From Chord Electronics – enter the ‘perfect’ Class D amp | audiofi.net
Power DACs that completely eliminate interstage analog seem like the next big thing in audio reproduction.
I know that NAD and Linn are using this technology in some of their products, are there any more manufacturers that have power DAC products?
What I'd really like to know is are there currently any power DAC chipsets that us DIY'ers can get our hands on to build ourselves or is it just too early?
Power DACs that completely eliminate interstage analog seem like the next big thing in audio reproduction.
I know that NAD and Linn are using this technology in some of their products, are there any more manufacturers that have power DAC products?
What I'd really like to know is are there currently any power DAC chipsets that us DIY'ers can get our hands on to build ourselves or is it just too early?
Sony's higher end home theatre receivers from a few years ago had "S-Master Pro" amplifiers which were few directly either PCM or DSD digital signals. (STR-DAx000ES and STR-DAx100ES models). There was also a set of DACs in the box, but they were used only for the pre-amp outputs.
Digital PWM generation or any other control pattern for the switches of class D amps obviously is one of the coming steps.
And of course it is an obvios target, since the music storage went digital in the 80s.
Still the existing affordable approaches so far are not on the level of analog modulators.
If you go for 500kHz and want to bring the dominant harmonics below -90db, which you usually need to achieve an overall THD around -80db.. ..just calculate how few ps resolution do you have to ensure.
And this is just the most obvious cost driver.
The Chord interview is just advertising without relevant content.
Of course he does not tell their key knowledge - and he is a master in making a story from 5 year aged standard knowledge.
It's all about catching customers.
And of course it is an obvios target, since the music storage went digital in the 80s.
Still the existing affordable approaches so far are not on the level of analog modulators.
If you go for 500kHz and want to bring the dominant harmonics below -90db, which you usually need to achieve an overall THD around -80db.. ..just calculate how few ps resolution do you have to ensure.
And this is just the most obvious cost driver.
The Chord interview is just advertising without relevant content.
Of course he does not tell their key knowledge - and he is a master in making a story from 5 year aged standard knowledge.
It's all about catching customers.
Last edited:
DDX320 from Hifimediy
Mentioned in this thread:
http://www.diyaudio.com/forums/clas...gital-amp-project-s-may-rival-nad-c390dd.html
Mentioned in this thread:
http://www.diyaudio.com/forums/clas...gital-amp-project-s-may-rival-nad-c390dd.html
Do this "digital amps" use feedback from analogue speaker output back to the digital input stream? Otherwise it won t work well I assume. How could this work?
Nonono... as quoted above, it's digital all the way to the speaker! The speaker doesn't have a voltage applied to it ('cause that would be analog) instead the speaker is fed ones and zeros!
sigh.
interesting, Chord stuff isn't very good? I thought it was interesting that they were using proprietarily-programmed FPGAs in their DACs instead of off-the-shelf chips, I don't know of any other company that does that.
While GaNFETs are no longer "military technology", I know of only EPC that is using Digikey as a distributor, which means anyone can buy them, and they aren't too expensive. The problems start when you try to do anything with them, either in reality or simulation. Reality is they are so fast that "wiring" is not possible. To achieve the fractions of a nH lead inductances required to let these transistors fly you basically can't have any leads, so the the tiny factor is not a development, it's an unfortunate requirement. EPC's device models are pretty complex, to the point where you need a small parallel processing PC farm to simulate in minutes rather than days. So unless you have the computing power, the knowledge to rewrite simplified models that still basically work (which would not be easy in any case since it's the picosecond stuff going on with these transistors that makes them so special), or the ability to easily fabricate hybrid circuits around dice and pack it into a TO-3 can like in the suggestive but vague pictures with the suited man, the truth is that you wont get very far. The interesting thing to me is that the semiconductor technology is apparently approaching other physical limits, where hard switching class D simply can't pass. It might be time for a completely new way. And, by the way, Mr. Chord is apparently not describing a "Power DAC", just the same old with new parts.
Fortunately the achievable results with monster MosFets and monster series diodes and monster freewheeling diodes and the related huge geometries are still about factor 10 better than what today's of-the-shelf-digital-classD-modulators serve.
So the today's fun killers for a fully digital modulator, are not located in the power stage.
Nevertheless GaN is very promising, especially because it completely avoids the headache with the Qrr of the freewheeling diodes.
IMHO the tiny factor could be solved with silicone already pretty good, which becomes visible with the direct fets.
GaN got a little bit delayed (say 2 years? 😛 ), because nobody had a look to the dynamic Rdson before some lonely circuitry developers started to wonder about losses in their protos, which strongly contradicted vs the promises they believed earlier... More than one year ago on the PCIM it was promoted that the issue was solved. Today there is still not much of a portfolio available in the market place. IMHO we have to stay patient, it's not a mature technology yet.
The often promoted cascoded configuration with an additional normal MosFet appears convinient, because the GaNfet alone asks for unpleasant well controlled gate drive voltages. The cascode allows our well established gate drive methods, good for dinosaurs like me. No lunch for free: The cascode adds again a body diode with Qrr (fortunately just little Qrr, because the MosFet only needs to be a low voltage type) and further series impedances.
