A and T labs K6 SMPS for Class D amp use - Page 3 - diyAudio
Go Back   Home > Forums > Amplifiers > Class D

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
Reply
 
Thread Tools Search this Thread
Old 18th April 2005, 02:05 PM   #21
ghemink is offline ghemink  Netherlands
diyAudio Member
 
Join Date: Jun 2004
Location: Japan
Quote:
Originally posted by ghemink



With 0V at the input of the Sg3525, nothing will happen, it will have 0 output. The sg3525 starts to work from about 0.9V at its input, I think it will give a 50% duty cycle at around 2V in or so. Effectively, the duty cycle will be almost 100%, so output voltage would be about 130V, this will blow my caps and I don't want to try that.

I will soon try to modify the feedback loop in the simplest possible way. Not adding an zero, just keep it as it is, just increase the size of the cap in the feedback loop to compensate for the lower resonance frequency of the LC filter with the bigger output cap. This should then still work with the original caps as well.

Best regards

Gertjan

OK, here are the results. I added a cap of 0.47uF in parallel to C25 see the schematics at:

http://www.a-and-t-labs.com/K6_Sw_Amp/index.htm

This moves the compensating zero in the compensation network from 100Hz down to 33Hz, this is about the factor 3 that the resonance frequency of the LC output filter would go down with a 10.000uF cap at the output.

I measured the smps with the default setting and with the added compensation cap with the default output caps and with an additional 8200uF cap at each rail. Plan to use high grade caps for that, but now just used a few caps that were lying around (so not super high quality).

The results are very good. The SMPS is completely stable with and without the added cap, as expected of course. Of course, the feedback loop reacts slower to any load changes, but it recovers very gentle, no overshoots, so lots of phase margin. I could increase the gain (have not done that), to reduce the 100Hz ripple more.

I heated up my dummy loads and went as high as 5.3A load. The output voltage stayed stable at about 63.8V at any load between 0.38A and 5.3A, at 5.3A we are talking 675W power, the dummy load got pretty hot.

The 100Hz ripple at the highest load of 5.3A was about 1.9V peak/peak, so about 670mV RMS with the standard caps and went down to 660mV peak/peak or 230mV RMS with the added 8200uF caps. So I got about a factor 3 reduction in 100Hz ripple.

The HF ripple consisting of mainly 150kHz and a smaller 75kHz component reached about 250mV peak/peak at 5.3A and is about 100mV peak/peak at 3A and lower.

I'm already very happy with this performance, I will in a later stage hook up my Elna Cerafines and I will also try to increase the gain since I'm sure I have a lot of phase margin. Possibly with the Elna's the HF ripple is further reduced. I can also play with some additional black gates that I have to see if I can reduce that HF ripple further. Besides that, I also have the white noise active supplies that I could hookup after the SMPS, expect than a super clean output without any 100Hz ripple and without any HF ripple (hopefully :-)).

Maybe in the weekend or later next week, I plan to hook it up to a couple of UcD40 amps.

Attached are two scope measurements for the default case and the case with added 8200uF per rails. The load is switched from 164 Ohm to 41 Ohm, so from about 0.75A to about 3A. Note that the scales are blown up big time to 2V per division in both cases. Both + and - rails are shown.

Looks quite good I think.

Best regards

Gertjan
Attached Images
File Type: jpg default.jpg (47.5 KB, 908 views)
  Reply With Quote
Old 18th April 2005, 02:06 PM   #22
ghemink is offline ghemink  Netherlands
diyAudio Member
 
Join Date: Jun 2004
Location: Japan
This is with the 8200uF caps.


I know which one I prefer :-)
Attached Images
File Type: jpg 8200uf.jpg (46.3 KB, 900 views)
  Reply With Quote
Old 18th April 2005, 02:18 PM   #23
ghemink is offline ghemink  Netherlands
diyAudio Member
 
Join Date: Jun 2004
Location: Japan
Quote:
Originally posted by ghemink
This is with the 8200uF caps.


I know which one I prefer :-)

In case anybody wonders, the scope is a PC based scope with 12 bit resolution up to 50MHz sampling and 8 bit up to 100Mhz. It can do up to 14 bit and 16 bit resolution at lower sample frequencies (it uses oversampling). The scope is the HS3-100 from http://www.tiepie.nl (but better buy it in the US, a lot cheaper than in Europe). It also has an AWG (arbitrary waveforg generator) that is a great tool that I used to generate signals for the open loop measurements, no seperate generator needed, very nice.

Best regards

Gertjan
  Reply With Quote
Old 19th April 2005, 02:41 AM   #24
diyAudio Member
 
Join Date: Jan 2002
Location: Wheaton, Il.
Default K6PS

I'm delighted to see DIY so alive and well, and the design changes/ improvemnts being undertaken! Very interesting!

FYI, one simplification that a builder told me about, as they adapted the design to their application, was to eliminate the op amps in the feedback loop, and fed the positive output directly to the SG3525 (via a filter network). The reason for the op amps is to regulate on both the positive and negative outputs of the supply. In retrospect, given that most audio loads are symetric on average, this was an excessively complicated approach.

