This is true but I believe the transconductance droop of these devices goes beyond the bias that is typically used for class AB. In this experiment done a while back I noticed that the 'error signal' still contains some error, although much smaller, even at 300mA bias. I think that much above 300mA bias per device begins to nullify the intended benifits of using class AB. Obviously employing EC with pure class B bias stretches the limits of the EC circuitry and from experience I can say this can create a problem if the HEC follower circuit is used within the global loop. There is a range of optimal bias comprimise using HEC with vertical mosfets as it pertains to bias current and I suspect it depends on the devices used and how good the error amplifier and gate drive circuit is. Devices with larger overall Gm may have a larger percent change in Gm as the device approaches cut off.
Personally for home Hi-Fi, if the output stage supply rails are +/-24V maximum then the supply caps can be 25V rated which saves on space per uf and price. This allows for ~65Wrms @ 4R, or to bridge two channels for ~250Wrms @ 4R, multiple devices of course. Do you really need more than 250Wrms X 2 in a home stereo system? I suppose some folks could justify it, their neighbors might dissagree.😀 My point is that the power transformer for this supply is 17.5-0-17.5VAC. Using a simple half wave voltage doubler will produce +/-48VDC. Needing to only regulate a few mA, you can use smalle caps, 470uF or so, and easlily drop 8V from each rail without a significant heat penalty giving you +/-40VDC. The maximum swing for the output stage is 25Vp + 10V of bias so the VAS must swing 25Vp within +/-35V. +/-40V gives enough room for this plus 5V for the cascode on each side. It just works out so that the power supply is much simpler, not needing a second transformer or winding for the VAS.😉
"In my middle stage concept I'm splitting things even further. The cascoded device is no longer responsible for gain either. All it does is to provide voltage domain conversion through mirrorring. What I mean by that is that unlike powerrail referenced VAS device, my actual gain devices no longer need to be referenced against supply rails either, they can be referenced at any preferential voltage, allowing for a much wider control range and much more flexible adaption on how the gain devices are controlled. It allows these gain devices to work in class A fully and since their drain voltages never change, it's like the gain devices are inherently cascoded. In addition, each side's (north/south) gain element consists of an N/P combination so the gain behavior is 100% identical for both the north and south swing, especially when the Ns are matched and the Ps are matched. No N vs P matching required."
I am surprised that you have concerns about the gain stages in your amplifier running in class A Magicbox. In a class AB amplifier, everything usually runs in class A up until the output devices, which run in class AB. The VAS should definitely be running class A, and the pre and driver stages.
The BJT vs MOSFET debate comes down to preference I believe - both have their pluses and minuses. To Bob's point, and one that I make in my e-Amp article that you can download on my site below, a good design will will exploit the positives and manage the negatives. For mosfets, HEC seems a very useful technique, while for bipolars, an EFT (aka 'triple') brings significant advantages. I have recently designed and built two big amplifiers using bipolars, and this is what I am most comfortable with. Of course, with me, always fully symmetrical as well!
I am surprised that you have concerns about the gain stages in your amplifier running in class A Magicbox. In a class AB amplifier, everything usually runs in class A up until the output devices, which run in class AB. The VAS should definitely be running class A, and the pre and driver stages.
The BJT vs MOSFET debate comes down to preference I believe - both have their pluses and minuses. To Bob's point, and one that I make in my e-Amp article that you can download on my site below, a good design will will exploit the positives and manage the negatives. For mosfets, HEC seems a very useful technique, while for bipolars, an EFT (aka 'triple') brings significant advantages. I have recently designed and built two big amplifiers using bipolars, and this is what I am most comfortable with. Of course, with me, always fully symmetrical as well!
Hi Bonsai 🙂
I agree up untill a point you see, when the amp is happy and doing its thing the VAS indeed should not ever leave class A or be saturated. But as soon as the OPS bandwidth is exceded VAS devices can be saturated in quite an ugly way when there is an increasing phase shift in the chain. I guess I could say that with my construction the devices can be pinched off better/tighter (regulatory spikes in the control signal) than a powerail referenced device could. All in all it makes the amp's life easier in the difficult high-frequency area.
