The on resistance does increase with temperature, but the extremely low voltage high current FETs Bruno mentions are specified at significant currents already. They are being used extensively in low voltage processor power supplies outputting 100's of amps, so they definitely work!
And of course they can be paralleled if a few miiliohms isn't low enough. As well, a multiphase design could be explored at least for DIY (there are some patents here that need to be considered for production).
And of course they can be paralleled if a few miiliohms isn't low enough. As well, a multiphase design could be explored at least for DIY (there are some patents here that need to be considered for production).
Re: High current
0.1 ohms? Do not underestimate the kind of devices that power your PC. If your reservations were anything like true, PC development would have stopped dead at the 386.
http://www.irf.com/product-info/datasheets/data/irf6609.pdf
http://www.vishay.com/docs/73384/73384.pdf
http://www.semiconductors.philips.com/acrobat_download/datasheets/PH2520U-01.pdf
On resistance does increase with current, but when Ron becomes much higher than Ron at low currents, the MOSFET has effectively entered saturation (becomes a current source). Obviously any "Ron" value (as defined by Vds/Id) is possible once you have the device in its constant-current region.
The MOSFET type and gate drive voltage should be chosen such as to insure saturation current is well above operation current. When you stay within half of the saturation current, Rdson may be considered to be fairly constant. For the IR device, saturation current at Vgs=4V is specified at around 200A. Below 100A, Ron is already nicely linear at just over 2 milliohms. Extrapolating to Vgs=10V, Vdsat is around 800A, a highly academic value indeed.
In a full bridge, the mosfets account for 4 to 5 milliohms. If the load is 100 milliohms, on losses are 5%. Accounting for another 5% switching and inductor losses, 90% efficiency is by all means feasible.
CarlBenton said:
0.1 ohms? Do not underestimate the kind of devices that power your PC. If your reservations were anything like true, PC development would have stopped dead at the 386.
http://www.irf.com/product-info/datasheets/data/irf6609.pdf
http://www.vishay.com/docs/73384/73384.pdf
http://www.semiconductors.philips.com/acrobat_download/datasheets/PH2520U-01.pdf
On resistance does increase with current, but when Ron becomes much higher than Ron at low currents, the MOSFET has effectively entered saturation (becomes a current source). Obviously any "Ron" value (as defined by Vds/Id) is possible once you have the device in its constant-current region.
The MOSFET type and gate drive voltage should be chosen such as to insure saturation current is well above operation current. When you stay within half of the saturation current, Rdson may be considered to be fairly constant. For the IR device, saturation current at Vgs=4V is specified at around 200A. Below 100A, Ron is already nicely linear at just over 2 milliohms. Extrapolating to Vgs=10V, Vdsat is around 800A, a highly academic value indeed.
In a full bridge, the mosfets account for 4 to 5 milliohms. If the load is 100 milliohms, on losses are 5%. Accounting for another 5% switching and inductor losses, 90% efficiency is by all means feasible.
Wow, thanx for pointing that out Bruno. Those FETs you showed are quite good. Looking at the gate charges and assuming a drive current of 2A or so, the losses due to rise and fall times are quite insignificant for anything below 400kHz. Very good choice.
Most of the losses will be due to the Rdson, and the drive circuitry.
Although what freq will it be running at, and what rise/fall times are you aiming for? Do you think the rise and fall times are not a factor in this design. Larger Rise/Fall times reuire more drive current per cycle. Of course we all know this, but if we require large drive currents, this will impede on efficiency. Remember, efficiency includes all stages of the amp.
I would like to see 90% achieved here too.
Thankyou.
Most of the losses will be due to the Rdson, and the drive circuitry.
Although what freq will it be running at, and what rise/fall times are you aiming for? Do you think the rise and fall times are not a factor in this design. Larger Rise/Fall times reuire more drive current per cycle. Of course we all know this, but if we require large drive currents, this will impede on efficiency. Remember, efficiency includes all stages of the amp.
I would like to see 90% achieved here too.
Thankyou.
Generally I don't shoot for very fast rise times. My amps usually switch in the 20..50ns area. Actually switching the gates faster does not incur greater gate drive losses because the amount of charge transferred is the same. If you do it quickly, the peak current is higher but the spike is shorter too. Average gate driver supply current per fet = Qg*fsw, with switching speed not a factor.
Switching too fast does create EMI problems, which is why you need to compromise on gate switching speed. Proper gate drive design allows you to get optimal efficiency and EMI at the same time.
Switching too fast does create EMI problems, which is why you need to compromise on gate switching speed. Proper gate drive design allows you to get optimal efficiency and EMI at the same time.