Ironically, I sell far more stand-alone K6 power supplies than the amps - a bit surprising, as the design intent of the supply was only weight reduction, while the K6 amp was intended to be the star. They do really well in listenning tests, even compared to the $xK stuff out there. Oh well!

Reinhard
__________________
R Metz
  Reply With Quote
Old 19th April 2005, 03:42 AM   #25
ghemink is offline ghemink  Netherlands
diyAudio Member
 
Join Date: Jun 2004
Location: Japan
Default Re: K6PS

Quote:
Originally posted by rmetz
I'm delighted to see DIY so alive and well, and the design changes/ improvemnts being undertaken! Very interesting!

FYI, one simplification that a builder told me about, as they adapted the design to their application, was to eliminate the op amps in the feedback loop, and fed the positive output directly to the SG3525 (via a filter network). The reason for the op amps is to regulate on both the positive and negative outputs of the supply. In retrospect, given that most audio loads are symetric on average, this was an excessively complicated approach.

Ironically, I sell far more stand-alone K6 power supplies than the amps - a bit surprising, as the design intent of the supply was only weight reduction, while the K6 amp was intended to be the star. They do really well in listenning tests, even compared to the $xK stuff out there. Oh well!

Reinhard
Hello Reinhard,

Well for some applications, it maybe nice to have the feedback from both rails as you have implemented it, in many cases both rails are not equally loaded (especially with Class D amps). Nice thing is also that you limit the maximum voltage from each rail that goes into the feedback network by the added zener diodes. This could actually be nice when used with class D amps since power supply pumping may occur and result in a higher rail voltage on one of the rails. As a result, the other rail would go down excessivly when trying to compensate for the rail that is being pumped up. Because of this, power supply pumping potential, I wanted to use bigger caps. Probably for use with AB amps, the default configuration is good enough. However, taking feedback from only one rail, is probably even tricky with class AB amps.

I also like the current limiting function, have not tried it yet :-) would need to heat up my dummy loads and then switch on some lightbulbs in addition to that. Don`t know whether I`m going to try that.

So I would say, don`t simplify the design, I think the features it has now and the feedback taken from both rails are desired. I just did not like the overshoot during startup. The output went up to about 77V or so before settling at about 63.5V (had expected 58V), so using it with 63V caps is a no-go (the 77V surge maybe too much). So I added the RC filter for Vref that solved it. However, it is not the complete solution, when the protection kicks in, Vref will probably stay as it is, so when it comes back out of protection, I likely will have the same overshoot. I`m thinking of adding a few components to discharge the C of the RC filter when the protection kicks in, then after that, the supply should recover nicely even after a shutdown due to overload.

I was wondering whether you have some special reason not to use the error amp that is built in in the SG3525?

I will optimize the feedback network further as with the bigger caps, the Q factor of the LC filter goes down, so phase margin is increased because of that. I also use it at 60V while it is designed to be used over a range of 40 to 82V. The 60V case should have more phase margin than the 40V case, so I can get some (3-4dB) gain from that as well. On top of that, the mains voltage in Japan is only 100V, giving margin as well (plan to keep that as additional margin).

I`m tempted to try adding an additional zero in the feedback network. The additional zero improves phase margin and allows thus more gain and even more suppression of the 100Hz mains related ripple. It would also reduce the voltage drop after a sudden load change. Don`t want to overdo it. I don`t want to increase the gain at high frequencies (say above 20kHz or so) to avoid that the switching noise can have a negative influence on the feedback loop. So plan to use that 2nd zero (and pole) in a moderate way.

Best regards

Gertjan
  Reply With Quote
Old 19th April 2005, 05:39 AM   #26
Pierre is offline Pierre  France
Banned
 
Join Date: Nov 2004
Location: Paris
Sorry, Ghemink, I told you to put the input of the SG3525 to GND in order to get 50% duty cycle because I usually connect the non inverting input to the reference and use the inverting one to control. So if you tie the inverting input to GND you get full output. I have just realized that the K6 SMPS input is non-inverting. That's why you have to put more than 2 volts there to get full duty-cycle output (ooops).

If you would get +/-130V with that output, I understand that you don't want to make the test, of course!

BTW: How hot does the PSU get with that 650W continous output? Are you using an additional heatsink or only the bracket shown in the webpage? Have you measured efficiency? (I know, difficult with offline supplies without PFC)

Best regards,
Pierre
  Reply With Quote
Old 19th April 2005, 11:35 AM   #27
ghemink is offline ghemink  Netherlands
diyAudio Member
 
Join Date: Jun 2004
Location: Japan
Quote:
Originally posted by Pierre
Sorry, Ghemink, I told you to put the input of the SG3525 to GND in order to get 50% duty cycle because I usually connect the non inverting input to the reference and use the inverting one to control. So if you tie the inverting input to GND you get full output. I have just realized that the K6 SMPS input is non-inverting. That's why you have to put more than 2 volts there to get full duty-cycle output (ooops).

