One possibility: You have built a magnetic loop antenna with the probe ground-lead that picks up switching noise:
Regards
Charles
Charles
Welcome back. We have been doing some furious brainstorming.
I see great promise in the self-oscillating circuit. I plan to name it either "The Jamp" or "The N-champ" I like the second because using the 6N137, there is theoretically no limit to the output power attainable, up to the isolation voltage rating of the 6N137 and the current and voltge rating of the mosfets--not that I would build one that does more than 1.5kW peak. 😉
Welcome back. We have been doing some furious brainstorming.
I see great promise in the self-oscillating circuit. I plan to name it either "The Jamp" or "The N-champ" I like the second because using the 6N137, there is theoretically no limit to the output power attainable, up to the isolation voltage rating of the 6N137 and the current and voltge rating of the mosfets--not that I would build one that does more than 1.5kW peak. 😉
Sergio
It is a very good idea to LPF the input to the amp. Non switching circuits also do it because it lowers distortion from any noise from all sources. I think the noise could be the magnetcally coupled noise into the scope wires too.
It is a very good idea to LPF the input to the amp. Non switching circuits also do it because it lowers distortion from any noise from all sources. I think the noise could be the magnetcally coupled noise into the scope wires too.
Hi IVX
Did you buffer the outputs of your IR2110? That is one of nice features of the N-champ. It uses low gate threshold voltage mosfets configured as source followers as buffers without the penalty of drawing any more net current from the control power supply. Another benefit is that as the voltage reaches its switching limit, it softly settles into place, reducing ringing and providing damping. Another feature of the N-champ is the use of the 6N137 for level shifting the signal to the IR2110/3.
Did you buffer the outputs of your IR2110? That is one of nice features of the N-champ. It uses low gate threshold voltage mosfets configured as source followers as buffers without the penalty of drawing any more net current from the control power supply. Another benefit is that as the voltage reaches its switching limit, it softly settles into place, reducing ringing and providing damping. Another feature of the N-champ is the use of the 6N137 for level shifting the signal to the IR2110/3.
SMD Not
Jaka Racman
Thanks for your idea. But I looked up that LT1394 comparator, and unfortunately it is only available in SMD. The only SMD I will use is the IRF7343 beause I can solder it directly onto the power supply pins of the IR2113 for buffering purposes. SMD, in my opinion is suitable only for machines and robots, its size and awkwardness is not suitable for human physiology. Time to get down off my soapbox again.
Jaka Racman
Thanks for your idea. But I looked up that LT1394 comparator, and unfortunately it is only available in SMD. The only SMD I will use is the IRF7343 beause I can solder it directly onto the power supply pins of the IR2113 for buffering purposes. SMD, in my opinion is suitable only for machines and robots, its size and awkwardness is not suitable for human physiology. Time to get down off my soapbox again.
Subwo1,
I didn't know your supply rails were 90V, my simulation was with 45V. When i tried 90V I also found some ringing on start of 80V square pulse. My solution would be lower filter inductance to 50uH and change filter capacitor to 20uF. In my opinion it is good to have some (2 to 5A) ripple on filter inductor since it helps to achieve zero voltage switching on output FETs. Switching signal is much cleaner this way. Regarding use of LM6172, you might have problems with overload recovery time, when opamp used as a comparator recovers from saturation. But maybe you could use U1b as 10x gain stage in front of a real comparator. That would relax comparator input requierments by an order of magnitude.
Regarding SMD, I use headband magnifier similar to optivisor and soldering is very comfortable. I can use it for several hours without fatigue, in fact I would not solder through hole parts without it.
Sergio,
welcome to the club of SMPS problems. What you see is definitely magneticaly coupled interference. Try LP filter at the input (classic orientation, from input first resistor and then cap, I suppose you have some resistor from input of error amplifier). Then get rid of probe ground wire and measure input signal directly with naked probe tip with grounding sleve at the probe top touching ground plane. I think you will be a little surprised.
Could you post a link to your output filter choke? If it is not magnetically closed design it might radiate all over the place. I also do not remember if you constructed your output filter on PCB. If this is some P2P design with long wires it is also prone to radiation. The idea is not to use as short wires as possible, but to reduce loop area as much as possible, even when this means longer wires or tracks.
