Why not 😉
Hi Christophe, actually your post gave me a great idea (well, at least it looks interesting to me 😉). I have never tried high-power FETs in my amps. And here is a good opportunity to try. I've got a second SlewMaster board un-populated so far, and I have checked with the local store - I can buy some IRFP240/9240 quickly. And their pins (GDS) are in the same order as the ones (BCE) of BJTs, the boards are designed for.
I did some simulations. Perfect at 1KHz, however, at 20 KHz, being connected directly to the VAS cascode, running at 4.7 mA, it seems to be difficult for the cascode to drive at least 3 x 1300 pF input capacitance (3 pairs of MOSFETs). 20KHz THD increases dramatically. However, using a EF buffer, running at 12 mA, makes the situation way better. I will measure both options on the prototype, so we'll see. There are strange peaks on the spectrum, repeating every 4 KHz between harmonics at some lower level - I normally don't see them with BJTs, got no idea of their nature so far, but we'll see what the real prototype will show.
Regarding your other points - I like VAS with a cascode and in most cases use the balanced symmetric topology with current mirror(s).
Diamond input - no problem, I have tried Damir's input buffer in one of my CFA designs with excellent results - level of linearity it provides allows for lower OLG - and therefore lower LG - something I like.
With regards to the front-ends, I've got another passion recently - hybrid topologies with vacuum tubes at the input. [HERE] is my design with the tube LTP at the input and directly connection to the next stage. It is a "classic" VFA and I really like the way it sounds. So, I'm going to dig deeper in this direction, trying to design a front-end, where all the voltage gain is done by the tubes, and then the current gain will be done with SS - don't know BJT or FET based yet.
This makes the MOSFET experiment even more interesting 😛
OK, I'm out to buy some FETS 😎
This is going to be fun!
Cheers,
Valery
Happy to see, Valery, that you correlate my 'instinctive' habits ;-)
I will add that, at the end, i tend to prefer Laterals for OPS. For two reasons: they are fast and they don't have secondary breakdown risk, making them reliable.
As i never noticed some kind of 'FET coloration' with them, i stay stuck with this technology. I like too the fact that they provide a better 'isolation' with the previous stages.
Their high impedance is an advantage too, as, most of the time, we don't need a pre-driver: Simple stages IPS, VAS and OPS are achievable, despite the high gate parasitic capacitances at HF: Simple is beautiful and good to keep the source's 'details".
For VAs, the only question that remains is Cascode or not.
For IPS and a CFB topology (that i definitively prefer after years of listening comparisons), apart CCS or not, the choice is a cap in the signal path (VSSA like), or some kind of diamond input to cure the offset. Despite the bad opinion of 'Lazy Cat' about Diamonds, based on his listening experience with this topology, i am sure that it is the way to dig. Because the extra transistor is not in the feedback loop, because it reduce the distortion at low frequencies, due to the phase shifts of the feedback cap there, because it produce a better isolation between the input line and the loop. Because it is easy to fix a low pass filter not influenced by the source's impedance.
I am interested by your comments about all those expectations.
Hi Christophe, actually your post gave me a great idea (well, at least it looks interesting to me 😉). I have never tried high-power FETs in my amps. And here is a good opportunity to try. I've got a second SlewMaster board un-populated so far, and I have checked with the local store - I can buy some IRFP240/9240 quickly. And their pins (GDS) are in the same order as the ones (BCE) of BJTs, the boards are designed for.
I did some simulations. Perfect at 1KHz, however, at 20 KHz, being connected directly to the VAS cascode, running at 4.7 mA, it seems to be difficult for the cascode to drive at least 3 x 1300 pF input capacitance (3 pairs of MOSFETs). 20KHz THD increases dramatically. However, using a EF buffer, running at 12 mA, makes the situation way better. I will measure both options on the prototype, so we'll see. There are strange peaks on the spectrum, repeating every 4 KHz between harmonics at some lower level - I normally don't see them with BJTs, got no idea of their nature so far, but we'll see what the real prototype will show.
Regarding your other points - I like VAS with a cascode and in most cases use the balanced symmetric topology with current mirror(s).
Diamond input - no problem, I have tried Damir's input buffer in one of my CFA designs with excellent results - level of linearity it provides allows for lower OLG - and therefore lower LG - something I like.