The output inductor will probably have a lot more losses than the mosfets... it will probably have a higher resistance than the ribbon ! thus you might need a higher frequency to balance the losses between switches and inductor...
Also Ribbons, once you take the transformer out, are very efficient (100dB/W/1m ?) so, unless you listen primarily to ultrasonic program material, high power is a bit unnecessary. People have been doing tweeter amps with 1W triodes !
Also Ribbons, once you take the transformer out, are very efficient (100dB/W/1m ?) so, unless you listen primarily to ultrasonic program material, high power is a bit unnecessary. People have been doing tweeter amps with 1W triodes !
That ribbon efficiency will certainly be helpful.
I would imagine that the same developments that produced such low Ron FETs have given us some pretty low loss inductors as well. Certainly these would be made from wide foil and so forth and would not look much like the conventional class D toroids etc.
I would imagine that the same developments that produced such low Ron FETs have given us some pretty low loss inductors as well. Certainly these would be made from wide foil and so forth and would not look much like the conventional class D toroids etc.
I'm excited to see so much activity and hope to see some sample schematics and Spice simulations soon.
The ribbon is 100 db/watt efficient, but at 0.1 ohms, we still need 10 amps to reach 1 watt. I design my speakers for 110 db @ 1m max SPL, as I do most of my listening below 80 db @1m, about 74db at the couch. This is the magic of high efficiency ribbon linesources, you can get goosebump audio at kid safe SPLs.
My current amp is a DIY Class-A that uses a SuperSym folded cascode topology to drive 16 Sanken 180 watt bipolar output transistors. The output stage is a bridge with 4 NPN-PNP pairs on each symmetrical side. I am currently using a +/- 7V power supply for the output stage and +/- 12V for the driver circuits.
I am looking forward to a fully digital signal path: digital outs from the DVD/CD player, going over digital cables to each speaker where there is a digital Xover with room equalization and YOUR DIGITAL AMP.
The ribbon is 100 db/watt efficient, but at 0.1 ohms, we still need 10 amps to reach 1 watt. I design my speakers for 110 db @ 1m max SPL, as I do most of my listening below 80 db @1m, about 74db at the couch. This is the magic of high efficiency ribbon linesources, you can get goosebump audio at kid safe SPLs.
My current amp is a DIY Class-A that uses a SuperSym folded cascode topology to drive 16 Sanken 180 watt bipolar output transistors. The output stage is a bridge with 4 NPN-PNP pairs on each symmetrical side. I am currently using a +/- 7V power supply for the output stage and +/- 12V for the driver circuits.
I am looking forward to a fully digital signal path: digital outs from the DVD/CD player, going over digital cables to each speaker where there is a digital Xover with room equalization and YOUR DIGITAL AMP.
LineSource!
You have a very impressive speaker!
Unfortunately I don't see much developers of digital amps here. Most of us makes analog ClassD amp. If you want to build digital, then you must use an IC, eg. from Texas, but digital amps needs extremely fast and amasingly matched dead-time output devices to produce low distortion, wich are hard to find with such a low impedance. Analog designs are much less critical. Main designs are known, you just have to select proper devices, and design an extremely low inductance PCB.
I think this can be adapted and improved easily:
http://www.hszk.bme.hu/~sp215/elektro/PWMamp_2x3W.gif
You have a very impressive speaker!
Unfortunately I don't see much developers of digital amps here. Most of us makes analog ClassD amp. If you want to build digital, then you must use an IC, eg. from Texas, but digital amps needs extremely fast and amasingly matched dead-time output devices to produce low distortion, wich are hard to find with such a low impedance. Analog designs are much less critical. Main designs are known, you just have to select proper devices, and design an extremely low inductance PCB.
I think this can be adapted and improved easily:
http://www.hszk.bme.hu/~sp215/elektro/PWMamp_2x3W.gif
I think it is a nobel goal to try and keep digital music digital throughout the system. Generally it should be a lot easier to control all the interface between equipment. Although, having heard a report from a friend that even optical cables can alter the sound, the signal interface. buffering, and timing probably needs a more thorough investigation.
Looks like getting low enough self-noise on the power supply for satisfying subjective dynamics is going to be the biggest challenge here. Building on Bruno's earlier calculations for the capacitor size (around 0.5F) the ESR needs to scale down too alongside the capacitance scale-up. Given a typical pair of reservoir caps might have 10-20mohms ESR the 0.5F cap should have ESR below 0.5mohm. This level of ESR is achieved with some of the larger Maxwell ultracaps however the large capacitance needed to achieve this (into 1000s of Farads) is inconvenient in terms of switch-on time.
I think using an output trafo is a more practical option.
I think using an output trafo is a more practical option.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.