SystemD_2kW, any interest for an open design?

Hi vovatver,
you will need patience,...

I will wait humbly
... and i try to repeat Your feat.
I was hoping that You had a stage of Your project, when You said to yourself -
ok! this is the best that can make my half-bridge! and it is time to tell it about "Errors" by loop-back.

Ambitious project that You have already done - is worthy of admiration.
Thank You for the pleasure of seeing Your research.
I hope, that soon I will be able not only to strain You by questions,
but to say something on the merits.
 
Edit1:
@Reactance

Yeah - soft soaping
...feeling better again... :eek:
UcD is hard to beat in terms of simplicity. Well, bender's version is already pretty grown up with lots of additional components, but they all should be easily available and low cost - valid point.

I cannot comment on the Arnold cores, never used them and never had a detail check through their data sheets. 'Less famous' is not automatically bad.

Enjoy your DSP !



Edit2
@vovatver:
If I have time - yes, I do optimize THD in open loop situation for curiosity and then close the loop.
But up to now this open design was definitely a project which I pressed to bring it to a reasonable level within short time.
The more relaxed time of tweaking derivative version(s?) will start this summer.
 
Last edited:
... and i try to repeat Your feat.
....
I hope, that soon I will be able not only to strain You by questions,
but to say something on the merits.

You are going to build? :cool:
Please make sure to use the instruction set in the BOM of posting #329 (page 33), not the earlier one!
Let us see how things are going, and when struggling or something in the builders pack might be unclear, please let me know.
Thread or PM - both is fine.
 
500MHz ! ...sounding like you will need a 10bit frequency divider... :D
The power stage of SystemD_2k4 is fine for 500kHz, but you will have higher losses.
THD typically also increases with frequency (except you can win inversely more loop gain), because dead time distortion is a function of the ratio of dead time by period.
Keep us updated :cool:
 
Something new for rainy days

Thanks Markus
 

Attachments

  • IMG_20130506_111012.jpg
    IMG_20130506_111012.jpg
    880.9 KB · Views: 1,306
Hi CPX,
I guess your question is not linked to the school book beating, but to disturbances like you described it in posting #1968 of the 25-1200 thread:
http://www.diyaudio.com/forums/class-d/166214-ucd-25-watts-1200-watts-using-2-mosfets-197.html

To me this is sounding like you were suffering from one or both of these two effects.
a)
Known from selfoscillating SMPS:
Putting multiple of them close together can lead to uncontrolled sync and unsync by chance and repeatingly. ==> Often visible as fluctuating output.
If this effect happening between two class D amps, it is likely that the fluctuations are audible.
b)
Class D modulators are unrivaled champions in catching HF noise and converting it to audible noise. ;)

Most obvious difference vs your set up is that I am using copper shields for Q309, D311 & D313 which keep the parasitic capacitive charging of of the switching node inside small loops right at the switching stage, instead of spreading the noise over the entire heatsink. Connecting these shields to GND (like I do in V1.3) is just one option. It can also be fortunate to connect them to pos or neg rail right at the device. But without EMI measurement it is hard to tell the most fortunate connection.
Less obvious is optimizing the snubbering to the layout.
And of course every detail of the wiring of the PSU & signals.

So far I am not in trouble at all, because my proto is still single channel.
A first stereo set up will come with my faster version V1.4. Most likely within this year. I will post the stereo results, when ready.
...confident that this build in one enclosure goes trouble free. But only the real build can prove it.
Planned size for 2x1200W into 2R or bridged 2400W into 4R, including PSU, fanless:
19" rack, 1.5 HE, 300mm deep
 
Sync

Hi CPX,
I guess your question is not linked to the school book beating, but to disturbances like you described it in posting #1968 of the 25-1200 thread:
http://www.diyaudio.com/forums/class-d/166214-ucd-25-watts-1200-watts-using-2-mosfets-197.html

To me this is sounding like you were suffering from one or both of these two effects.
a)
Known from selfoscillating SMPS:
Putting multiple of them close together can lead to uncontrolled sync and unsync by chance and repeatingly. ==> Often visible as fluctuating output.
If this effect happening between two class D amps, it is likely that the fluctuations are audible.
b)
Class D modulators are unrivaled champions in catching HF noise and converting it to audible noise. ;)

Most obvious difference vs your set up is that I am using copper shields for Q309, D311 & D313 which keep the parasitic capacitive charging of of the switching node inside small loops right at the switching stage, instead of spreading the noise over the entire heatsink. Connecting these shields to GND (like I do in V1.3) is just one option. It can also be fortunate to connect them to pos or neg rail right at the device. But without EMI measurement it is hard to tell the most fortunate connection.
Less obvious is optimizing the snubbering to the layout.
And of course every detail of the wiring of the PSU & signals.

