Yet another N-Channel Amplifier

[efren sabio]

Mike,

Thanks for your reply on cross-conduction.

Efren

 

[keypunch]

Thanks for Definition

hi efren !

"Cross conduction" is something evil...
It is related to speed of transistors, if one transistor opens faster than
the complement closes, you suddenly get a big current through both
transistors, as both transistor are open at the same time.
This is mainly a classB (AB) problem, as transistors have to close/open.
Switch on time can be shorter than switch off time.
This results in very high currents for high frequencies and can also
create crossoverdistortions. These currents can reach values high
enough to burn up an amplifier.

Mike

Mike,

Thanks very much for the definition of what is crosss conduction.

Does cross conduction boil down to TD (on) and TD (off) of output device pairs? If so, in a N/P-Channel design usually the pairs are not very complementary with respect to many of their specs, TD being one of those specs of course. Is this cross conduction matter strickly a N-Channel topology issue or can it also be in N/P-Channel topologies? If yes, same, lessor or to greater extent?


Regards,

John L. Males
Willowdale, Ontario
Canada
09 January 2004 10:52

 

[richie00boy]

It can happen with transistors or MOSFETs, and in complementary or quasi designs. Ton and Toff can give an indication, but these parameters are really for switching, and other factors such as drive symmetry and capacitance/charge are more relevant to audio class-b amps.

 

[keypunch]

It can happen with transistors or MOSFETs, and in complementary or quasi designs. Ton and Toff can give an indication, but these parameters are really for switching, and other factors such as drive symmetry and capacitance/charge are more relevant to audio class-b amps.

Hi richie00boy,

Not that I am any EE type or even EE savvy, but your point can happen in basically any design or device type makes sense.

You indicate "drive symmetryand capacitance/charge are more relevant to audio class-b amps". I am not questioning your knowledge, I do have questions to enable me to understand better. I was keen on the N-Channel design topogogy for many reasons related to avoiding differences in complementary output device pair issues. When I found the Holton design I really liked it as it seem more symentical, if not almost, than any othr N-Channel design I had seen. I am trying to read and narrow what candidates for an amp I will build. This cross cross conduction issue for me seems to be of some concern as parly related to device/parts values coupled with board layout.

As result I have some questions. No details required, but questions for yes/no like simple reply are:

1) Can Td(on) and Td(off) ratios of output device be a good indication of possible cross conduction issues?

2) Can Td(on) and Td(off) ratios of driver device be a good indication of possible cross conduction issues?

3) Is it not basically a switching mode of operation in the "B" mode of AB class amp, therefore "switching" related Td's not be larger determining factor?

4) I assume given what Class A is about there is no switching issues to be concerned about in Calss A part of operation, and no pure Class A only amp has to be concerned about cross conduction?

5) Is capacitance you make reference to Ciss or Coss?

6) Is shorter charge time better?

7) Given the nature of an N-Channel amp being all same output devices, how is it that charge time is factor when all the output devices are same, therefore charge times should be close? Or are these otehr not as mentioned items that should be better matched due to production manufacturing tolerances?

8) Some mention to effect that cross conduction seems not to be issue when single pair of output devices are used. Just luck or there is good reason for this? If true, then could it be implied that cross conduction is realted to paralleling of output devices?


Regards,

John L. Males
Willowdale, Ontario
Canada
09 January 2005 12:22

 

[jleaman]

Now I have done the first real test with +-30V. But the result was not what I expected. I got oscillation at 10MHz, about 100mV peak-peak. Didn't have time to test any modifications today, so it has to wait untill next week.
Here is a picture of the schematic I used for testing:

An externally hosted image should be here but it was not working when we last tested it.

And here is the amplifier:

An externally hosted image should be here but it was not working when we last tested it.

An externally hosted image should be here but it was not working when we last tested it.

-Pelle

Can i have the overlay for this board layout ?

 

[Tekko]

How does it sound ? Very nice boards, jealous

 

[richie00boy]

1) Can Td(on) and Td(off) ratios of output device be a good indication of possible cross conduction issues?

Not really, although if they are excessive they may hint that the device has high charge/capacitance issues. See answer to 3).

2) Can Td(on) and Td(off) ratios of driver device be a good indication of possible cross conduction issues?

Same applies as 1).

3) Is it not basically a switching mode of operation in the "B" mode of AB class amp, therefore "switching" related Td's not be larger determining factor?

No. Switching refers to 'hard' i.e. fully on or fully off use, changing state in the shortest possible time, and thus is pretty irrelevant to audio class-b amplifiers.

4) I assume given what Class A is about there is no switching issues to be concerned about in Calss A part of operation, and no pure Class A only amp has to be concerned about cross conduction?

No. See answer to 3). Although in class-a the problem may be slighly less exaggerated as the devices do not swing as much.

5) Is capacitance you make reference to Ciss or Coss?

Ciss.

6) Is shorter charge time better?

Yes. Although rather than charge time, you are more likely to find things stated as the amount of charge. Obviously the smaller the amount the faster/less current will be required.

7) Given the nature of an N-Channel amp being all same output devices, how is it that charge time is factor when all the output devices are same, therefore charge times should be close? Or are these otehr not as mentioned items that should be better matched due to production manufacturing tolerances?

Because some amps -- especially quasi-complementary with their different driver scheme for upper and lower -- do not have symetrical source/sink current ability in the driver stage.

8) Some mention to effect that cross conduction seems not to be issue when single pair of output devices are used. Just luck or there is good reason for this? If true, then could it be implied that cross conduction is realted to paralleling of output devices?

Mostly luck/sound design, but because the amount of charge the driver stage has to sink/source goes up in proportion to the number of paralleled output devices, then issues are less likely with single pairs and typical VAS/driver currents.

