So what have changed since 2005 regarding drive circuit and mosfets?
Tripath was very aware of what to look at to get the mosfets perform at its best... So was zetex now owned by NAD ( there class D Technology )
So was/is B&O, TI and IRF (but who knows what happens after infineon takes over)
New packages arrives but it is only to make class D cheaper to high volume markets This drives the class D market... Not quality ... Except of course for B&O, lyngdorf, mbl, T+A to name a few... But they are low quantity and not exactly good business for the large semi fabs.
Sent from my iPhone using Tapatalk
Hypex arrived. Look at the list of high end manufacturers using their boards including
NAD
Theta
MBL
Can't be bothered to Xref more but there are plenty of others launching high end products on Bruno's designs. Never heard them myself as not currently in the market for that much power, but some of the people using it are well enough established not to launch with something that would affect their market.
And all the above review well, unlike the Harman offering in the levinson 53. Go figure.
NAD
Theta
MBL
Can't be bothered to Xref more but there are plenty of others launching high end products on Bruno's designs. Never heard them myself as not currently in the market for that much power, but some of the people using it are well enough established not to launch with something that would affect their market.
And all the above review well, unlike the Harman offering in the levinson 53. Go figure.
Tripath desigh dont have post filter feedback so its out right away and old B&O Icepower has alot problems there is MC33078 inside the self oscillation path that can't pass undistored signal at needed frequency this is inside the inside feedback, there is two feedback loops the outside loop drives the inside feedback loop to improve linearity by a bit because of badly designed output filter with big inductance parasitics on the output filter caps it can not run with one feedback like UCD that uses carrier residual as a pseudo triangle wave and it does not have ground planes and the layout is bad plus the whole design does not have linear dc transfer in theory.
If you got a chance you should test hypex nCore design but dont use AP passive filter it just adds noise (well mine does) active AES17 filter is enough i did use dScope with prism own passive filter and it worked as it should.
If you got a chance you should test hypex nCore design but dont use AP passive filter it just adds noise (well mine does) active AES17 filter is enough i did use dScope with prism own passive filter and it worked as it should.
Sorry to forget hypex... But his solution is discrete (no integrated parts)...
They are doing a good job ....
Sent from my iPhone using Tapatalk
and don't call me heterodyne
The Crown/Levinson reception is a puzzle, and I don't know enough about the internal details to speculate very much. But one of the great features of Stanley's patented back-to-back buck converters and modulation scheme is to render deadtime essentially a non-issue.
Bruno's UcD achievement is particularly interesting, as it was the result of trying to make the cheapest and lowest parts-count switchmode amp! Philips is still the patent holder and will be happy for you to use the topology as long as you use NXP components---or so I last heard. UcD does have some potential deadtime issues that are handled very well in the Hypex embodiments. Since they are variable-frequency designs the spectre of audible beat frequencies is there, although this is a sensitive subject to bring up with Hypex, who guarantee you will have no difficulties provided you buy the modules from them.
I looked at ways to make the UcD-like approach work for fixed frequencies and concluded it would be difficult to achieve similar performance.
NXP still makes among the best of the integrated class D amps, and I cannot understand why they didn't get UcD versions on a chip. Their control of deadtime is generally very effective, and also embodied in their properly-named one-bit D-A chips, used by some of the "digital" amp people.
The Crown/Levinson reception is a puzzle, and I don't know enough about the internal details to speculate very much. But one of the great features of Stanley's patented back-to-back buck converters and modulation scheme is to render deadtime essentially a non-issue.
Bruno's UcD achievement is particularly interesting, as it was the result of trying to make the cheapest and lowest parts-count switchmode amp! Philips is still the patent holder and will be happy for you to use the topology as long as you use NXP components---or so I last heard. UcD does have some potential deadtime issues that are handled very well in the Hypex embodiments. Since they are variable-frequency designs the spectre of audible beat frequencies is there, although this is a sensitive subject to bring up with Hypex, who guarantee you will have no difficulties provided you buy the modules from them.
I looked at ways to make the UcD-like approach work for fixed frequencies and concluded it would be difficult to achieve similar performance.
NXP still makes among the best of the integrated class D amps, and I cannot understand why they didn't get UcD versions on a chip. Their control of deadtime is generally very effective, and also embodied in their properly-named one-bit D-A chips, used by some of the "digital" amp people.
