It is basically class G with a switched mode current source shunting the outer device in each rail.
Since the outer devices are the ones that dissipate the power in a class G amp (They get the voltage drop when there is real current in play) this is all to the good.
One interesting saving here is in the power transformer, and DC rail caps.
Because the output stage rail voltage is limited essentially by the GAN switches, diodes and inductor ratings, it might be reasonable to build an amp using very high rails with little smoothing (Hence relatively large conduction angles on the rectifiers and relatively good PF) with the power being set by limiting the drive to the amplifier, let the switched mode current source deal with the excess voltage. A 150V rail with 50V or so of ripple on it driving an amp that can only swing 80V pk would save one hell of a lot of DC smoothing cap, and actually has advantages in that the reduced ratio of voltages across the inductor means that the current can slew faster when the amp is near maximum output V=L.dI/dt and all that (Which is a potential weakness of this design, current ramp up rate is inversely proportional to load voltage).
In the subwoofer case, for an amp built right into the box you could conceivably do NON isolated power stage run bridged off the 380 odd volts directly from the PFC caps. I would hate to have to fault find the lethal thing, but it would be insane power for nearly no weight.
Since the outer devices are the ones that dissipate the power in a class G amp (They get the voltage drop when there is real current in play) this is all to the good.
One interesting saving here is in the power transformer, and DC rail caps.
Because the output stage rail voltage is limited essentially by the GAN switches, diodes and inductor ratings, it might be reasonable to build an amp using very high rails with little smoothing (Hence relatively large conduction angles on the rectifiers and relatively good PF) with the power being set by limiting the drive to the amplifier, let the switched mode current source deal with the excess voltage. A 150V rail with 50V or so of ripple on it driving an amp that can only swing 80V pk would save one hell of a lot of DC smoothing cap, and actually has advantages in that the reduced ratio of voltages across the inductor means that the current can slew faster when the amp is near maximum output V=L.dI/dt and all that (Which is a potential weakness of this design, current ramp up rate is inversely proportional to load voltage).
In the subwoofer case, for an amp built right into the box you could conceivably do NON isolated power stage run bridged off the 380 odd volts directly from the PFC caps. I would hate to have to fault find the lethal thing, but it would be insane power for nearly no weight.
I'm rather partial to the NAD 3240PE design, whereby the outer rails are switched in when the the output signal approaches the inner rails, rather than actually modulating the inner rail. It means the transistor used to source current in to the inner rail doesn't have to dissipate appreciable power, on account of being mostly on or off.
The inner drivers then are able to function as a completely normal power amp.
The subtlety is in ensuring the rail switch devices do so fast enough not to get in the way during fast transients, but smoothly so as not to inject currents to the output quicker than the error amp can compensate for.
You're completely out of luck for square wave response for outputs that are higher than the inner rail. There's just no way that switch is going to respond in time not to mangle the leading edge.
Plus of course the myriad of interesting fault currents that blow things up, and ensure I have a never ending supply of cheap dead 3240's to play with.
The inner drivers then are able to function as a completely normal power amp.
The subtlety is in ensuring the rail switch devices do so fast enough not to get in the way during fast transients, but smoothly so as not to inject currents to the output quicker than the error amp can compensate for.
You're completely out of luck for square wave response for outputs that are higher than the inner rail. There's just no way that switch is going to respond in time not to mangle the leading edge.
Plus of course the myriad of interesting fault currents that blow things up, and ensure I have a never ending supply of cheap dead 3240's to play with.
Or just insane?
I have 3 phase power so I don't actually need DC caps at all.
A 6 phase transformer array should work, I have no experience of this kind of industrial power tech but I think it looks possible.
However, I would like to keep the amp power supply essentially standard so other DIY members can use it easily (not that there's much interest outside a select few, so far).
So I tentatively plan to use a Hypex SMPS that looks pretty well matched to the job.
I don't care about heavy, certainly not a reason to risk a non isolated supply.
I must admit the idea had occurred to me for cost reasons but I'm not prepared to risk it just for that.
Nor for more power, 600 W is overkill in a domestic system anyway.
Maybe if I was asked to build a multi kW tour system where heavy mattered.
Link or schematic?
Yes, hard to do nice square waves.
The ASC I posted actually can handle square waves, the outer transistors simply act as conventional class H until the switch mode stuff catches up.
Which is quite nice, efficiency deteriorates but that hardly matters because it should rarely or never happen in reality.
Even less so because I plan to use the switch mode Class H for the subs and woofers and plain class H for the mids.
I have played around with both GaN and Si FETs for the switcher.
The GaN use less power, even when I tried Si FET models with 1 m.ohm drain-source resistance and 1 pF capacitances.
Not sure why yet, it's all due to slower turn off however fast the driver.
Anyone understand this or know the physics?
Best wishes
David
Last edited:
My point was that I wouldn't even need capacitors, the transformer phases overlap so it's not needed.
Probably use a small amount of capacitance for EMI but still save quite a bit of money because nice fat capacitors are not cheap.
Unfortunately, unlike semiconductors that continue to fall in price, capacitors just seem to increase.
But the 6 phase idea is not as simple as I expected, probably not worth the trouble.
Best wishes
David
Probably use a small amount of capacitance for EMI but still save quite a bit of money because nice fat capacitors are not cheap.
Unfortunately, unlike semiconductors that continue to fall in price, capacitors just seem to increase.
But the 6 phase idea is not as simple as I expected, probably not worth the trouble.
Best wishes
David
Last edited:
I flip flop on GaN or Si FETs.
The availability of complementary silicon simplifies life.
Whereas GaN Systems have a neat and tidy form factor.
Wish there was some equivalent silicon pack- it's low inductance, easy to layout, decent thermal performance.
And GaN has some nice characteristics, SOA is quite decent.
But no avalanche breakdown number to compare because " GaN E-HEMTs do not avalanche and thus" are not rated.
Hmm, so what's the equivalent comparison?
David
Have you ever considered the benefits of verbs?
Possibly in sentences?
Best wishes
David
We did this for work. The supply was +/-6.5V, about 30A, from a 415V three phase source. We used Δ and Y windings on the transformer to essentially get six phases.
While ripple was significantly lower than a normal single-phase supply, it was still there due to mismatches between the phases. We still used capacitors (brutally expensive wet tantalums to cope with the heat) but they were only a few hundred μf.
The transformer was extremely expensive, and the wiring was way over the top. I wouldn’t do it again if I had my druthers.
The Δ and Y connection was what I had in mind but after I looked a bit closer I did decide "probably not worth the trouble".
Thanks for the story of your experience to confirm that.
3 phase may still happen but not 6.
Not really my main focus here any more, I now take the power supply as a done deal.
I did post some new power supply circuit ideas earlier in the thread Very Efficient Very Low Distortion Linear+ Amp but there wasn't much interest, too hard for most people I suppose.
It was too hard for me
Well, at the same time as the Class H switcher anyway, maybe later.Best wishes
David
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.