So far I am not in trouble at all, because my proto is still single channel.
A first stereo set up will come with my faster version V1.4. Most likely within this year. I will post the stereo results, when ready.
...confident that this build in one enclosure goes trouble free. But only the real build can prove it.
Planned size for 2x1200W into 2R or bridged 2400W into 4R, including PSU, fanless:
19" rack, 1.5 HE, 300mm deep

Hi Choc ,

Are you planning to sync the oscillators in a multichannel version of your design ? and what would be your motivation to do so or not ?
Great thread here! as an aged engineer I'm learning a lot here and I love your clear explanations ! very educational .

Thanks and cheers ,

Rens
 
Yes,i was referring to rf interferences in post 1968, you guessed correctly:).
I am working at a new design and i want to solve this problem.
I didn't notice any fluctuating output but i have variable high frequency(5-15khz) noise that sounds like am radio and increases with load current.I will see if i can implement rf shielding like in your project and/or use a clocked design.
I am also very interested to see how you will implement power and signal ground connections between each amplifier,power supply and sound source and how you will reduce crosstalk between channels.

:cheers:
 
Hi Choc ,
Are you planning to sync the oscillators in a multichannel version of your design ? and what would be your motivation to do so or not ?

Hi Rens,
I do not intend to sync, because I came to the conclusion that the pure school book beating is not much of an issue. It only forms an amplitude modulation of an HF signal. It is not audible as long we do not connect an AM demodulator to the beaten HF.
Some tests and thoughts here:
http://www.diyaudio.com/forums/class-d/201738-clock-not-clock.html
So the only remaining issue from school book beating is in bridged configuration, where the non linearities of poor speakers could act as an AM demodulator. Using this amp in bridged configuration to drive your tweeters in the living room bears the chance to make the effect audible.
...well, driving a tweeter for a living room from a bridged 2400W classD amp is not the most high end approach anyway. (Curious Choco will most likely test even this, but the relevance of the result is low.)

Regarding the non school book issues with undesired sync/unsync effects, Bruno has already proven in mass production that it is possible to avoid headache by a proper layout and set up.
So there will be one or the other learning ahead, but I do not expect that I will need wonders.



I will see if i can implement rf shielding like in your project and/or use a clocked design.
I am also very interested to see how you will implement power and signal ground connections between each amplifier,power supply and sound source and how you will reduce crosstalk between channels.
:cheers:
Wishing you good luck that the shielding will help - I am curious, keep us updated.

My wiring concept for stereo/bridged is still not finalised.
I intend to incorporate also some common mode filtering in the wiring.
Before building I have to analyse current paths, loops and fields.
Most ugly will again be the bridged configuration, because in this case you will have high currents in the GND connection between both channels.
...will publish the solution after the build is ready.
 
Hi Dimitri,
in order to allow stereo with asymetric input signals, GND of both chanels must be connected together.
In bridged config the GND will remain to be connected together. The speaker is inbetween both amp outputs.

Situation 1, amp A delivers pos output voltage & and Amp B delivers neg output voltage.
Current loop:
Pos supply of amp A ==> Amp A ==> speaker ==> Amp B ==> neg supply of Amp B ==> GND ==> Pos supply of amp A

Situation 2, amp A delivers neg output voltage & and Amp B delivers pos output voltage.
Current loop:
Pos supply of amp B ==> Amp B ==> speaker ==> Amp A ==> neg supply of Amp A ==> GND ==> Pos supply of amp B
 
Thank you Markus,
stereo SE - two bipolar power supplies - no problem, grounds are isolated
stereo SE - one bipolar power supply - you can invert the input signal for one of the channels and change the load polarity in the same channel to prevent rail pumping and low frequency current through GND.
mono bridge - why do you need two bipolar power supplies for bridge?
 
stereo SE - two bipolar power supplies - no problem, grounds are isolated
.. no problem: Yes.
...grounds isolated: Not really. Even if not connected in the amp, you will have both GNDs connected through the signal wires to your pre amp or signal source and there they will be connected together.

stereo SE - one bipolar power supply - you can invert the input signal for one of the channels and change the load polarity in the same channel to prevent rail pumping and low frequency current through GND.
Valid constellation.
But I decided to go for separate supplies, in order to achieve a high channel separation also at higher frequencies where the PSRR of classD amps is not at all convincing.
On the other hand I would not fight your view, if you say that channel separation is just for measurements and not so important for sound.


mono bridge - why do you need two bipolar power supplies for bridge?
It should become one amp, which can handle stereo and also mono bridged and also offer symmetric and asymmetric inputs.

:D Obviously, I am a victim of my own I-want-it-all-design-requirements. :D
In the moment I am considering your first arrangement, but with an ultra low impedance equipotential plate in the amp which connects both GNDs.
Basically this is just extending the GND plane concept of each PCB - still have to make up my mind where and how to connect it to the PCBs.
On the PCBs I found a geometric arrangement of the components, which ensures that the resulting current distributions in the GND plane do not badly mess signal and power... looking to the equipotential plate and considering signal cables which have connected GNDs also at the signal source - there is a new challenge :trapper: