Go Back   Home > Forums > >
Home Forums Rules Articles diyAudio Store Blogs Gallery Wiki Register Donations FAQ Calendar Search Today's Posts Mark Forums Read

Class D Switching Power Amplifiers and Power D/A conversion

Class D Design Issues
Class D Design Issues
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
Reply
 
Thread Tools Search this Thread
Old 23rd August 2019, 12:30 PM   #141
kees52 is offline kees52  Netherlands
diyAudio Member
 
Join Date: May 2009
Location: Sprang-capelle Holland.
I have experience that a class d with only prefeedback sounds warmer, and has more air, but need a good quality coil and cap, prefere aircoil, that is with multilevel systems more easy, can keep it small.
  Reply With Quote
Old 23rd August 2019, 06:02 PM   #142
kees52 is offline kees52  Netherlands
diyAudio Member
 
Join Date: May 2009
Location: Sprang-capelle Holland.
This paper is also interesting, did read this a long time ago, contains very usefull info.

https://www.hobbielektronika.hu/foru....php?id=253790
  Reply With Quote
Old 25th August 2019, 12:36 PM   #143
MorbidFractal is offline MorbidFractal
diyAudio Member
 
Join Date: Nov 2017
Things such as...

https://www.hobbielektronika.hu/foru....php?id=253790

With 4 foot long equations tend to make me run a mile. I have looked elsewhere for information about your topology and there are claims that the available closed loop bandwidth is improved by factors of N where N are the number of levels. I would be wary about the interpretation of such claims.

In the above paper reference is made to Feed Forward. As far as I can see this is Pre-Filter Feedback so why not call it Pre-Filter Feedback? I do not have a model for your topology and am uncertain that I could recreate it. In particular the manner in which the flying capacitors are charged/recharged. That might happen as a matter of course during operation.

If you could provide such a base LTSpice model, .asc file, I would be happy to have a play. Failing that you might consider the following but bear in mind that it might be incomplete or slightly incorrect.

You suggest that the distortion performance of such a topology is worse as a result of the added switching nodes. You should be able to use Pre Filter Feedback to improve that as follows.

Let's say you have a Section operating frequency of Fs and a Total operating frequency of Ft (N.Fs). As I suggest I do not believe that with Post Filter Feedback you can exceed a crossover frequency of Fs/PI. However with Pre Filter Feedback a crossover frequency for that particular loop of Ft/PI should be realisable.

Consider the first picture. This is standard practice for Pre Filter Feedback. Values are nominal and the value for CINT is omitted. You calculate CINT such that the slope of the integrated output square wave matches the slope of the modulator triangle wave. At zero output the triangle at VEA is half the amplitude of the modulator triangle.

Under those conditions the loop crosses over at Fs/Pi. Single section.

In your half bridge, three level[?], instance at first sight things look slightly complicated due to the level shifts involved but it would appear that such shifts, top to bottom, will cancel. It also appears to be the case that the averaged outputs, top and bottom, are half of the main summed output.

Assuming this is the case then you can proceed with the second picture. CINT does not change. You just add in the extra resistors and comparators. You may wish to consider adding a DC servo around this section should you find that the Top/Bottom DC offset match is not precise. I would guess that would be unlikely.

Having become inconsistent. Switch Top, STOP, Switch Mid, SMID, Switch Bottom, SBOT and Drive Top, DRVT, Drive Mid, DRVM, and Drive Bottom, DRVB.

Assuming my suggestion about scaling is correct then the new resistors will have the same values as the original one. In the half bridge case your new crossover frequency becomes 2Fs/PI. Again this is strictly Pre Filter.

This is extended to your full bridge case in the third picture. Again I am slightly losing the plot over labelling but Switch Left Top, Mid, Bottom. Switch Right Top, Mid, Bottom and similar with the comparator outputs. Then whatever with the triangle waves.

Also just in case when I refer to Switch I am really referring to the associated switched nodes.

Again all component values remain the same. You just add some more resistors and comparators. This time your crossover frequency becomes 4Fs/PI or Ft/PI. Once complete then the closed loop response, DEMand to Pre Filter Output for the values chosen is 39 Volts per Volt.