If you would get +/-130V with that output, I understand that you don't want to make the test, of course!

BTW: How hot does the PSU get with that 650W continous output? Are you using an additional heatsink or only the bracket shown in the webpage? Have you measured efficiency? (I know, difficult with offline supplies without PFC)

Best regards,
Pierre

Hi Pierre,

I'm still using that small L profile only. I don't test for long times with large loads because my load resistors would probably burn out before the SMPS would. Of course it needs some heat sinking. Once I finish tweaking the performance. I plan to mount it to a bigger L profile that will be mounted to an aluminum plate (bottom of the case). I hope that is enough cooling since it will probably never need to run at full power for long times. I will use high efficiency class D amps and I never listen to dynamically compressed music, so average power is probably well below 100W. If more heatsinking needed, I will do that of course.

I don't know about the efficiency, I may measure it in the future but it is very low on my priority list. My priority is on performance and the main reason to use a SMPS was to get a regulated supply without the need of too bulky cooling that would be needed for a conventional series regulator.

Best regards

Gertjan
  Reply With Quote
Old 19th April 2005, 01:01 PM   #28
Pierre is offline Pierre  France
Banned
 
Join Date: Nov 2004
Location: Paris
So, if I get it right, in a non regulated SMPS, (always at max. duty cycle), the output inductors could be low valued, for example 5-10uH, with big capacitors, several thousands uF, right?

I think that that's the idea most car audio amplifiers use. Even if some of them are regulated, there is no necessity for a so wide range in duty cycles as K6 SMPS (and others) uses, as the car supply varies only between 11 and 14V aprox, and doesn't need to cope with 100 to 240V to cover worldwide mains. So with a narrower input range, a less varying duty cycle can be used and hence the transformer can be optimized for almost 50% duty cycle, so the output inductors can be reduced to a few uH.

Just thinking in loud...
  Reply With Quote
Old 19th April 2005, 01:48 PM   #29
ghemink is offline ghemink  Netherlands
diyAudio Member
 
Join Date: Jun 2004
Location: Japan
Quote:
Originally posted by Pierre
So, if I get it right, in a non regulated SMPS, (always at max. duty cycle), the output inductors could be low valued, for example 5-10uH, with big capacitors, several thousands uF, right?

I think that that's the idea most car audio amplifiers use. Even if some of them are regulated, there is no necessity for a so wide range in duty cycles as K6 SMPS (and others) uses, as the car supply varies only between 11 and 14V aprox, and doesn't need to cope with 100 to 240V to cover worldwide mains. So with a narrower input range, a less varying duty cycle can be used and hence the transformer can be optimized for almost 50% duty cycle, so the output inductors can be reduced to a few uH.

Just thinking in loud...

You maybe right, I'm not an expert on SMPS. In a car, you probably have a pretty stable battery that is stronger than the 100V mains that I have here in Japan. So regulation may not be needed. I can't answer on whether a small L in combination with a big C is ok in such a case.

I prefer large L and large C :-)

Best regards

Gertjan
  Reply With Quote
Old 20th April 2005, 02:24 PM   #30
ghemink is offline ghemink  Netherlands
diyAudio Member
 
Join Date: Jun 2004
Location: Japan
Quote:
Originally posted by ghemink
This is with the 8200uF caps.


I know which one I prefer :-)

And this one is with the original 0.22uF cap but now resistor R25 changed from 7k to 33k. This increases the gain and reduces the frequency of the first zero needed to keep things stable with the big 8200uF caps. The gain can be increased because the LC filter output essentially falls of with 6dB/Oct due to the low Q factor and due to the series resistance of the cap itself (0.05Ohm). I tried even higher values, but then at high currents (more than 5A), you get a slight oscillating behavior, not heavy, but larger than the 100Hz ripple. Everything stayed clean up to 43k, but when I switched to 56k, I got some weak oscillations at about 250Hz when switching from the 41 Ohm load to the 23.4 Ohm load (from about 3A to 5.5A). With 56k, the 3A load gives no problem at all. For safety I decide to use 33k to have a safe margin for even higher currents.

When I hook up the elna caps I may have to do this exercise again because the elna's probably have a lower series resistance.

Note that the drop due to the switching in the load is now almost invisible (switching from 0.8A to about 3A).
Attached Images
File Type: jpg 8200uf feedback 33k.jpg (90.5 KB, 601 views)
  Reply With Quote

Reply


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
SMPS for class D module NewTesla Class D 40 16th May 2010 07:38 PM
SMPS vs. Class-D. Dag Class D 17 2nd March 2007 05:23 PM
JLH class a with a SMPS?? Possible? JZatopa Solid State 2 27th March 2005 04:47 PM
help with class D amp and SMPS Evan Shultz Class D 9 23rd February 2005 12:36 AM
SMPS and class D parts gearheaddruid Swap Meet 0 31st January 2005 05:08 AM


New To Site? Need Help?

All times are GMT. The time now is 06:52 PM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright 1999-2014 diyAudio

Content Relevant URLs by vBSEO 3.3.2