UcD like topology amp.

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
It does. (Depends on the impedance around switching freq, not the audio impedance.)
Thanks for your reply Pafi. Sorry for the confusion. However the situation is same with or without load.
I suspect from the output inductor, Ferroxcube TN 26/11-3c20-a113 wound 17sp gives 31uH (remaining 4 of them TN 23/7.5-3c20-a65, 22sp). I measured next channel and they are in same value. However, the next channel do not have same issue. Do you think it is still output impedance/zobel network problem.
BR
 
Last edited:
However the situation is same with or without load.

Yes, because its impedance is too high at fsw compared to the impedance of Coutput. I bet your load is not inductanceless, is it?

Obviously fsw does depend on load impedance in UcD.

Do you think it is still output impedance/zobel network problem.

Im not sure if it's a problem or not. Lets say fenomenon. I wont tell what is the problem actually. No time, no access to your circuit. But you can't exclude the possibility of such a trivial thing. Basic operation first, side-effects (like parasitic feedback) second! If you really sure about that not all channels behave equally, then side-effects are more probable.

Parasitic feedback is probably magnetic/inductive at this low impedance.
 
Yes, because its impedance is too high at fsw compared to the impedance of Coutput. I bet your load is not inductanceless, is it?

Obviously fsw does depend on load impedance in UcD.

Ok, Load is a DVC MA 12" sub bass speaker. Hence it is inductive. If i add a ferrite bead between the two caps as shown below, it makes again a difference. The sw goes up 420-430K residual noise decreases to 200mV pk-pk. The strange thing is that, i am not changing the position of the FB it is between the L1 and L2. Then, side effects? do not know yet.

Thanks
 

Attachments

  • Bead.jpg
    Bead.jpg
    49 KB · Views: 1,043
UcD is a phase shift oscillator. If you are trying to use a zobel on it, you are not understanding how and why it works, you are just copying!! (In a creative but not thoughtful way). Note that UcD was developed to avoid zobels on the output!!

On the other hand, anything that changes the phase shift from comparator output node to the feedback node will affect self oscillating frequency. This includes: output capacitor parasitic and layout inductance, loads that don't become high impedance above 20khz (ie: not a speaker), adding a cap with wires (it's a LRC), adding a cap even without wires (it's again a LRC with lower L), adding a RC (which changes phase response of output filter). Getting UcD right requires some RF thinking, the actual frequency response of everything in the power path matters (unlike in clocked amps, triangle wave modulators were invented to cope with "dirty" power stages at the expense of linearity and freq. response).
 
Last edited:
UcD is a phase shift oscillator. If you are trying to use a zobel on it, you are not understanding how and why it works, you are just copying!! (In a creative but not thoughtful way). Note that UcD was developed to avoid zobels on the output!!

On the other hand, anything that changes the phase shift from comparator output node to the feedback node will affect self oscillating frequency. This includes: output capacitor parasitic and layout inductance, loads that don't become high impedance above 20khz (ie: not a speaker), adding a cap with wires (it's a LRC), adding a cap even without wires (it's again a LRC with lower L), adding a RC (which changes phase response of output filter). Getting UcD right requires some RF thinking, the actual frequency response of everything in the power path matters (unlike in clocked amps, triangle wave modulators were invented to cope with "dirty" power stages at the expense of linearity and freq. response).

Thanks for the reply Eva,

However, there should be some misunderstanding about Zobel network. Because I did say it is not due to the zobel network. I am not copying someones design (ie. Bender.ru). If i would do so, i would completely copy his desing. The actual thing is Bender.ru and I thought the same thing at same time and I was already desinging the circuit with MAX1016 while he was doing with LM311( Logic is same and if something logical two or more people can think the same thing at same time.). However he posted complete desing on thread and i gave up working on it.
Anyways, We agree in some points and disagree in some others. I said the sw frequency of UCD does not depend on load in one of the previous post. However, i did not mean in the understood manner. The sw frequency of UCD does depend on load but not as much as i faced with actually. In my case i kept the inductor same and put an additional C to change the filter response and get higher self osc frequency. This is true logic and as is the case not true always. Hence, the sw frequency of UCD does depend on the Hloop and this is effected by Rlead, Clead, Rin, Rfb and Lout Cout+ propagation delay from input to output. We know that the typical propagation delay of IRS20957s is 90ns and LM311 has an 200ns of response time whereas the original design discrete UCD amp has an 210ns propagation delay. Therefore, one should expect a different sw frequency with the same component values used in discrete design. The interesting thing is the self osc frequency difference between the channels. I am not using %1 tolerence resistors or capacitors. But, one can not expect so much difference from channel to channel. I tried 220,330,470,680,1K and 2.2K for Rlead and believe or not i always get the same sw frequency for the LF oscillating channel, which is around 300kHz. Then i changed Ri from 2K2 to 1K5 and get 330kHz and leave it like this. But, there should be something wrong with this channel because all the others running around 410kHz with the Rfb=22K, Ri=2K2(should be changed into 1K5 to ensure same gain), Rlead=2K2, Clead=220p(outputfilter inductor always same inductance =30uH and the Cout=1u).
I also swapt the Lout from another channel and the sw did not changed. Hence, the IRS20957 or LM311 can be the reason or the PCB itself. What can be else?

Best wishes.
 
Hi All,

After some measurements, I found the reason why one channel oscillating at a lower frequency. The reason is the LM311 itself. Interestingly it is ~100ns slow than the others. And hence replacing it with a new one solve all the problems. Now all the channels running around 400-430kHz. The idling loss for SMPS itself is around 2Ws(none of the channels connected) and the idling loss for two Channels around 5Ws. It looks the channels not fighting to eachother in the sw frequency means. So the amp is fine just now and i am going on to complete it up to final version. I have to build output protection relay board and fit it into the place. Then I can start tests at idle, mid and full power.

Chreers
 
Hi all.
Here is light version amp docs & pcb.
Please connect me if you find bugs.

Cakeamp.zip
(link will be available for 30 days)

http://rapidshare.com/files/107380802/Cakeamp.zip.html
(i dont know how long it will be available ;) )


Bender
The Specs, and the compact looks like, more than for a DIYers. Thanks for sharing.

If your tool can generate BOM with footprints of components. Can I request you to post it here. So that I can check if its available for me to buy and built one for myself. I am very much eager to hear this.
 
Last situation of my amp

Hi all,

I think there are several questions around and there are many others similar to the amp discussed here. Anyway, I started to do the amp below for a very long time ago. And now it went into a good PCB. The entire PCB is 105um Cu thickness and i had it made from a China manufacturer (hell, it was a difficult process to get them). The photos below are nearly completed amp but, still 3 auxiliary PCB are missing. They will be available soon.

I hope you all like it.
Best wishes
 

Attachments

  • 1.jpg
    1.jpg
    1 MB · Views: 1,040
  • 2.jpg
    2.jpg
    533.2 KB · Views: 982
  • 3.jpg
    3.jpg
    522.1 KB · Views: 924
  • 4.jpg
    4.jpg
    618.1 KB · Views: 891
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.