With regards to the front-ends, I've got another passion recently - hybrid topologies with vacuum tubes at the input. [HERE] is my design with the tube LTP at the input and directly connection to the next stage. It is a "classic" VFA and I really like the way it sounds. So, I'm going to dig deeper in this direction, trying to design a front-end, where all the voltage gain is done by the tubes, and then the current gain will be done with SS - don't know BJT or FET based yet.
This makes the MOSFET experiment even more interesting 😛
OK, I'm out to buy some FETS 😎
This is going to be fun!

Cheers,
Valery
Working on the Wolverine and it is nearing completion. A couple of queries, maybe someone remembers discussion about these points in OS's absence.
1) In reference to either the v1.2 or v1.3 Wolverine schematics do we need Q10 to be the same device as Q11? I based my art on Q10 being a TO-92 device but can change it if OS had a good reason for making it a TO-126 and coupled to Q11.
2) If 'Option 1' is invoked, it appears that Q3 is to be left off but OS provided no note as such.
3) There is also an 'Option 2' that appears related to compensation without notation as to what the pads are for. This is on OS's art and not the schematics.
1) In reference to either the v1.2 or v1.3 Wolverine schematics do we need Q10 to be the same device as Q11? I based my art on Q10 being a TO-92 device but can change it if OS had a good reason for making it a TO-126 and coupled to Q11.
2) If 'Option 1' is invoked, it appears that Q3 is to be left off but OS provided no note as such.
3) There is also an 'Option 2' that appears related to compensation without notation as to what the pads are for. This is on OS's art and not the schematics.
Working on the Wolverine and it is nearing completion. A couple of queries, maybe someone remembers discussion about these points in OS's absence.
1) In reference to either the v1.2 or v1.3 Wolverine schematics do we need Q10 to be the same device as Q11? I based my art on Q10 being a TO-92 device but can change it if OS had a good reason for making it a TO-126 and coupled to Q11.
2) If 'Option 1' is invoked, it appears that Q3 is to be left off but OS provided no note as such.
3) There is also an 'Option 2' that appears related to compensation without notation as to what the pads are for. This is on OS's art and not the schematics.
Ok, I found my answers to 2 and 3. Question 1 still stands.
Hi Christophe, actually your post gave me a great idea (well, at least it looks interesting to me 😉). I have never tried high-power FETs in my amps. And here is a good opportunity to try. I've got a second SlewMaster board un-populated so far, and I have checked with the local store - I can buy some IRFP240/9240 quickly. And their pins (GDS) are in the same order as the ones (BCE) of BJTs, the boards are designed for.
I did some simulations. Perfect at 1KHz, however, at 20 KHz, being connected directly to the VAS cascode, running at 4.7 mA, it seems to be difficult for the cascode to drive at least 3 x 1300 pF input capacitance (3 pairs of MOSFETs). 20KHz THD increases dramatically. However, using a EF buffer, running at 12 mA, makes the situation way better. I will measure both options on the prototype, so we'll see. There are strange peaks on the spectrum, repeating every 4 KHz between harmonics at some lower level - I normally don't see them with BJTs, got no idea of their nature so far, but we'll see what the real prototype will show.
Regarding your other points - I like VAS with a cascode and in most cases use the balanced symmetric topology with current mirror(s).
Diamond input - no problem, I have tried Damir's input buffer in one of my CFA designs with excellent results - level of linearity it provides allows for lower OLG - and therefore lower LG - something I like.
With regards to the front-ends, I've got another passion recently - hybrid topologies with vacuum tubes at the input. [HERE] is my design with the tube LTP at the input and directly connection to the next stage. It is a "classic" VFA and I really like the way it sounds. So, I'm going to dig deeper in this direction, trying to design a front-end, where all the voltage gain is done by the tubes, and then the current gain will be done with SS - don't know BJT or FET based yet.
This makes the MOSFET experiment even more interesting 😛
OK, I'm out to buy some FETS 😎
This is going to be fun!
Cheers,
Valery
Hi Valery
I've been working on a design that uses IRFP240/9240 - a single pair operating from 35V rails driven directly from the VAS, with an unusual double bootstrap arrangement to supply some extra drive voltage into the gates without resorting to boosted rails.
http://www.diyaudio.com/forums/soli...fier-based-rod-elliot-p3a-51.html#post4009564
I would be interested to see your VAS cascode design, and to know how well it simulates when having driving gate capacitance of a single pair of MOSFETs.