The tiny factor itself: Yes, small geometries are important and helpful, but already with geometries which humans can still handle manually, in class D one can achieve huge improvements vs state of the art designs of the year 2013.
FPGA:
It just means that Chord is not satisfied with what they get of the shelf and we have to trust that they can do better.
So the today's fun killers for a fully digital modulator, are not located in the power stage.
Nevertheless GaN is very promising, especially because it completely avoids the headache with the Qrr of the freewheeling diodes.
IMHO the tiny factor could be solved with silicone already pretty good, which becomes visible with the direct fets.
GaN got a little bit delayed (say 2 years? 😛 ), because nobody had a look to the dynamic Rdson before some lonely circuitry developers started to wonder about losses in their protos, which strongly contradicted vs the promises they believed earlier... More than one year ago on the PCIM it was promoted that the issue was solved. Today there is still not much of a portfolio available in the market place. IMHO we have to stay patient, it's not a mature technology yet.
The often promoted cascoded configuration with an additional normal MosFet appears convinient, because the GaNfet alone asks for unpleasant well controlled gate drive voltages. The cascode allows our well established gate drive methods, good for dinosaurs like me. No lunch for free: The cascode adds again a body diode with Qrr (fortunately just little Qrr, because the MosFet only needs to be a low voltage type) and further series impedances.
The tiny factor itself: Yes, small geometries are important and helpful, but already with geometries which humans can still handle manually, in class D one can achieve huge improvements vs state of the art designs of the year 2013.
FPGA:
It just means that Chord is not satisfied with what they get of the shelf and we have to trust that they can do better.
Well, did some experiments with GaN Fets as well, they are good with reverse recovery issues Qrr being near zero but the gate drive is very sensitive to noise because of the lesser VGS ON requirement. Some protos i have seen using 600V looks promising though in solar/fuel cell inverters.
I am more inclined towards multiphase class-D using IGBTs switched at 100khz in 4 phase bridged arrangement, keeping switching losses low and giving resultant ripple around 400k, enough for fullranger.......Further you don't have to bear with dreaded body diodes in that thing as well.
Not much to add.
Just wondering why/how:
Just less headache in terms of numbers - or avoiding the Qrr issue by principle?
Just wondering why/how:
Just less headache in terms of numbers - or avoiding the Qrr issue by principle?
Last edited:
Not much to add.
Just wondering why/how:
Just less headache in terms of numbers - or avoiding the Qrr issue by principle?
Choco,
IGBTs have no body diode issue, so one can use external diodes like Qspeed with very less Trr and Qrr to make the job easy. In one of my insane attempts i used a CREE SIC diode Qrr=Zero for 400V bus operated class-D with IXGH60N60C2 IXYS IGBT.
Multiphase Interleaving helps in better current conduction, less ripple, capability of low EMI and High current capacity to offer.
For lower power Infineon really has some good I2Pak mosfets as well. See the pics.
Kanwar
Attachments
For sure this isnt a bad idea....,but there are probably not many people with experience in this.
Fits to my view and is what I call less headache in terms of numbers.
Yes, SiC is the right way for high voltage switching diodes, if you are willing to spend the money.
For some reason there is a bunch of Cree zero recovery diodes also in my assortment.... CSD01060 and CSD05120. 😉
But up to 200V I am fine with Si-shottkys.
Frankly speaking - with normal house installations, I doubt that 10kW amps make much sense. Already my 2k4 causes an obvious fading of the lights in my flat during full power tests. And 2k4 into 4R is perfectly fine for 200V devices, without squeezing them beyond any reasonable limit.
Running 2x 2k4 (for bridged stereo) is definitely the limit which would make sense in locations which offer nothing but the usually available single phase power outlet - if not already to much.
...oohps we lost the focus of the thread! Sorry for OT.
Let's jump back to the power DAC topic.
Nice module - 12,288 taps is deliciously overkill. Here's hoping you don't build one with a big Virtex, the group delay would probably give you time to make dinner 😉
I've seen plenty of manufacturers in the recording/broadcast fields use their own DSP/FPGA instead of using an off-the-shelf device. Sometimes it's pride, usually it's because off-the-shelf offerings have some limitation in the application and the DSP/FPGA approach works better.
Whoaaaa, awesome
Also, you guys with all the diode-speak: interesting. I'm going to have to really look up how these things relate to eachother. I'm obviously not as well-versed as you guys and I'd like to be.
Well, the man behind Zetex DDFA was Rob Watts. Indeed a very clever guy. He did all the pulse Array stuff from the DAC64 onwards. He made some great DACs with his own brand back in the day (DPA), and John persuaded Rob to work with him. A very good business decision.
I have always wanted Rob to do DDFA+, and it looks like this is in the pipe.
Will be interesting to hear it.
I have always wanted Rob to do DDFA+, and it looks like this is in the pipe.
Will be interesting to hear it.
- Status
- Not open for further replies.