Hi,
Kenshin, excellent question, silly me, I took them as diodes, I should have looked closer, they must be those totem pole emitters mentioned... buffers.
Lumanauw, I would think not very many. IRF640 is kind of a fat mosfet to drive, and output current of those drivers isn't very high. So it depends on the speed you want to switch at, without having done any math you could possibly get away with 2 in parallel at your current frequency. If you want to parallel them at higher frequencies, just buffer the signals to get more current gain, but it would be easier to get a better mosfet. With a decent mosfet you shouldn't have a great need to parallel them at all.
If you want to learn more about UCD there's a bit of info on it in here:
http://www.diyaudio.com/forums/showthread/t-36852.html
I'm currently starting work on two improved versions over the ones you'll see in that thread, or I believe elsewhere on the forum. I plan on building one of them soon just with p2p.
I still think if you wanted to make your own, and pull out all the stops to do it, that it would cost you more than buying a module from Hypex, but this is more fun and educational.
I'll post schematics when ready.
Why does your oscillate? It will oscillate without the RC network, but not very well, it's pretty ugly.
Frequency with this type is determined by total circuit delay+total phase shift. That includes the phase shift of the LC filter, along with that of the RC lead network.
The frequency for which total phase shift =360 degrees (don't forget the 180 degrees from the inverting output stage counts) is where it oscillates.
Regards
Chris
Kenshin, excellent question, silly me, I took them as diodes, I should have looked closer, they must be those totem pole emitters mentioned... buffers.
Lumanauw, I would think not very many. IRF640 is kind of a fat mosfet to drive, and output current of those drivers isn't very high. So it depends on the speed you want to switch at, without having done any math you could possibly get away with 2 in parallel at your current frequency. If you want to parallel them at higher frequencies, just buffer the signals to get more current gain, but it would be easier to get a better mosfet. With a decent mosfet you shouldn't have a great need to parallel them at all.
If you want to learn more about UCD there's a bit of info on it in here:
http://www.diyaudio.com/forums/showthread/t-36852.html
I'm currently starting work on two improved versions over the ones you'll see in that thread, or I believe elsewhere on the forum. I plan on building one of them soon just with p2p.
I still think if you wanted to make your own, and pull out all the stops to do it, that it would cost you more than buying a module from Hypex, but this is more fun and educational.
I'll post schematics when ready.
Why does your oscillate? It will oscillate without the RC network, but not very well, it's pretty ugly.
Frequency with this type is determined by total circuit delay+total phase shift. That includes the phase shift of the LC filter, along with that of the RC lead network.
The frequency for which total phase shift =360 degrees (don't forget the 180 degrees from the inverting output stage counts) is where it oscillates.
Regards
Chris
lumanauw said:
Hi lumanauw,
Thanks. I am drawing a diagram of a bridge circuit which drives the MOSFETs with two IR2110 (IR2113) chips not unlike that subwoofer version. The new circuit should be able to reproduce full range nicely. In fact, the new circuit uses an LT1016 comparator instead of the 74C14. The result is the almost complete elimination of hysteresis. It also has the added benefit of a higher switching frequency. Also, DC offset is maintained automatically without adjustment.
Switchmode amp output inductors
I've done a lot of work on open-loop all digital power amplifiers (ala TacT Millenium) in the last 10 years and have found the linearity of the output inductor critical for obtaining low THD.
After much experimentation I've not found any core material completely satisfactory for this open-loop application, and the filter inductor I favour presently in my designs is an aircored type which is a 'foil wound' design. (The TacT Millenium also uses a foil-wound aircored output L made by Jensen).
Foil wound inductors can easily be made at home from transformer tape and self-adhesive copper foil, obtainable from RS or Farnell. I used a length of 16mm diam x 25mm long PVC conduit as a former. The overall length of the foil winding is around a meter for L of 8-16uH and the dc resistance is around 20-25mOhm. Running the amp at 250W continuous sinewave into 4 ohms the vertically mounted filter inductors in my PowerDAC 2 amplifier barely reach body temperature.
It's possible to reduce the dc resistance further by using wider tape/foil on a longer former, space permitting. The inductance is dependent on winding tension and must be trimmed to the desired value by progressively measurement-and-cut- from a value higher than that desired. The completed assy is then dunked in transformer varnish.
I originally suspected that I'd find the interwinding capacitance of these inductors would be high and the self-resonant frequency unacceptably low. It was a pleasant surprised to find the f-res at 8.5MHz about 30% higher than a conventional 2-layer aircored inductor of tyhe same L !