Best regards, Jaka Racman
I didn't know your supply rails were 90V, my simulation was with 45V. When i tried 90V I also found some ringing on start of 80V square pulse. My solution would be lower filter inductance to 50uH and change filter capacitor to 20uF. In my opinion it is good to have some (2 to 5A) ripple on filter inductor since it helps to achieve zero voltage switching on output FETs. Switching signal is much cleaner this way. Regarding use of LM6172, you might have problems with overload recovery time, when opamp used as a comparator recovers from saturation. But maybe you could use U1b as 10x gain stage in front of a real comparator. That would relax comparator input requierments by an order of magnitude.
Regarding SMD, I use headband magnifier similar to optivisor and soldering is very comfortable. I can use it for several hours without fatigue, in fact I would not solder through hole parts without it.
Sergio,
welcome to the club of SMPS problems. What you see is definitely magneticaly coupled interference. Try LP filter at the input (classic orientation, from input first resistor and then cap, I suppose you have some resistor from input of error amplifier). Then get rid of probe ground wire and measure input signal directly with naked probe tip with grounding sleve at the probe top touching ground plane. I think you will be a little surprised.
Could you post a link to your output filter choke? If it is not magnetically closed design it might radiate all over the place. I also do not remember if you constructed your output filter on PCB. If this is some P2P design with long wires it is also prone to radiation. The idea is not to use as short wires as possible, but to reduce loop area as much as possible, even when this means longer wires or tracks.
Best regards, Jaka Racman
SMD Revisited
Jaka Racman,
Thanks for those ideas. OOh, 50uH, that is small, feeding 20uF, large, large amounts of power consumed. Efficiency is the number one thing I am after. Reproduction quality would be second. I do know what you mean about it helping switching though. The idea of using the op amp to drive the comparator is very well considered. That is what I may actually do if necessary.
When I solder, I use strong reading glasses. But, the parts are just too hard to manipulate (SMD). Breadboarding SMD is virtually a total waste of time. P2p wiring is how I do things, and it is just not possible with SMD's. I guess you can see I qualify as a hobbyist, not a professional.
After working with SMD in the repair industry for so long, I hate it like the plague.
Jaka Racman,
Thanks for those ideas. OOh, 50uH, that is small, feeding 20uF, large, large amounts of power consumed. Efficiency is the number one thing I am after. Reproduction quality would be second. I do know what you mean about it helping switching though. The idea of using the op amp to drive the comparator is very well considered. That is what I may actually do if necessary.
When I solder, I use strong reading glasses. But, the parts are just too hard to manipulate (SMD). Breadboarding SMD is virtually a total waste of time. P2p wiring is how I do things, and it is just not possible with SMD's. I guess you can see I qualify as a hobbyist, not a professional.
After working with SMD in the repair industry for so long, I hate it like the plague.Hi subwo1,
could you elaborate a little about higher efficiency when using large L and smaller C versus smaller L and large C. Right now I can see no other reason except larger loses on capacitor ESR.
Best regards, Jaka Racman
p.s. Breeadbording SMD is definitely nosense, I agree. But there were occasions when I had to solder wires on SMD chips with 25mil pitch to test them. In such cases I use blank (no copper) board and glue chips backwards. I then connect with .2mm low temperature magnet wire.
could you elaborate a little about higher efficiency when using large L and smaller C versus smaller L and large C. Right now I can see no other reason except larger loses on capacitor ESR.
Best regards, Jaka Racman
p.s. Breeadbording SMD is definitely nosense, I agree. But there were occasions when I had to solder wires on SMD chips with 25mil pitch to test them. In such cases I use blank (no copper) board and glue chips backwards. I then connect with .2mm low temperature magnet wire.