ON second though, I like Fets in the output.
I like Fets in the output. 2sk135, 2sj50 if available. They just seem to have a more smooth, yet detailed sound. Fets in the input seem to relieve the problem of reflected
power back to the source causing scrambling of the detail in the music.
I like tube preamps. Someone should make a small gain of 1 circuit that
introduces the tubey sound goodness into a transistor system. Maybe a
6922 biased somewhat more nonlinear to introduce maximum goodness.
Distortion is not bad if it sounds good.
I like Fets in the output. 2sk135, 2sj50 if available. They just seem to have a more smooth, yet detailed sound. Fets in the input seem to relieve the problem of reflected
power back to the source causing scrambling of the detail in the music.
I like tube preamps. Someone should make a small gain of 1 circuit that
introduces the tubey sound goodness into a transistor system. Maybe a
6922 biased somewhat more nonlinear to introduce maximum goodness.
Distortion is not bad if it sounds good.
Member
Joined 2009
Paid Member
Hi,
hmmm, and I always thought that a amp would be a little more than just a output stage
jauu
Calvin
If you read Doug Self you'll find that most of the distortion from Class AB amplifiers is from the output stage, the rest of it can be made blameless. In my mind it kinda makes sense, after all the front end is a continuation of the signal chain that started back with the singer and includes a significant amount of electronics. The output stage is what interacts with the reactive load of your speakers and may have more influence than the front end on the sound, depending how it is designed.
That's actually a project I want to do when I finished my amp 🙂 A unity gain tube stage that clips sooner than the amp would (takes away some headroom from your amp) but it should give the authentic tube sound as is, unaltered through the wire with gain.
In addition, just a normal unity gain opamp stage to bypass the tube. Then send them both into a mixer and use a stereo pot to fade in/out the wire or tube channel:
"The Wire" (-o ) "The Tube" <- knob control 😉
Hi CBS240,
You are exactly right. Just because a MOSFET output stage is in the class A region does not mean that transconductance droop is not present.
However, its effect is much smaller. The origin of transconductance droop is the fact that the transconductance of the MOSFET decreases as current decreases at low current, such as in the crossover region. The smaller net transconductance at the lower currents in the crrossover region causes some greater open-loop attenuation against the load resistance. When the idle bias is increased, this smaller minimum transconductance in the crossover region is increased, lessening the effect of transconductance droop. The effect of going from 150mA idle bias up to 200mA idle bias is not stupendous, but every little bit helps.
Cheers,
Bob
I found 2SJ76-79 and 2SK213-216 have announced to be discontinued.
J313 and K2013 are announced EOL too...
IRF610 and IRF9610 may be the last complementary driver MOSFET.
J313 and K2013 are announced EOL too...
IRF610 and IRF9610 may be the last complementary driver MOSFET.
This is a major shame, soon components for true high end amplifiers will all be gone, well be stuck with ICs and loudness war junk toys like Ipods and gainclones. 🙁
That would really stink, driver mosfets seem rather difficult to come by, typical specs being a high Vceo, moderate, small current and a mountable casing like TO220 for dissipation in the several Watts region.
I found that at least in the simulation, IRF9620 is a closer 'match' to the 610 than the 9610 is when it comes to transfer curve; it seems to outweigh the increased output capacitance which, as a cascode, doesn't seem to affect much up to 120KHz with these as cascodes. That is in my particular situation, VAS design though.
And hi Homemodder, nice to see you again.
I found that at least in the simulation, IRF9620 is a closer 'match' to the 610 than the 9610 is when it comes to transfer curve; it seems to outweigh the increased output capacitance which, as a cascode, doesn't seem to affect much up to 120KHz with these as cascodes. That is in my particular situation, VAS design though.
And hi Homemodder, nice to see you again.
- Status
- Not open for further replies.