 

[keypunch]

Thanks so much for answers ... just few follow ups if may

richie00boy,

Thanks so much for your answers. I have a few follow ups if I may? See below:

Ciss.

So lower the Ciss the better not just from quality of high's reproduced perspective, but also from a greater opportunity of cross conduction?

Yes. Although rather than charge time, you are more likely to find things stated as the amount of charge. Obviously the smaller the amount the faster/less current will be required. [/B]

I hear alot of reference to the "miller" charge time. Is this bascailly the charge time you reger to or the collective charge times that can be listed, i.e. Qg, Qgs and Qgd, the latter of which is "miller" charge time. I have noticed how these values are listed against the Id, Vds, Vgs parameters of test. They are in nC units which I have no clue what that is, but no need to explain that for now, I can dig on internet to find out. The question is, and I have been thinking about it for few weeks, there any way to distill these by way of some sort of ratio so I can compare apples to apples of different devices of same company, let alone of different companies? I tried to apply what math I know, but not being a EE or of thatknowledge I have not figured out how to distill this to common ration I can use to compare different devices on this charge time issue.

Because some amps -- especially quasi-complementary with their different driver scheme for upper and lower -- do not have symetrical source/sink current ability in the driver stage. [/B]

Do any such topologies exist that have symetrical source/sink current ability in the driver stage that can work with N-Channel output stage or are these design oppsites challenges?

Mostly luck/sound design, but because the amount of charge the driver stage has to sink/source goes up in proportion to the number of paralleled output devices, then issues are less likely with single pairs and typical VAS/driver currents. [/B]

How would one modify the Holton N-Channel or is Pelle's design able to currently or with some modification handle up to 4 output device pairs? I have need for 4 pairs at 100-120W leve, then hope I can do a 80W at 2 pair level, a 40 and 20 at one pair level. So the only issue is current Holton desigh is for 3 pairs of output devices. I assume there are changes needed to accomodate a 4 pair design/modification. I am not sure 3 pairs will meet my higher power needs.

Any comment on pros or cons of N-Channel sound quality vs say the LYNX bi-polar design assuming reasonable excustion of circuit. I am aware there can be sloppy and exceptional designs that will skew the generally expected results.


Pelle,

While I am replying here excellent PCB work and effort.


Regards,

John L. Males
Willowdale, Ontario
Canada
09 January 2004 19:17

 

[jleaman]

I like the board layout my self too. so small good for a subwoofer amp id love a copy of the over lay : O )

 

[darkfenriz]

maybe i am wrong but in projects like this the 'lower' mosfet gets an extra current (so it is faster) amplification (so it is faster with same input C) and also extra delay through Q13
I would give an emitter follower to upper mosfet to equalize these driving currents

 

[richie00boy]

So lower the Ciss the better not just from quality of high's reproduced perspective, but also from a greater opportunity of cross conduction?

Yes.

I hear alot of reference to the "miller" charge time. Is this bascailly the charge time you reger to or the collective charge times that can be listed, i.e. Qg, Qgs and Qgd, the latter of which is "miller" charge time. I have noticed how these values are listed against the Id, Vds, Vgs parameters of test. They are in nC units which I have no clue what that is, but no need to explain that for now, I can dig on internet to find out. The question is, and I have been thinking about it for few weeks, there any way to distill these by way of some sort of ratio so I can compare apples to apples of different devices of same company, let alone of different companies? I tried to apply what math I know, but not being a EE or of thatknowledge I have not figured out how to distill this to common ration I can use to compare different devices on this charge time issue.

Miller is Qgd. nC Refers to the unit of charge which is C for columb, n refers to nano.

Device selection is a combination of various parameters, ranked according to the particular design and application you are trying to achieve. The parameters are all tradeoffs so you will never get a device with all ideal parameters.

Do any such topologies exist that have symetrical source/sink current ability in the driver stage that can work with N-Channel output stage or are these design oppsites challenges?

Almost. Designs with active current source loaded VAS tend to have this feature. However, once you decide that you must use quasi output then you spoil the symmetry somewhat as the upper half is working as a source/emitter follower and the lower half as CFP. This causes differences in the way that you can suck out charge from the devices, and their transfer characteristic especially around the transition. Using equal values for both emitter and collector resistor in the driver transistor for the lower device (R17 and R13 in Pelle's schematic) can help in this respect, as can the addition of a diode (for BJTs) or capacitor (for MOSFETs) across the emitter resistor.

How would one modify the Holton N-Channel or is Pelle's design able to currently or with some modification handle up to 4 output device pairs? I have need for 4 pairs at 100-120W leve, then hope I can do a 80W at 2 pair level, a 40 and 20 at one pair level. So the only issue is current Holton desigh is for 3 pairs of output devices. I assume there are changes needed to accomodate a 4 pair design/modification. I am not sure 3 pairs will meet my higher power needs.

From what I recall of this design, Holton claims that it will drive a large number of output devices without modification. Usually all that is required is to run the VAS and drivers at higher current.

As for Pelle's amp, darkfenriz's comment is exactly what I would say. I do not like to drive MOSFETs directly from the VAS because their capacitive input makes for a heavier load at HF, increasing distortion right when feedback is rolling away. He is running the VAS at 7mA which is OK to drive a single (non-high capacitance) pair of devices.

Any comment on pros or cons of N-Channel sound quality vs say the LYNX bi-polar design assuming reasonable excustion of circuit. I am aware there can be sloppy and exceptional designs that will skew the generally expected results.

True complementary designs have more linear output due to greater chance of symmetry around the transition point from upper to lower half conducting. That said, you can make a great quasi amp if you just pay attention to the right details.