If anyone is interested in detailed constructional diagrams for these inductors I can post them as a pdf.
I've done a lot of work on open-loop all digital power amplifiers (ala TacT Millenium) in the last 10 years and have found the linearity of the output inductor critical for obtaining low THD.
After much experimentation I've not found any core material completely satisfactory for this open-loop application, and the filter inductor I favour presently in my designs is an aircored type which is a 'foil wound' design. (The TacT Millenium also uses a foil-wound aircored output L made by Jensen).
Foil wound inductors can easily be made at home from transformer tape and self-adhesive copper foil, obtainable from RS or Farnell. I used a length of 16mm diam x 25mm long PVC conduit as a former. The overall length of the foil winding is around a meter for L of 8-16uH and the dc resistance is around 20-25mOhm. Running the amp at 250W continuous sinewave into 4 ohms the vertically mounted filter inductors in my PowerDAC 2 amplifier barely reach body temperature.
It's possible to reduce the dc resistance further by using wider tape/foil on a longer former, space permitting. The inductance is dependent on winding tension and must be trimmed to the desired value by progressively measurement-and-cut- from a value higher than that desired. The completed assy is then dunked in transformer varnish.
I originally suspected that I'd find the interwinding capacitance of these inductors would be high and the self-resonant frequency unacceptably low. It was a pleasant surprised to find the f-res at 8.5MHz about 30% higher than a conventional 2-layer aircored inductor of tyhe same L !
If anyone is interested in detailed constructional diagrams for these inductors I can post them as a pdf.
I definitely agree that air-cored coils would add the least amaount of distortion by themselves. This can be very important for amps that don't take NFB after the output filter or the ones without any sort of NFB like power -DACs (TacT et al).
For ordinary class-d amps that take NFB form the output filter this advantage may possibly not outweigh the negative effects of the stray fields of air-cored coils.
Edit: one possible exception just "jumped up": discrete-time delta-sigma amps might be less susceptible to such stray-fields and thus could greatly profit from these coils while being quite immune to their negative effects.
Regards
Charles
For ordinary class-d amps that take NFB form the output filter this advantage may possibly not outweigh the negative effects of the stray fields of air-cored coils.
Edit: one possible exception just "jumped up": discrete-time delta-sigma amps might be less susceptible to such stray-fields and thus could greatly profit from these coils while being quite immune to their negative effects.
Regards
Charles
Hi Guys
Here's the constructional info for the foil wound inductors I presently use. I documented this project meticulously because at one stage I contemplated making these PowerDAC amps commercially. But this would have meant extensive 'human interface' and I thought better of it and returned to the liberating process of scribbling notes on scruffy bits of paper and post-it pads.
In earlier prototypes of the PowerDAC amplifiers I used conventional aircore inductors wound on RM10 bobbins, (approx 33-37 t of 1.0mm e.c.w. for 10-12uH with 8-9 turns per layer. But these are only suitable for smaller amps as at 200-250W they fry after a while.
Regards
John Hope
Here's the constructional info for the foil wound inductors I presently use. I documented this project meticulously because at one stage I contemplated making these PowerDAC amps commercially. But this would have meant extensive 'human interface' and I thought better of it and returned to the liberating process of scribbling notes on scruffy bits of paper and post-it pads.
In earlier prototypes of the PowerDAC amplifiers I used conventional aircore inductors wound on RM10 bobbins, (approx 33-37 t of 1.0mm e.c.w. for 10-12uH with 8-9 turns per layer. But these are only suitable for smaller amps as at 200-250W they fry after a while.
Regards
John Hope
John Hope said:Hi Guys
In earlier prototypes of the PowerDAC amplifiers I used conventional aircore inductors wound on RM10 bobbins, (approx 33-37 t of 1.0mm e.c.w. for 10-12uH with 8-9 turns per layer. But these are only suitable for smaller amps as at 200-250W they fry after a while.
Regards
John Hope
The wound foil ones would seem to offer remediation for skin effect losses.
Hello,
Would you happen to have a link or two for suitable copper tape?
I've checked around and all I find is adhesive EMI shielding tape, where the adhesive is said to be conductive. It say's nothing about if the tape is insulated or not, since the adhesive is conductive I'm guessing not.
Regards
Would you happen to have a link or two for suitable copper tape?
I've checked around and all I find is adhesive EMI shielding tape, where the adhesive is said to be conductive. It say's nothing about if the tape is insulated or not, since the adhesive is conductive I'm guessing not.