I love to hear an elaboration. Again, I don't know the details, having been ejected from Harman in 2004.
From the get go the decision was made to use fixed frequency balance current style design, now this was effective 20 years ago when FET's were not capable as they are now so it can be run without worry about dead time and high frequency switching capability, today there is no such problems the transistors that are avaiable now are fast enough that the you can fine tune the output stage to very low distortions and small idle currents.
The biggest advantage of ML design is that you can run the FET's at low switching rate but they desided to use the extra capability to extend the bandwidth loosing all the advantages and receive only the drawbacks 2MHz is useless if the audible band is distorted.
Higher switching frequency gonna cause more distortion and on top of that with bca topology you need to time the output stage delay that will be the source of more distortion, the use of big Inductors gonna cause the need for substandard board layout that gonna radiate EMI everywhere that is gonna cause the need to not use negative feedback at higher frequency (ML 53 loses loop gain after 2kHz) and it's gonna use truckload of parts and gonna cost a arm and a leg (They brag that ML 53 uses more than 1000 parts) and not sound that impresive.
Every on of BCA design issues can be addresed with simple full bridge self oscilating design it will use 100 times less components and be twice as good.
The biggest advantage of ML design is that you can run the FET's at low switching rate but they desided to use the extra capability to extend the bandwidth loosing all the advantages and receive only the drawbacks 2MHz is useless if the audible band is distorted.
Higher switching frequency gonna cause more distortion and on top of that with bca topology you need to time the output stage delay that will be the source of more distortion, the use of big Inductors gonna cause the need for substandard board layout that gonna radiate EMI everywhere that is gonna cause the need to not use negative feedback at higher frequency (ML 53 loses loop gain after 2kHz) and it's gonna use truckload of parts and gonna cost a arm and a leg (They brag that ML 53 uses more than 1000 parts) and not sound that impresive.
Every on of BCA design issues can be addresed with simple full bridge self oscilating design it will use 100 times less components and be twice as good.
I would think the Class D suffers the most with the real worlds dynamically wearying loads of dynamic loudspeakers, so what looks good on the resistive bench, suffers a lot when actually asked to deliver.
Hypex amps proudly boast a flat, load invariate frequency response....I don't think you can make these types of generalizations.
better FETs
Is the improvement in FETs mostly in the better recovery characteristics of the body diodes?
Hypex is probably the best out there. We use it in one of our designs, but just for tracking voltage regulators.
The 53 did some things quite well and a couple of things very wrong. I can't say much more but Harmon did get rid of Brad after all.
😀
Corporate freaked out when I disclosed I had a tax problem (now long solved). The people I was working with were uniformly appalled but were powerless to change the decision from above.
Another departure around that time was that of COO Gregg Stapleton, who clashed with the then-CTO, who refused to consider going direct with some automotive suppliers. Sid sided with Geiger, and Gregg said Well if as COO I can't save 20 million dollars almost instantly, I'm out of here (or words to that effect)
Ah well. I guess about everyone in audio has to be touched by that company in one way or another. The CEO has been credited with the stock going up.
And of course lower control charge for a given Ron.
In a short-lived startup, I inherited a medium-power amp design based on an IR reference design with a little massaging from another engineer, including the incorporation of an extra L-C output filter section. The IR driver chip was located very close to the FETs in the DirectFET packages, and the performance was pretty nice overall as a fairly standard hysteretic converter. I could have improved the noise performance further by migrating to something other than the particular video amp at the heart of the loop, but I did reduce the overall gain of the power amp section and prefaced things with a nice front end. I think the output noise in the audio band unweighted was about 35uV rms, including the input section, but the maximum power was only about 100W into 4 ohms. It sounded good driving some big speakers, but I didn't have much time to audition. Standard big iron and copper and bulk cap unregulated supply, channels driven electrically inverted to assist reduction of rail pumping for bass mixed to both.
Although I never saw latchup behavior, evidently some did and IR respun the driver chip, and increased the minimum deadtime. I suspect performance was somewhat diminished, but four of us abandoned the founder, a sales guy, and his pest-control fortune finance guy.
I hope the folks at IR are surviving since the acquisition.