Hi DIYers -IR has recently released a very interesting new class-d audio half-bridge driver IC, the IRS20954 (data sheet). It is fabricated with their latest process, and has programmable dead-time and under-voltage and over-current protection available for both MOSFETs (latching or auto-recovering with adjustable hiccough mode delay), but perhaps most interesting of all is that its input is referenced to ground rather than the negative rail.
No more level shifting! Just front it with a low noise comparator and the drive train is done. 
... in theory, anyway.
PS: Not sure if these parts are actually available yet (single piece price seems to be about five US dollars).
Hmmm... if they only had an LTspice model.
analogspiceman: Thanks for bringing this chip about. It can make the amplifier slightly simper, because you don't need the PNP transistor for level conversion. A part from that, it is quite similar to earlier IR drivers.
It still doesn't have 'ultralow distortion' but might be ok for less critical applications.
It still doesn't have 'ultralow distortion' but might be ok for less critical applications.
Lars Clausen said:
Are you sure (about the distortion part)? All of the classical linear class-d parts would be outside of the IR driver IC - and it is capable of minimal dead-time and has seemingly constant delay. What more could be required?
Maybe I'm missing something, but it would be interesting to get some feedback from a knowledgeable someone who has recent hands-on experience with one of these latest generation ICs from IR.
Regards -- analogspiceman
fokker said:
Of course it has internal level shifters - that's the point - it has them built-in so that the non-engineer DIYer doesn't have to deal with designing them him or herself. This IC, although not quite the Holy Grail, 
is a big step forward in user convenience 
(sort of like what happened with switcher ICs for laptops).Regards -- analogspiceman
Lars Clausen said:
I suppose you are thinking of your fav open loop designs with that 0.02% remark, but the dead-time granularity certainly seems undesirable (don't know why IR didn't opt for a continuous analog adjustment scheme of DT). One could always fine tune the dead-time elsewhere (but if it's no longer a straightforward cookbook design, then it kind of defeats the point of simplicity, doesn't it).
I'd like to see a solution with parametric "on-the-fly" adjustment of dead-time via some sort of feedback or feedforward.
Regards -- analogspiceman
ideally it should be current controlled but that would be difficult to implement.
alternatively, the turn on of any mosfet can be conditioned on the Vgs of the other mosfet. That can be done fairly easily.
Although it is not as effective, but it should be reasonably OK.
Has anyone done that?
Lars Clausen said:
10ns steps are bold steps?
....I would guess there are only few guys in the world who really can handle better than 10ns precision with power devices..
I would not dare to say that I would be able to handle single nano seconds with high power devices in mass production.
And in DIY it might work out to fine tune some single protos by the gate drive resistors. Or not?
Honestly, I would really love to have samples of this driver.
I am right now picking up my switch mode sub woofer amp project.
Sub woofer amps seem to be a good starting point for my first switch mode amp. As it does not need to handle full audio bandwidth, I can work with lower switching frequencies. This makes dead time & distorsion considerations much easier and is pushing the efficiency. .... of course low switching frequencies are bad for the size of the output filter.... I am planning something around 100kHz...
Chokoholic: I think you project sound very sensible, and interesting. The output filter will not be so big after all, at 100 kHz.
10nS are steps the size of King Kong's feet when it comes to dead time distortion. Not only can you only get so close to the optimal dead time compensation, but you can also risk being 10 nS on the wrong side of the thin line. Then you will get some serious cross conduction.
You should be able to tune dead time out, down to below 1 nS. And also importantly, you should keep the dead time compensation independent of signal, temperature and drain current. I have seen THD as low as 0.0005% (10W in 4 Ohms) even on an open loop system after dead time compensation by this technique. I suspect it can go even lower.
10nS are steps the size of King Kong's feet when it comes to dead time distortion. Not only can you only get so close to the optimal dead time compensation, but you can also risk being 10 nS on the wrong side of the thin line. Then you will get some serious cross conduction.
You should be able to tune dead time out, down to below 1 nS. And also importantly, you should keep the dead time compensation independent of signal, temperature and drain current. I have seen THD as low as 0.0005% (10W in 4 Ohms) even on an open loop system after dead time compensation by this technique. I suspect it can go even lower.
10ns = King Kong's feet? ...oohps.
I feel that we are living in different worlds.
Also I feel that my home equipment USB scope might be a limiting factor... It has just 100Ms/s.