Regards
Phase_accurate
Yes indeed, you are right. Construction can be a bit fiddly if you approach it like a bull in a china shop, and I'll give some suggestions:
The transformer tape is both insulating and self-adhesive. You roll out a suitable length of this and stick it down onto a flat, smooth surface like a glass table, so that it can be peeled up again without getting into a mess.
The self-adhesive copper tape is, of course, conducting. The adhesive is also conducting but this is not relevant. You roll the copper tape off the roll and stick it down on the transformer tape, taking care not to make bubbles or creases, and to position the copper tape symetrically in the width of the transformer tape. (The copper tape is a few mm less in width than the transformer tape, as can be seen from the picture.)
After fitting the 'start' tab, you peel this copper-plastic sandwich off your flat smooth surface and wind it onto your bobbin, keeping a good tension on the tape. Roll the completed coil across the table under medium hand pressure a few times to even out bumps.
Begin the trim for the desired inductance value in a measure-rolloff-measure iteration and then fit the 'finish' tab .
Fitting the radial end leads is straightforward, whereafter the final assy can be dunked in varnish.
On the constructional sheet I attached there is a materials table which provides RS and Farnell part numbers for the copper foil and transformer tape. You can reference these on the respective vendor websites to get manufacturer data.
Regards
John Hope
Yes indeed, you are right. Construction can be a bit fiddly if you approach it like a bull in a china shop, and I'll give some suggestions:
The transformer tape is both insulating and self-adhesive. You roll out a suitable length of this and stick it down onto a flat, smooth surface like a glass table, so that it can be peeled up again without getting into a mess.
The self-adhesive copper tape is, of course, conducting. The adhesive is also conducting but this is not relevant. You roll the copper tape off the roll and stick it down on the transformer tape, taking care not to make bubbles or creases, and to position the copper tape symetrically in the width of the transformer tape. (The copper tape is a few mm less in width than the transformer tape, as can be seen from the picture.)
After fitting the 'start' tab, you peel this copper-plastic sandwich off your flat smooth surface and wind it onto your bobbin, keeping a good tension on the tape. Roll the completed coil across the table under medium hand pressure a few times to even out bumps.
Begin the trim for the desired inductance value in a measure-rolloff-measure iteration and then fit the 'finish' tab .
Fitting the radial end leads is straightforward, whereafter the final assy can be dunked in varnish.
On the constructional sheet I attached there is a materials table which provides RS and Farnell part numbers for the copper foil and transformer tape. You can reference these on the respective vendor websites to get manufacturer data.
Regards
John Hope
Hi,
You guys are right I had missed/forgotten about the transformer tape!
Thank you for the detailed info. This is actually something I've been looking into the past few weeks it is nice to know it's entirely possible to DIY.
I've also looked at those used for X-overs. Smallest I found was 70uH. Can also have them wound to spec but I doubt they'd share the same reasonable price.
I dont' find the DIY method at all intimidating.
Regards,
Chris
You guys are right I had missed/forgotten about the transformer tape!
Thank you for the detailed info. This is actually something I've been looking into the past few weeks it is nice to know it's entirely possible to DIY.
I've also looked at those used for X-overs. Smallest I found was 70uH. Can also have them wound to spec but I doubt they'd share the same reasonable price.
I dont' find the DIY method at all intimidating.
Regards,
Chris
Thanks for the info. I have a big suspicion on how powerfull SMPS radiation is. Once I track ground with ground that differs by distance only 20cm. (but near smps transformer) It has spikes (caused by the pcb copper tracks picks up pulses from smps radiation?).
How if I put the SMPS transformer and LC filter core with a "can" or wrap them with aluminum foil, then ground them for shielding. Is this way safe to have LC filter next to SMPS transformer?
If not, what is the minimal distance required between smps transformer and LC filter core?
And one more question. If I put LC filter next to SMPS transformer (without any shielding), could it that the classD amp becomes oscilating caused by radiation of SMPS is picked up by LC filter core? Or the radiation is not powerfull enough do that (destructive oscilation at classD power amp)?
Hi Lumanauw
Measuring ground noise with an ordinary scope probe is prone to error because you get noise pickup from the scope probe's ground clip lead which often makes things look twice as bad as they really are. Linear Tech covers this issue in one of their application notes - the LT1074 design manual if I recall.
