficient than Class-AB ones...
regards,
K a n w a r
Hi Kanwar,
It depends on how exactly you trigger commutation. On Carver's, there was a minimum on time. High frequency signals had a much shorter period, so a high level signal at high frequency would trip the commutators on and keep them there. Oscillation would do this on any design I would think, feedback could be a problem.
Build it and test for this in particular.
-Chris
It depends on how exactly you trigger commutation. On Carver's, there was a minimum on time. High frequency signals had a much shorter period, so a high level signal at high frequency would trip the commutators on and keep them there. Oscillation would do this on any design I would think, feedback could be a problem.
Build it and test for this in particular.
-Chris
Hi Kanwar,
When stuff goes wrong at high power, it goes really, very wrong. Quickly too.
-Chris
Not by itself. What happens is that the ouput device dissipation goes up to the level that it would be as a simple class A-B amp. This is very hard on your SOA. You may end up with failed ouputs if the amp is under load at the time, depending on the load of course. At high frequencies I would expect the tweeters to open, crossover may fail caps short. Generally ugly.
When stuff goes wrong at high power, it goes really, very wrong. Quickly too.
-Chris
Hi Chris,
very good info...Then what should be done to overcome this situation of rail lockup at HF......
meanwhile I came across QSC MX1500 Class-G amp...which features Parallel connected upper/lower Tier Devices...rather than series connected...I think this parallel topology of Class-G works better than series connected devices...transition noise could be very well eliminated.........
please do see this link........
http://www.qscaudio.com/support/library/schems/mx1500.pdf
regards,
K a n w a r
very good info...Then what should be done to overcome this situation of rail lockup at HF......
meanwhile I came across QSC MX1500 Class-G amp...which features Parallel connected upper/lower Tier Devices...rather than series connected...I think this parallel topology of Class-G works better than series connected devices...transition noise could be very well eliminated.........
please do see this link........
http://www.qscaudio.com/support/library/schems/mx1500.pdf
regards,
K a n w a r
Hi Kanwar,
The envelope follower type of switching did not introduce switching noise. False problem. Only the earlier Carver types that switched the entire supply on at once did that.
Whether the devices are in parallel or series will not affect the outcome. If you switch on and off with a stepped waveform, expect transients in the output.
The Lightstar model may work even better for you. It used high frequency switching mosfets to reduce the +-125 VDC rails down to 10 VDC or so above the musical waveform. It was just a tracking regulator fed by the music signal. PWM determined the DC voltage, filtered by an LC network.
To keep the rails from locking up was a constant problem. They go better at it over time, but it could still happen. Carver did use an HF detect circuit to trigger the protection. That's about all you can do I think.
-Chris
The envelope follower type of switching did not introduce switching noise. False problem. Only the earlier Carver types that switched the entire supply on at once did that.
Whether the devices are in parallel or series will not affect the outcome. If you switch on and off with a stepped waveform, expect transients in the output.
The Lightstar model may work even better for you. It used high frequency switching mosfets to reduce the +-125 VDC rails down to 10 VDC or so above the musical waveform. It was just a tracking regulator fed by the music signal. PWM determined the DC voltage, filtered by an LC network.
To keep the rails from locking up was a constant problem. They go better at it over time, but it could still happen. Carver did use an HF detect circuit to trigger the protection. That's about all you can do I think.
-Chris
The original QSC MX series used the paralleled outputs on different rails, later they went to the rail switch.
That's why the MX2000 brings more money on the used market than the MX2000A.
The Crest method for modulated rail sounds better, but reguires the front end to do rail-to-rail vs the QSC only needing to swing +/- a few volts with their grounded output stage.
That's why the MX2000 brings more money on the used market than the MX2000A.
The Crest method for modulated rail sounds better, but reguires the front end to do rail-to-rail vs the QSC only needing to swing +/- a few volts with their grounded output stage.
anatech said:
Hi Chris,
Therefore, the envelope follower as in Class-G mx1500 is a good outcome...and hence it sounds better than mx1500a which is Class-H ...with step switched waveform and hence transients......thus sounds bad at HF.......therefore Class-G is more linear than Class-H....is it right....
Now I understand , whenever I place mobile phone ringing or severe HF clipping than the competitor's Class-H amp, it goes in to protection mode just to save the devices from rail locking........
but this protection mode could be activated during "Shows & events" and thus turned out to be more catastrophic....
regards,
K a n w a r
Hi Kanwar,
Yes, an envelope follower will sound better in theory. Execution will can make a difference too.
For protection, we are now talking about degree. If you ever have full output at high frequencies, the high frequency driver(s) just failed. That would suck more don't you think?
So when you experiment, you can set frequency, level and duration limits for the protection circuit. Don't forget, the amp will come back out of protection. The tweeters will not come back from the dead without a recone (diaphram) kit.
-Chris
Yes, an envelope follower will sound better in theory. Execution will can make a difference too.
For protection, we are now talking about degree. If you ever have full output at high frequencies, the high frequency driver(s) just failed. That would suck more don't you think?
So when you experiment, you can set frequency, level and duration limits for the protection circuit. Don't forget, the amp will come back out of protection. The tweeters will not come back from the dead without a recone (diaphram) kit.
-Chris
Hi Chris,
Does Rail locking also exsit in Class-G amps, as they modulate the rail, but dont switch them ON/OFF at once.....unlike Class-H amps.......
Here is the Link to Carver's FTP site....
ftp://208.187.38.55/Carver
regards,
K a n w a r
Does Rail locking also exsit in Class-G amps, as they modulate the rail, but dont switch them ON/OFF at once.....unlike Class-H amps.......
Here is the Link to Carver's FTP site....
ftp://208.187.38.55/Carver
regards,
K a n w a r
anatech said:
Hi Chris,
Majority of Pro-amps used Stepped waveforms for rail switching these days....To suppress transients these employ snubbers accross commutation diodes and output devices also......
regarding Rail lok problem..I think the decision must be taken before the waveform advances the output....or just sense the transition threshold from the input rather than output...Secondly loss of Saturation of output device must be detected to prevent further rail lok...
regards,
K a n w a r
