diyAudio

diyAudio (http://www.diyaudio.com/forums/)
-   Class D (http://www.diyaudio.com/forums/class-d/)
-   -   hf ringing after inductor on positive halfwave. (http://www.diyaudio.com/forums/class-d/221047-hf-ringing-after-inductor-positive-halfwave.html)

switchleg 6th October 2012 02:32 PM

hf ringing after inductor on positive halfwave.
 
4 Attachment(s)
I am getting what appears to be 28mhz ringing on the output of my filter (28uh, 680nf). I have been searching threads and still not sure why i have ringing on the output but very little on the power stage. Just wondering if anyone has any suggestions on how to resolve this.


The yellow line is taken at the output terminals (after filter) and the blue line is taken btw the mosfets (before filter).

Salas 6th October 2012 03:04 PM

Welcome to DIYA. Smaller images next time please. 1200X900 will do. I resized and attached for this time. :captain:

switchleg 6th October 2012 03:16 PM

ok will do, thanks.

darkfenriz 7th October 2012 02:36 PM

Hi

Good to see the measurement setup.
Such arrangement is not very good for measement of dozens of MHz signals, try reducing the are of a probe scope and ground tip to minimum.
A hint: I solder a piece of UTP directly to what I measure and connect the probe several centimeters away with ground tip twisted around the probe, that's be e.g. at switching node and most adjustent ground reference.
Secondly, can you show both signals at the same scale? Keep in mind that the LC filter may be out of any attentuation at 30MHz.

Pafi 7th October 2012 03:40 PM

Parasitic (parallel) capacitance of inductor, parasitic (series) inductance of capacitor.

On the first picture rise time seems to be 50 ns, but on the second 20 ns. I guess there is some problem with the measurements.

High amplitude square on one channel of the scope can disturb the sensitive other channel.

Ground loop.

Agilent Technical Forums : Noise problem with DSO1014a? ...

switchleg 7th October 2012 06:45 PM

honestly i wasn't being careful with the probes after a bunch of testing i was getting lazy :P. but i will try twisting the ground around the probe! and use a single ground point for both probes (as i usually do!). also when i only use one probe at the output terminals i also get the same waveform... i will try to post more pictures after dinner tonight. (maybe some schematics to make things easier)

not sure if its visible on these images but the power stage does have a slight ring maybe the same amplitude but its also inverted to the signal after the inductor.


this is my fourth board with the same components but new layout. last builds where ok with only a very very small ringing around 5mhz but nothing appeared at the output. noise was ok but not as good as my class b or a amps. also had a dc offset around 80mv with -+50v supply:(. after a lot of time designing this board i was sure this layout was much better :/

switchleg 7th October 2012 07:03 PM

3 Attachment(s)
i have a feeling its the lc filter and lack of attenuation at such frequencies...

i am posting pictures of the power stage pcb and schematics. the red circles are the snubbers and decoupling capacitors the blue line is the only trace that is on the other side of the board (the rest being a ground plane). also the toroid is a t106-2 laying flat on top of the ground plane. the filter cap is a cbb metal film cap at 680nf 600v. also the decoupling caps are 1206 format 100nf XR7 100v. the snubber caps are 1206 format 100pf NPO 1000v.

oh i also changed bootstrap diodes from mur120 to ES1D on this design....

using irf530n only for testing...

Eva 8th October 2012 02:32 AM

The falling edge is slow because it's soft switched, it's inductor current what charges parasitic capacitance rather than the opposite FET turning on. This is what happens in the rising edge, and the high di/dt excites the 28Mhz resonance that is made of L and C elements spread through the power stage layout. Once the resonance starts, you get AC voltages even across ground tracks, and some energy is radiated too, so the resonance appears "everywhere" unless you use the proper measurement techniques (for example: direct ground to scope probe ring, without relying on the classic alligator clip ground lead, which acts as a receiving antenna).

The C usually comes from the FET dies themselves, drain to source capacitance (few dozen to few hundred pF, changes with Vd-s, should be considered at proper Vd-s), although other capacitances connected to this node can alter the main resonance. The L comes from series elements: PCB traces from FETs to supply capacitors and the series inductance of these parts (in the order of a few dozen nH). The R is low enough to get an underdamped system. Personally I prefer to shift these resonances as high as possible in frequency because copper is more resistive at 100Mhz than at 30Mhz, getting more self damping to start with. Also, the snubbers needed to damp lower frequencies waste more power.

Things to do to improve it:
- Get good switching timing, the two halves of the output stage have not equal delays or dead times, this is important to keep good audio signal integrity.
- Try to move up the 28Mhz resonance by reducing layout and component lead inductances. It usually comes at ~1nH per mm of lead length/spacing for capacitors and transistor packages. Use ground plane, this reduces the ~1nH per mm of "unshielded" PCB tracks to much less, and makes tracks more lossy and leass inductive at RF.
- Use snubbers to damp the resonance. The right combination of C, R and parasitic L, just tuned to the main resonance, is what does the trick best. Fr=1/(2*pi*sqr(L*C))

gootee 8th October 2012 03:42 AM

Here is a practical procedure for finding the optimal R for the snubber you need, and the C to go with it:

http://www.diyaudio.com/forums/power...ml#post2828689

switchleg 8th October 2012 04:38 AM

well the majority of the board is covered with a ground plane on the other side not shown(including the power stage and gate driver). with the 1nh per mm is starting to make sense if i add up the length of all the leads coming off both fets + the trace btw the two. also i wrote down the wrong frequency, the current ringing is at 35mhz (28ns). if i can manage to shorten the leads on the fets (and use better fets with lower COSS) i might be able to get upto 50mhz or so. then the only other option i have is tune the snubber to the ringing frequency and hope for the best. correct?


All times are GMT. The time now is 07:13 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