RFI problems can drive you BANANAS and eradicating RFI once it's raised its ugly head involves a lot of experimentation and countless reworks. In general it's better to try get rid of RFI at source rather than take measures to prevent it getting into sensitive stuff. So let's look at the SMPS:
Screening the SMPS transformer alone will probably not totally prevent RFI. There are several sources of RFI from an SMPS and they can be conductive or radiated or both.
If I were you I would Screen the ENTIRE SMPS, using a grid of small holes for ventilation where you need it. Tinplate is cheap, easy to work with and can readily be soldered. The screen case should have a solid ground connection to chassis ground.
SMPS transformers with an airgap radiate more RFI than those without. If it hasn't got one already, fit a copper foil shield around the OUTSIDE of the transformer, so that it's wound over both the core and the windings. Connect this shield to the main SMPS ground.
I would also put high current common mode power filters on all power leads leaving the SMPS. The Murata BNX002 series are rated up to 50V 10A. If these are not rated enough for you or prove difficult to get you can 'crib' them and make your own with ferrites and caps.
As to your last question, all I can say is that The Law of Conservation of Misery really works. . .
Hope this helps
John Hope
Measuring ground noise with an ordinary scope probe is prone to error because you get noise pickup from the scope probe's ground clip lead which often makes things look twice as bad as they really are. Linear Tech covers this issue in one of their application notes - the LT1074 design manual if I recall.
RFI problems can drive you BANANAS and eradicating RFI once it's raised its ugly head involves a lot of experimentation and countless reworks. In general it's better to try get rid of RFI at source rather than take measures to prevent it getting into sensitive stuff. So let's look at the SMPS:
Screening the SMPS transformer alone will probably not totally prevent RFI. There are several sources of RFI from an SMPS and they can be conductive or radiated or both.
If I were you I would Screen the ENTIRE SMPS, using a grid of small holes for ventilation where you need it. Tinplate is cheap, easy to work with and can readily be soldered. The screen case should have a solid ground connection to chassis ground.
SMPS transformers with an airgap radiate more RFI than those without. If it hasn't got one already, fit a copper foil shield around the OUTSIDE of the transformer, so that it's wound over both the core and the windings. Connect this shield to the main SMPS ground.
I would also put high current common mode power filters on all power leads leaving the SMPS. The Murata BNX002 series are rated up to 50V 10A. If these are not rated enough for you or prove difficult to get you can 'crib' them and make your own with ferrites and caps.
As to your last question, all I can say is that The Law of Conservation of Misery really works. . .
Hope this helps
John Hope
I think you are right. I start to smell a BANANA PIE in my experiment desk
At first I wanted to make use the small dimension needed for classD and SMPS, to make a small but powerfull poweramp.
I never think about the component placement issue, until I smell this BANANA PIE. Layout and component placement is very important for these 2 switching cct, especially when they are on the same board.
Help...help....
Besides the core material problem, the topology seems to have problem.
I'm about to give up with topology on post #18. The gate signals are not square and I cannot raise the switching speed from 29khz.
What is basically wrong with this topology?
Before I'm leaving this topology, I would like to know wheter this topology can be make better like :
1. How to make clean square gate signal with this topology? The MC's are totem pole. Since the topology itself is very "analog amplifier" like, I think it's hard to make clean square with this analog topology. The differential and VAS always driven in linear mode, giving big rise and fall time that cannot make clean square
2. How to raise self-oscilating frequency? Nothing I do can raise the frequency from 29khz. I can make lower but cannot make higher frequency. I even remove all the caps that lower frequency, and the frequency doesn't rise any.
I suspect the mA of the totem pole is not big enough to charge the mosfets. How to make bigger mA available for driving gate of mosfets with this topology (without using mosfet driver IC?)
Besides the core material problem, the topology seems to have problem.
I'm about to give up with topology on post #18. The gate signals are not square and I cannot raise the switching speed from 29khz.
What is basically wrong with this topology?
Before I'm leaving this topology, I would like to know wheter this topology can be make better like :
1. How to make clean square gate signal with this topology? The MC's are totem pole. Since the topology itself is very "analog amplifier" like, I think it's hard to make clean square with this analog topology. The differential and VAS always driven in linear mode, giving big rise and fall time that cannot make clean square
2. How to raise self-oscilating frequency? Nothing I do can raise the frequency from 29khz. I can make lower but cannot make higher frequency. I even remove all the caps that lower frequency, and the frequency doesn't rise any.
I suspect the mA of the totem pole is not big enough to charge the mosfets. How to make bigger mA available for driving gate of mosfets with this topology (without using mosfet driver IC?)
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.