SystemD_2kW, any interest for an open design? - Page 6 - diyAudio
Go Back   Home > Forums > Amplifiers > Class D
Home Forums Rules Articles diyAudio Store Gallery Wiki Blogs Register Donations FAQ Calendar Search Today's Posts Mark Forums Read

Class D Switching Power Amplifiers and Power D/A conversion

Please consider donating to help us continue to serve you.

Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
Thread Tools Search this Thread
Old 26th November 2012, 10:37 PM   #51
diyAudio Member
Join Date: Dec 2003
Location: Munich
...good point with the BSS - must have something like this somewhere in my component boxes....
Idle frequency is 380kHz, with the shown hysteresis and delays (in theory/simulation). But I trust that the delays are more or less realistic.
In SystemD_MD I have 140ns delay from IRS-input to halfbridge output, so 230ns with the IRFP4668 is already a slow guess, I think/hope it will be faster which would ask for a slightly larger hysteresis to fit the same idling frequency and would allow even better control of the frequency by the dynamic hysteresis.
  Reply With Quote
Old 28th November 2012, 09:07 AM   #52
diyAudio Member
Join Date: May 2002
Location: Switzerland
Default Dynamic hysteresis control

Here you can see a simulation model of my proposal. It uses a 2nd order feedback loop with post-filter feedback as general topology.

It would of course also work with pre-filter or other post filter topologies (than the one shown here).

Because it is only a simulation model with an output voltage of +- 1 V the "comparator and the output stage is the same (maybe it would have been better to separte them for clarity).

With a real implementation one would take the positive feedback for the hysteresis from the comparator and the NFB from the power output.

The hysteresis is controlled by modulating a reference voltage. This voltage is the actual voltage that is positively fed-back (to cause the hysteresis). The multiplier is actually just acting as a polarity switch. I.e. the reference voltage is either inverted or not - depending on the output state of the comparator.
In practice one would implement this "multiplier" either with a CMOS analog switch that is controlled by the comparator output voltage (and which is switching between a positive and a negative reference voltage) or by modulating the supply voltage of a CMOS -buffer/-levelshifter whose input is connected to the comparator output.

The amount of hysteresis is then controlled by subtracting a voltage - that is dependant on the input voltage of the amp - from the reference voltage. In this example here the subtracted voltage is directly proportional to the input amplitude. A deeper analysis would be needed in order to find out whether a different mathematical function (or just a coarse approximation of such a function) than just proportional would be advantageous. A practical circuit would also need some additional simple means to tame the hyteresis control under overload conditions.

I don't know whether something like this has already been patented. If not - maybe heaps of amps using it may pop up from China quickly

For the impatient ones:
If driven to around 10% of max outpult voltage the influence on THD (which is low in this situation anyway) is benign, at 90% drive it means a THD improvement of about tenfold. But these are simulations and have to be taken with the usual grain of salt of course.


Attached Images
File Type: jpg hysteresis_contr.JPG (45.5 KB, 1256 views)
  Reply With Quote
Old 28th November 2012, 10:13 AM   #53
diyAudio Member
Join Date: Dec 2003
Location: Munich
Great! - now I have to think and cannot comment immediately...

In order to get my brain free to think about Charles' version,
first an update on the comparator topic.
The ultra simple comparator is doing a fairly good job in simulation.
From data sheet of IRS20957 I came to the conclusion that it would
be better to shift the floating input stage towards positive, because
VSS must not be pulled below COM. In case of missing high power rails
COM is likely to be equal with GND and in this case putting VSS to -8V is not allowed. Clamping by a diode is ugh, because this would then pull the -8V towards GND...
So shifting VSS some volts towards positive is better - and also possible with a triple-BJT-comparator.