Hi Jaka Racman,
I was thinking the same. An ideal LC tank circuit is lossless. But I think there are factors like RDS-on of the mosfets, and the loss in the mosfet during the transition between high and low states. And maybe EMF generation, especially from any circuit paths. Even though the inductance speeds up the transition, the cost is forcing current through the channel during its highly resistive transition phase. Also, with a 100khz square wave I wonder how much loss there is due to skin effect of the wire. That supposedly is why some transformer windings are many small guage wires instead of one thicker one. So I avoid those problems by using a larger choke with higher impedance. Some earlier power supply circuits I made suffered from big losses not only in the filter capacitor, but also had severe problems with overheating filter inductors. The inductors were even air core. I learned how critical the way the choke is used at that time. 😱
That is an excellent idea about gluing the SMD upside down and soldering fine wires onto the pins. One thing though is that those wires cannot carry much current. My main problem is that my soldering iron is never in good enough shape to do such fine work. I always forgot to turn it off and kept burning up tips and heating elements. So, I decided it was better for me to use a piece of solid 8ga wire and clip off a little to reform the tip when necessary. Aslo, I had problems with mess from having to use extra flux.
I was thinking the same. An ideal LC tank circuit is lossless. But I think there are factors like RDS-on of the mosfets, and the loss in the mosfet during the transition between high and low states. And maybe EMF generation, especially from any circuit paths. Even though the inductance speeds up the transition, the cost is forcing current through the channel during its highly resistive transition phase. Also, with a 100khz square wave I wonder how much loss there is due to skin effect of the wire. That supposedly is why some transformer windings are many small guage wires instead of one thicker one. So I avoid those problems by using a larger choke with higher impedance. Some earlier power supply circuits I made suffered from big losses not only in the filter capacitor, but also had severe problems with overheating filter inductors. The inductors were even air core. I learned how critical the way the choke is used at that time. 😱
That is an excellent idea about gluing the SMD upside down and soldering fine wires onto the pins. One thing though is that those wires cannot carry much current. My main problem is that my soldering iron is never in good enough shape to do such fine work. I always forgot to turn it off and kept burning up tips and heating elements. So, I decided it was better for me to use a piece of solid 8ga wire and clip off a little to reform the tip when necessary. Aslo, I had problems with mess from having to use extra flux.
Jaka,
Thanks for your advices. Yesterday I included the LPF and take a little more care about the grounding of the tip of the probe and thinks did improve.
Now I can consider the amplifier is working. I include the schematics in PDF format. It is the only thing I can do after all, as the design has been improved a lot since its first steps thanks to the help of many people in this forum. Thanks to all!
If anyone is interested in the board also, please tell me and I will post the design in Protel format.
I plan to make 6 copys of the board to build a multichannel 5.1 amplifier. The input selector (2x5.1 channels plus 2xstereo) and the volume controls will be controlled digitally by a microcontroller using digital potentiometers and integrated multiplexers.
Another issue is the circuit to control the power-up of the transformer, as it is relatively large (350VA) and the filter capacitors are very big (100.000uF/63V), so I have to include a ballast resistor bypassed after a while by a relay.
Best regards.
Sergio
Thanks for your advices. Yesterday I included the LPF and take a little more care about the grounding of the tip of the probe and thinks did improve.
Now I can consider the amplifier is working. I include the schematics in PDF format. It is the only thing I can do after all, as the design has been improved a lot since its first steps thanks to the help of many people in this forum. Thanks to all!
If anyone is interested in the board also, please tell me and I will post the design in Protel format.
I plan to make 6 copys of the board to build a multichannel 5.1 amplifier. The input selector (2x5.1 channels plus 2xstereo) and the volume controls will be controlled digitally by a microcontroller using digital potentiometers and integrated multiplexers.
Another issue is the circuit to control the power-up of the transformer, as it is relatively large (350VA) and the filter capacitors are very big (100.000uF/63V), so I have to include a ballast resistor bypassed after a while by a relay.
Best regards.
Sergio
Hi Sergio
Congrats to your success !
But one burning question remains: HOW DOES IT SOUND ??????
Regards
Charles
Congrats to your success !
But one burning question remains: HOW DOES IT SOUND ??????
Regards
Charles
Aha!
Well, I must say that it has only been tested with a not-so good loudspeaker, but anyway sounds very well. I am waiting to receive a new woofer for a much better 3-way enclosure that I have, and that will be the definitive test.
I also have only a 3300 uF per rail capacitor, and that obviously degrades the low frequency response at high power levels.