How are you planning to use for the bias spreader?
I guess my question is why are we discussing MOSFETs in this thread? Just seems like it would fit better somewhere else. Maybe I'm missing something.
To be fair, I have thought about it also. There is no technical reason why MOSFETs couldn't be used as outputs in the OPS. I just haven't had the stones to try it. Of course some adjustments would have to be made to some component values, but it should be a valid option for those adventurous enough.
Guess that would be the SlewMOSter...
Guess that would be the SlewMOSter...
Do you think there is any benefit to using MOSFET outputs in this design? Better Slew? Lower THD? Lower PSSR? What would be the reason for the change? MOSFET's are more expensive and more difficult to find. I'm just wondering why we would try to fix something that ain't broke.
You are right, it isn't broke. Devices like the IRFP240/IRFP9240 are pretty cheap and readily available. They are also fast and recent discussions show two things. We are looking for fast devices and need to drive them with at least one EF stage to hide the capacitive load of the MOSFETS gates from the VAS.
I think I might just try it with a baby OPS. I have some 240/9140 on hand to try with a 45V set of rails.
I think I might just try it with a baby OPS. I have some 240/9140 on hand to try with a 45V set of rails.
I tried to find some IRFP9240 a while back and had a heck of a time finding them. The IRFP240 are around but I couldn't find the 9240 except on ebay and who knows if they are legit. Looking forward to what you find out.
Really? Folks seem to be using them all over the place. Mind you I bought the P channel as the 9140, which except for voltage rating is an almost perfect compliment to the 240.
Quick peek shows over 1000 of them (the 9240) by Vishay at Mouser (Canada).
Quick peek shows over 1000 of them (the 9240) by Vishay at Mouser (Canada).
Last edited:
Hey, I just checked and looks like Mouser has them on order so maybe they will still be around. Anyway, still looking forward to hearing how you do.
Hafler uses Laterals.
Yea, vertical versus lateral but still a related tech. My point was simply that I have fond memories of non BJT amplifiers that I have owned. I think it would be worth exploring in the spirit of DIY. I wonder how different they would sound? I would guess probably not better or worse, just different.
Last edited:
While on the subject of Mosfets for OPS, there is nothing quite like designing a quasi-complementary setup using N-channel Mosfets only as an alternative cos they are available easily and sound great, provided cross-conduction is avoided and one doesn't bother too much about asymmetrical clipping. One can also experiment with changing tonal characteristics by swapping Mosfets with different Gate capacitance.
Maybe, but not likely here.
I'm interested in seeing if this could be a 'drop-in' option with thermal compensation / bias values adjusted accordingly rather than a total redesign. A quasi OPS doesn't do a thing for me at this point and would be enough of a departure to warrant its own thread.
The idea of putting in MOSFETs in this OPS is already pushing things a little, but I'm already assembling one to test.
I'm interested in seeing if this could be a 'drop-in' option with thermal compensation / bias values adjusted accordingly rather than a total redesign. A quasi OPS doesn't do a thing for me at this point and would be enough of a departure to warrant its own thread.
The idea of putting in MOSFETs in this OPS is already pushing things a little, but I'm already assembling one to test.
Do you think there is any benefit to using MOSFET outputs in this design? Better Slew? Lower THD? Lower PSSR? What would be the reason for the change? MOSFET's are more expensive and more difficult to find. I'm just wondering why we would try to fix something that ain't broke.
Well none of the IPS designs are 'broke' either but that hasn't stopped OStripper from publishing several.
Personally I like the idea of having a few different OP boards to try.
You are right, it isn't broke. Devices like the IRFP240/IRFP9240 are pretty cheap and readily available. They are also fast and recent discussions show two things. We are looking for fast devices and need to drive them with at least one EF stage to hide the capacitive load of the MOSFETS gates from the VAS.
I think I might just try it with a baby OPS. I have some 240/9140 on hand to try with a 45V set of rails.
Strongly endorse this idea, particularly if its a 1-2 pair version with compact dimensions. I'll be first in line for a couple of PCBs if it ever gets to that point.
- Home
- Amplifiers
- Solid State
- Slewmaster - CFA vs. VFA "Rumble"