Nevertheless I am intending to use integrated comparators - just because of component count and number solder joints. The headaches for debugging a new built grow proportional with the number of solder joints and exponential with the number of components. Adding up my debugging times when building SystemD_MD (317 components, unbridged) is likely to sum up to a full day for debugging only.
Impossible to transfer such a design to other DIYers, except the ones who could design it on their own anyway.
I definitely want to avoid that this happens to SystemD_2kW.

Many thanks for all your proposals of comparators.
I think there is a simple way to set a PCB which allows LT1016 and LM306.
Depending on personal taste and availability one or the other can be placed.
Have a look to the attached sketch, which is showing one possible PCB solution for both (final solution will depend on the very layout details).
Well, allowing this flexibility again generates additional chances to build something wrong, but I think it is easy to handle and worth the flexibility.
Attached Images
File Type: jpg DoubleFit.JPG (55.3 KB, 955 views)
  Reply With Quote
Old 28th November 2012, 10:51 AM   #54
diyAudio Member
Join Date: May 2002
Location: Switzerland
So here are the sims.
Keep in mind that the improvement will most probably be better as soon as the phase-lag between the amp input and the input to the modulator is taken into consideration as well. The gain of E87 was set to 0.7 when dynamic control was enabled.

The graphics show the following:

Output spectrum with 10% drive at 1 kHz and dynamic control disabled.

Dito with dynamic control enabled -> not a big difference.

Output spectrum with 90% drive and dynamic control disabled.

The same with dynamic control enabled - at least 20 dB improvement in THD and a clear restriction of the frequency range that is swept by the switching frequency. Not so elegant is the ratio between k3 and k5.

Output signal without dynamic hysteresis control at 90% drive.

The same with hysteresis control enabled.


Attached Images
File Type: jpg low_inp_no_contr.JPG (51.2 KB, 883 views)
File Type: jpg low_inp_contr.JPG (50.0 KB, 766 views)
File Type: jpg hi_inp_no_contr.JPG (53.9 KB, 719 views)
File Type: jpg hi_inp_contr.JPG (49.4 KB, 82 views)
File Type: jpg sin_no_contr.JPG (43.6 KB, 137 views)
File Type: jpg sin_contr.JPG (41.0 KB, 149 views)
  Reply With Quote
Old 28th November 2012, 11:59 AM   #55
diyAudio Member
Join Date: Dec 2003
Location: Munich
Hi Charles,
....I think I got it. That's nice. And your sim results are looking pretty nice.
You are not just controlling the hysteresis in order to avoid massive voltage ripple at high output levels - you vary the hysteresis over a large voltage range in order to optimize distortion and avoid massive voltage ripple at the same time.
I did not compare distortion at 90% with and without dynamic hyteresis in my approach. ==> coming soon (one day off today :-) )
Furtheron I have to check if I can parametrize to optimize for this dual functionality in my implementation as well. Great input!

But let's stick with your proposal of implementation.
As already visible in my previous posting - for this project I am acting like a "Erbsenzähler" (English: Nitpicker) in terms of component count and solder joints.
I even considered to skip the triangle shaping and dynamic hysteresis - but no, that would be to poor!

5x resistor
1x dual OP amp
2x supply cap for dual OP amp

As long as we stay with the proportional approach, this gain can be incorporated into the ABS circuit.

For sure something reasonable available without additional effort (i.e. pick from LED chain...)

1x OP amp
2x supply cap for OP amp
4x resistor

Multiplier (Buffer with modulated supply):
6x resistor
1x dual OP amp
2x supply cap for dual OP amp
1x buffer
2x supply cap for buffer (RC between OP amp and supply pin of buffer)
2x resistor (RC between OP amp and supply pin of buffer)
1x summing resistor towards hysteresis


Hm, overall not to bad, but already extensive.
Basically I would prefer to use the CMOS-Switch-Multiplier, but from my understanding we would not get right the polarity information from the PWM signal.
The analogue input of the CMOS-Switch is needed to make use of the analog information from the substractor.
So only the ON/OFF input of the CMOS can be used for the PWM information... -or do I miss something?