Referring to the noise, I can say that, although I haven't measured it (only the 250KHz ripple: about 60mV rms), you must put your ear right into the tweeter to hear something. I am very happy with this figure, too.
If you think that I can do some representative measurements with my equipment (multimeters, waveform generator and 100 MHz oscilloscope), please tell me and I will give you some real numbers.
Thanks, Charles. Your help has been specially useful.
Sergio
Well, I must say that it has only been tested with a not-so good loudspeaker, but anyway sounds very well. I am waiting to receive a new woofer for a much better 3-way enclosure that I have, and that will be the definitive test.
I also have only a 3300 uF per rail capacitor, and that obviously degrades the low frequency response at high power levels.
Referring to the noise, I can say that, although I haven't measured it (only the 250KHz ripple: about 60mV rms), you must put your ear right into the tweeter to hear something. I am very happy with this figure, too.
If you think that I can do some representative measurements with my equipment (multimeters, waveform generator and 100 MHz oscilloscope), please tell me and I will give you some real numbers.
Thanks, Charles. Your help has been specially useful.
Sergio
Sergio,
good luck with your project. Maybe you would like to look at US patent 6046636. It is from Bruno Putzeys, head of Philips lab that developed much acclaimed (but little heard) UCD amplifier. The patent deals especially with grounding and power supply distribution in multi channel switching amplifiers in order to minimize mutual interference.
Best regards, Jaka Racman
good luck with your project. Maybe you would like to look at US patent 6046636. It is from Bruno Putzeys, head of Philips lab that developed much acclaimed (but little heard) UCD amplifier. The patent deals especially with grounding and power supply distribution in multi channel switching amplifiers in order to minimize mutual interference.
Best regards, Jaka Racman
Thanks, Jaka.
Please, can you provide me with a link for the search of that patent? It could be very useful for me.
Best regards.
Please, can you provide me with a link for the search of that patent? It could be very useful for me.
Best regards.
Hi ssanmore:
Here is the link:
http://patft.uspto.gov/netacgi/nph-...0&s1=6046636.WKU.&OS=PN/6046636&RS=PN/6046636
Pretty long link 🙂
Koldby
Here is the link:
http://patft.uspto.gov/netacgi/nph-...0&s1=6046636.WKU.&OS=PN/6046636&RS=PN/6046636
Pretty long link 🙂
Koldby
Thanks, Charles. Much better.
By the way, I am thinking in trying to get the feedback at the very output of the amplifier, mainly to make the frequency response independent on the load. Do you think it may work? Just for curiosity ;-)
Sergio
By the way, I am thinking in trying to get the feedback at the very output of the amplifier, mainly to make the frequency response independent on the load. Do you think it may work? Just for curiosity ;-)
Sergio
Feedback from the filter is possible but that's exactly where it is getting tough!
BTW you can't make the behaviour fully load-independant.
You will achieve flatter response, you can iron-out inductor-related THD and you can achieve better transient response. But it comes at the cost of stability problems with extreme loads. The excellent stability (with its intrinsic poor frequency response) is one of the advantages of feedback takeoff before the output filter.
Regards
Charles
BTW you can't make the behaviour fully load-independant.
You will achieve flatter response, you can iron-out inductor-related THD and you can achieve better transient response. But it comes at the cost of stability problems with extreme loads. The excellent stability (with its intrinsic poor frequency response) is one of the advantages of feedback takeoff before the output filter.
Regards
Charles
OK, charles, give me some clues, please...
What if I try with the circuit just as it is now but connecting the feedback resistor to the speaker output?
(Of course, you can tell me... just try it, but I am curious and I won't be able to try until this weeked 🙁 )
Best regards
What if I try with the circuit just as it is now but connecting the feedback resistor to the speaker output?
(Of course, you can tell me... just try it, but I am curious and I won't be able to try until this weeked 🙁 )
Best regards
ssanmor.
If you can´t read the pictures, you do not hav a TIFF viewer in your browser.
Find one here:
http://www.uspto.gov/patft/help/images.htm#req
Koldby
If you can´t read the pictures, you do not hav a TIFF viewer in your browser.
Find one here:
http://www.uspto.gov/patft/help/images.htm#req
Koldby
- Status
- Not open for further replies.