I am not experienced in fast switch mode applications... just in slow switch mode stuff... and there I am used to much worse situation.
Just imagine these cheap BJTs which increase their storage time by 400ns when heating up from RT to 110C... OK MOSFETs are less annoying, but also there the treshold voltage is drifting with temperature and with this the resulting real deadtime will vary..
Hints from your side, how to reduce the effect of the temperatures are quite welcome , also dependency on current .... '10ns=King Kong's feet' ... you are really threatening me..
For my project I will probably set up a new thread as soon as I have something on hand that is worth to be posted..
It might still need some time. I already prooved with my Rookie-thread that my slow project progress can stress the patience of the readers , and there is no chance for me to speed up.
I am thinking of a simple self oscillating structure with full bridge (+/-45V) in order to run two paralleled 4 Ohm sadhara drivers... ...should end up in something like 1kW....1.5kW at 2 Ohms. Hysteresis oscillator. Not UCD, because I am avoiding to build things that I do not fully understand.
I feel that we are living in different worlds.
Also I feel that my home equipment USB scope might be a limiting factor... It has just 100Ms/s.
I am not experienced in fast switch mode applications... just in slow switch mode stuff... and there I am used to much worse situation.
Just imagine these cheap BJTs which increase their storage time by 400ns when heating up from RT to 110C... OK MOSFETs are less annoying, but also there the treshold voltage is drifting with temperature and with this the resulting real deadtime will vary..
Hints from your side, how to reduce the effect of the temperatures are quite welcome , also dependency on current .... '10ns=King Kong's feet' ... you are really threatening me..
For my project I will probably set up a new thread as soon as I have something on hand that is worth to be posted..
It might still need some time. I already prooved with my Rookie-thread that my slow project progress can stress the patience of the readers , and there is no chance for me to speed up.
I am thinking of a simple self oscillating structure with full bridge (+/-45V) in order to run two paralleled 4 Ohm sadhara drivers... ...should end up in something like 1kW....1.5kW at 2 Ohms. Hysteresis oscillator. Not UCD, because I am avoiding to build things that I do not fully understand.
Hi Markus
Do you remember magnetic snubbers? They allow for no deadtime at all while controlling cross-conduction and reverse recovery di/dt. They produce an apparent dead time proportional to inductor current.
Note that diode reverse recovery times and currents also have their impact on apparent dead time.
BTW (off-topic): A working sample of my full-bridge switching amplifier project has been already produced and sent to the final customer for evaluation. It's powered direcly from rectified mains (350V DC) and it employs 2 magnetic snubbers and 12 TO-220 IGBTs. It has been tested up to 2.2KW or so for an hour (that's sine, so it's 4.4KW peak).
At 2.2KW, EMI is barely measurable with the oscilloscope at 1mV/div while looking into the signal picked up by a 10cm diameter loop antenna placed 1 meter away from the prototype. That's despite the absence of any casing or any common-mode EMI filtering in the input or the output leads, so it may be regarded as "quiet". I believe efficiency to be in the 95% range.
Do you remember magnetic snubbers? They allow for no deadtime at all while controlling cross-conduction and reverse recovery di/dt. They produce an apparent dead time proportional to inductor current.
Note that diode reverse recovery times and currents also have their impact on apparent dead time.
BTW (off-topic): A working sample of my full-bridge switching amplifier project has been already produced and sent to the final customer for evaluation. It's powered direcly from rectified mains (350V DC) and it employs 2 magnetic snubbers and 12 TO-220 IGBTs. It has been tested up to 2.2KW or so for an hour (that's sine, so it's 4.4KW peak).
At 2.2KW, EMI is barely measurable with the oscilloscope at 1mV/div while looking into the signal picked up by a 10cm diameter loop antenna placed 1 meter away from the prototype. That's despite the absence of any casing or any common-mode EMI filtering in the input or the output leads, so it may be regarded as "quiet". I believe efficiency to be in the 95% range.
Check that thread in order to see the basic approach for single quadrant switching:
http://www.diyaudio.com/forums/showthread.php?postid=609651#post609651
Other more complex (and patented) approaches for four-quadrant switching have been shown recently on the class D forum, but I can't find them now.
http://www.diyaudio.com/forums/showthread.php?postid=609651#post609651
Other more complex (and patented) approaches for four-quadrant switching have been shown recently on the class D forum, but I can't find them now.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.