...looks like your control loop also incorporates a shaped D portion..
  Reply With Quote
Old 28th November 2012, 12:22 PM   #56
diyAudio Member
Join Date: May 2002
Location: Switzerland
The "analog" part of a CD4053 (3x 2 channel multiplexer or just atoggle switch IMHO) can be run from symmetrical suplies and handle negative voltages whil the input control is done with H = Vdd and L = Vss.
But you would need to make a negative mirror image of the reference voltage as well. I.e. one more OP-AMP plus two resistors plus means for decoupling.


  Reply With Quote
Old 28th November 2012, 12:37 PM   #57
AP2 is offline AP2  Italy
Join Date: Dec 2009
Location: Italy
Congratulations to you two, this is the right way to play well in a class D.
In particular, the idea of Charlie is very similar to that used by me. (I also control a proportion value of the audio frequency), voltage control is also reported to the frequency audio I mean. (my DXA use just a similar concept)
This system also, delete the errors added in the modulator when applying a scheme of nfb, such as self-oscillating post-filter feedback.

  Reply With Quote
Old 28th November 2012, 12:42 PM   #58
diyAudio Member
Join Date: Dec 2003
Location: Munich
Yes, symmetrical supply for the CMOS always worked fine in my circuits.
Good idea to use it as a toggled switch with information from two substractors and references.
Hm, I think then we would not need the ABS anymore, don't we?
  Reply With Quote
Old 28th November 2012, 12:54 PM   #59
diyAudio Member
Join Date: May 2002
Location: Switzerland
Not sure - I simulated with symmetrical hysteresis. Maybe it works as well when each control voltage is just derived from either half-wave of the input signal. Although I somehow don't like it. and you would need two half-wave rectifiers which wouldn't lower parts-count much.


  Reply With Quote
Old 28th November 2012, 01:47 PM   #60
diyAudio Member
Join Date: Dec 2003
Location: Munich
I took my simulation - values and schematic circuit according the posting #3 and #14 - and checked the influence of the dynamic hysteresis on the harmonics at 90%.

Also in this circuit the dynamic hysteresis is helpful for harmonics at high levels. Already unmodified it improves the situation about 15db. Not sure if it can reach identical results as Charles' solution, just by optimized parameters, but already 15db positive effect is nice.
This result is fitting to the obvious effect that the shape of the feedback carrier turns less and less triangular the more the frequencies drops.
Many thanks to Charles for pointing to this.

1. Output shape with constant hysteresis
2. Output shape with dynamic hysteresis
3. FFT with constant hysteresis
4. FFT with dynamic hysteresis
Attached Images
File Type: jpeg Transient_fixedHyst_900.jpeg (54.5 KB, 171 views)
File Type: jpeg Transient_dynHyst_900.jpeg (51.6 KB, 120 views)
File Type: jpeg FFT_fixedHyst_900.jpeg (91.8 KB, 121 views)
File Type: jpeg FFT_dynHyst_900.jpeg (94.7 KB, 101 views)
  Reply With Quote


Hide this!Advertise here!
Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off

Similar Threads
Thread Thread Starter Forum Replies Last Post
Probing interest in My inrush limiting design nattawa Power Supplies 5 8th October 2016 08:28 PM
O2 Front Panel INTEREST POLL your own design Llama16 Group Buys 1 12th September 2012 05:23 PM
Custom open baffle 12" woofer - any interest? SimontY Group Buys 2 4th March 2009 01:53 PM
Open baffle design lawbadman Multi-Way 1 17th January 2008 12:15 AM
Open call for suggestions on Open Source DIY Audio Design gfergy Everything Else 1 15th April 2007 07:33 AM

New To Site? Need Help?

All times are GMT. The time now is 04:51 AM.

vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2017 DragonByte Technologies Ltd.
Copyright ©1999-2017 diyAudio

Content Relevant URLs by vBSEO 3.3.2