At this point you can add the Post Filter Ouput Feedback around this internal loop as proposed in your other paper...

https://www.meng-engineering.ch/_dow...upply-170V.pdf

This also tries to introduce four foot long equations but the simple solution is to mess about in Spice. The fourth picture shows a 40KHz cutoff filter scaled for a 4R load, yes speakers are not but. E1 represents the previously designed internal loop with a gain of 39 Volts per Volt. It is ideal but all you care about is that it crosses over, runs out of steam, at Ft/PI and you, almost, don't care about that because your external loop is going to be limited to Fs/PI.

Plotting the control to output response you find out where the -7dB point is. That's because we are going to introduce a phase bump, similar to your paper, of 7dB in the external loop to cancel one of the filter poles. In this case it is at 375KHz. Your section switching frequency is 240KHz so your maximum loop crossover is 240K/PI or 76KHz. That sets the gain in the external loop at this frequency as (76/375)^2 or 0.04, at the moment the response is second order hence the squared term.

We end up with the fifth picture. Here I have used the trick of breaking the loop with VAC and plotting V(a)/V(b) to get the loop behaviour and -7dB point. R1, standard value, is low and we do not precisely hit 76KHz. Such is life and other things may matter more.

Now we add the phase bump as in the sixth picture. It involves attenuating the low frequency gain by a factor of five so we also have to increase the error amplifier gain by the same factor to bring things back. The capacitor value is chosen to have an impedance of 13K5 at the desired frequency of 76KHz.

The point... If we unload the filter as shown in the seventh picture then crossover is still occurring with 42 degrees phase margin as a result of the added phase bump and therefore we hope that the system is still stable.

We also get to nominally break the DC path, add a pole, around the error amplifier as shown in the eighth picture to improve the low frequency performance. In this case at about 40KHz but caveats here and elsewhere.

The almost don't care in respect of ignoring Ft/PI given you are limited to Fs/PI assumes the closed internal loop will not contribute phase shifts to the external loop. For Ft = 4Fs that assumption might be valid. Ft = 2Fs probably not. Also breaking the external loop as mentioned above robs phase margin.

Then you get into the wonderful world of clamping or limiting error amplifiers when the loop loses the plot in order to help it gracefully find its way back again.

Now I'll try to add the right pictures in the right order. Just a moment. Hopefully that works.


...
  Reply With Quote
Old 26th August 2019, 09:58 AM   #144
kees52 is offline kees52  Netherlands
diyAudio Member
 
Join Date: May 2009
Location: Sprang-capelle Holland.
Hi orbitfractal.


First I have I have from reading, so the answers or this is better or not is from papers not mine, what I do is play with it and try to improve.

However with multilevel, we have the issue that fase is important to stay into the multiplication factor and keep multilevel switching, I havd done from three to 8, but it get pretty complicated with more levels, I stick on max 5 with two three levels bridget and 90 degree shift between, that makes a 5 level with less complicated circuit.

I have afcourse the need of A good driver ic for the mosfets, or a good discrete schematic, I do not now if the multilevel switching has inpact on the upper fet who needs to have a voltage pump, this pump do not switch when that fet is not active in a multilevel, but maybe I am wrong, there chips who does 5 level and it works.

Do I see right you do feedback from every level? and sum it?

I like discrete, can play more then.

Here a two .asc to play with, and have fun, if you have not some models ask me I will send.

regards

Last edited by kees52; 26th August 2019 at 10:03 AM.
  Reply With Quote
Old 26th August 2019, 07:45 PM   #145
MorbidFractal is offline MorbidFractal
diyAudio Member
 
Join Date: Nov 2017
Here's my attempt at your half bridge.

You'll notice that my model is very much simplified and, for the moment just includes the proposed Pre Filter Feedback. I was more than surprised when it worked first time.

My original suggestion as to scaling of resistors was slightly wrong. You have to ratio them 2:1. I hacked the slope matching in Spice as well.

The performance looks a bit dirt. Maybe I got something wrong but...

Have a poke about.
Attached Images
File Type: png Screenshot from 2019-08-26 20-41-12.png (54.8 KB, 22 views)
File Type: png Screenshot from 2019-08-26 20-41-50.png (23.1 KB, 23 views)
File Type: png Screenshot from 2019-08-26 20-43-13.png (25.0 KB, 13 views)
Attached Files
File Type: asc halfflykees52.asc (7.7 KB, 2 views)
  Reply With Quote
Old 27th August 2019, 12:58 PM   #146
kees52 is offline kees52  Netherlands
diyAudio Member
 
Join Date: May 2009
Location: Sprang-capelle Holland.
Quote:
Originally Posted by MorbidFractal View Post
Here's my attempt at your half bridge.

You'll notice that my model is very much simplified and, for the moment just includes the proposed Pre Filter Feedback. I was more than surprised when it worked first time.

My original suggestion as to scaling of resistors was slightly wrong. You have to ratio them 2:1. I hacked the slope matching in Spice as well.

The performance looks a bit dirt. Maybe I got something wrong but...

Have a poke about.

Thanks for the help, Do you now a good mosfet driver ic for use with multilevel? I have some worries about the bootstrap circuit when I get multilevel, is it fast enough to provide the voltages in such a ic, I think it works because more people are use them in inverters.

Thanks for the help, I go simplify also the system, using models in stead of complete ic,s. You have clearly much more experience with LTspice then me, and learn it at mine age 62 takes longer as normal.

I do now the circlotron who starts again when weather get less hot, sitting behind a soldering iron with 32 degrees is hot, like today, so it is fall project (hybrid amp, autobias).

Your design do the same, the multilevel switching get bad with feedback and a extra faseshift that it cause, only the multilevel btl with slef oscillation do work, the faseshift is because the triangles do not get changed, maybe a faseshifting there can repair the switching, with self oscillating it do that automatically and do multiplying the frequency.

I have included the feedbacked self oscilating type, probe the resistor on the output before the filter to see it.

Stil it is that feedback multilevel is not that easy except self osc types until three level noth higher.

regards
Attached Images
File Type: jpg ScreenHunter_2517 Aug. 27 14.02.jpg (204.8 KB, 12 views)
File Type: jpg ScreenHunter_2519 Aug. 27 14.26.jpg (199.9 KB, 13 views)
Attached Files
File Type: asc 5 level offset-flycap-2019-5level-single.asc (13.2 KB, 1 views)
File Type: asc ucd orig_sinus_ucd-threelevel.asc (15.8 KB, 2 views)
  Reply With Quote
Old 27th August 2019, 09:53 PM   #147
kees52 is offline kees52  Netherlands
diyAudio Member
 
Join Date: May 2009
Location: Sprang-capelle Holland.
Here I have test with cascade stages, this does better and is open loop already nice.

Do you setup your schematic for deadtime?

When you try it, set the opamps vack because I had use models, and you did universal, need fast opamps. First plot op pic is openloop test, rest is feedback with two integrators.

What means the numbers between brackets into the ltspice harmonic distortion plot. And why I have such a lot 3th harmonics? deatime maybe?.

regards
Attached Images
File Type: jpg ScreenHunter_2520 Aug. 27 18.53.jpg (242.0 KB, 13 views)
File Type: jpg ScreenHunter_2521 Aug. 27 19.27.jpg (328.6 KB, 12 views)
File Type: jpg ScreenHunter_2523 Aug. 27 23.46.jpg (394.8 KB, 12 views)
Attached Files
File Type: asc halfflykees52cascade.asc (10.6 KB, 2 views)

Last edited by kees52; 27th August 2019 at 10:10 PM.
  Reply With Quote

Reply


Class D Design IssuesHide 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

Forum Jump

Similar Threads
Thread Thread Starter Forum Replies Last Post
Forum for issues related to the design of Class D amplifiers sovadk Class D 57 13th December 2006 06:45 AM


New To Site? Need Help?

All times are GMT. The time now is 02:28 AM.


Search Engine Optimisation provided by DragonByte SEO (Pro) - vBulletin Mods & Addons Copyright © 2019 DragonByte Technologies Ltd.
Resources saved on this page: MySQL 15.00%
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2019 DragonByte Technologies Ltd.
Copyright ©1999-2019 diyAudio
Wiki