So, I know Mr. Pass seems to prefer MOSFETs, particularly vertical MOSFET's, in his designs. Now, granted, their lack of current crowding and resultant hot-spotting is a plus in a class A design, which may be a big factor in his use.
So, why do each of you prefer MOSFETs? I have built amps with both (as well as toooobs
), and while the MOSFETs have done the job I have always wished I could get devices with closer complementary specs. This is one area where I feel bipolars have an advantage...
So, why do each of you prefer MOSFETs? I have built amps with both (as well as toooobs
), and while the MOSFETs have done the job I have always wished I could get devices with closer complementary specs. This is one area where I feel bipolars have an advantage...
For vertical type mosfets there is no such thing as true complementary. Bipolars have an advantage here in keeping good symmetry without over complicating the circuit design and layout. It is my opinion that the 'complementary' operation of the mosfet output stage is achieved in the driving stage.
IMHO, mosfets are more complicated to use properly than BJTs, but for all the extra effort, I think they are more suited for driving reactive loads such as speakers. The effective Ft is much higher and they provide excellent current gain with no secondary breakdown and typically larger SOA for the same package........you just have to do something about the non-linear Gm at low currents.....or just use them in class A.
IMHO, mosfets are more complicated to use properly than BJTs, but for all the extra effort, I think they are more suited for driving reactive loads such as speakers. The effective Ft is much higher and they provide excellent current gain with no secondary breakdown and typically larger SOA for the same package
........you just have to do something about the non-linear Gm at low currents.....or just use them in class A.
The clear winner is the well designed circuit.
They can all sound great or bad for that matter, it's up to the designer to work his/her magic.
Personally I like hollow state with it's simplicity and good looks. I would be hesitant to claim sonic dominance due solely to the tubes themselves. It might be easier in some ways to achieve a good sound than using other techs but they can all sound really good if designed right.
They can all sound great or bad for that matter, it's up to the designer to work his/her magic.
Personally I like hollow state with it's simplicity and good looks. I would be hesitant to claim sonic dominance due solely to the tubes themselves. It might be easier in some ways to achieve a good sound than using other techs but they can all sound really good if designed right.
I believe every transistor builds have their own sound character, as a result of combination of their technical parameters or characteristics, the possible circuit designs and the speaker. And I like the general sound character of a mosfet.
I don't like heat (because I have to pay for the bill, where I usually use my amp almost 24x24), so lateral is preferred over hexfet, as they have lower capacitance which is important in low bias application. Currently I'm hoping that the LazyCat's SSA amp will be the best place for some of my laterals. The Moskido was just too hot (it made my Aikido became useless).
For hexfets, my "best" transistor is the 100W 2SK1341 (Ci=980pF, Co=400pF) that I use for Aleph J.
Bipolars such as the thermaltrack (NJL) MIGHT be better (or not), but I don't have them, and I don't want to purchase them to find out how they "sound". The best bipolar I have (forgot the model number) currently goes to my AKSA-clone.
Last edited:
For many, originally the motivation was that the mosfest were less linear but faster, so despite more correction via feedback you get less time-related distortion due to the feedback and you come out ahead.
I just like that they don't blow out as easily.
Now the new v-fets, that looks like another reason, and I have no relevant experience.
I just like that they don't blow out as easily.
Now the new v-fets, that looks like another reason, and I have no relevant experience.
I would be happy if the price of SiC JFETs would come down to lateral MOSFET levels. I want to experiment with them, but they are a bit spendy at the moment
Otherwise, I am dabbling with a Nelson-ish design right now. Originally I was using ON NJL's, but I may switch to Fairchild verticals. The grand plan is to have jfet-driven outputs somewhat like the F5, but with a bit less heat. That involves a microcontroller keeping a watch over the output temp and bias levels. The idea is to run the bias low if the amp is idling with no signal, then bias up into class-A low-moderate levels (under a few watts out), and finally slide the bias back down to a lean AB at high outputs. The uC makes all sorts of crazy sliding-bias schemes easy. Plus, I want the amp to be mobile-friendly, and full-on class A operation isn't
Otherwise, I am dabbling with a Nelson-ish design right now. Originally I was using ON NJL's, but I may switch to Fairchild verticals. The grand plan is to have jfet-driven outputs somewhat like the F5, but with a bit less heat. That involves a microcontroller keeping a watch over the output temp and bias levels. The idea is to run the bias low if the amp is idling with no signal, then bias up into class-A low-moderate levels (under a few watts out), and finally slide the bias back down to a lean AB at high outputs. The uC makes all sorts of crazy sliding-bias schemes easy. Plus, I want the amp to be mobile-friendly, and full-on class A operation isn't
Class A is obviously preferred but then there is the heat and inefficiency.
I think people put too much emphasis on Ciss when choosing a mosfet to use as a SF class AB output stage. If the driver stage is designed properly it doesn't matter much, mostly due to the bootstrapping effect of Ciss reducing the effective capacitance as seen by the driver stage by an order or so. The AC voltage seen across Cgs is the change in Vgs that is related to the change in conductance. A higher transconductance device will have larger Cgs but require much smaller change in Vgs as per the transfer curve so the effective capacitance loading of the driver stage is not really increased. Cgd is another story but is usually small except close to Vds saturation. The higher voltage mosfet has less transconductance so it requires a larger change in Vgs for the same range of conductance. If a 'complementary' SF stage is desired, one should choose devices that match relativity close in transconductance not input capacitance or breakdown voltage. Different value source ballast resistors can also be used to better match the transfer if one chooses not to use local EC. Mosfets are less prone to thermal run away because of the waning transconductance at lower bias levels. Laterals have an advantage here requiring no thermal feedback. With vertical fets the thermal monitoring circuit does not have to be precise and this allows it to be more versatile. Then there is the higher operating temperature and SOA. Mosfets do not hot-spot (assuming you don't use Trench architecture Fets because they do). Hex-type structure is quite robust but I have found planer stripe architecture like Fairchild's Q-fet to be very rugged. They require less charge vs change in conductance and using a driver stage with local EC, make excellent outputs, IMHO. There are other companies making plainer stripe fets now so the price is quite low. The amp I'm listening to now can output 60Wrms @ 4R and uses one pair of TO-220 Q-fets, P-type = $1.80, N-type = $0.69.
Unfortunately I have never experimented with class AB, but for class A SF, Ciss determines the quality of the sound (especially the critical high frequencies which usually is the weakest point). Even I have compared 2 devices with the same "Ciss per gfs", and the lowest Ciss was better.
In DOGC MkIII, mosfets are used for the input LTP (And I think some of Holton's amp), which is class-B. Here, I found lower Ciss mosfet (even with relatively lower gfs) also sound better, but not enough evidence as there are limited number of mosfets suitable for this job to be compared.
The problem is that it is never that "easy" to design properly, so it does matter imo. Isn't mosfet has been long unpopular due to this "mosfet mist"?
There is something that I don't understand regarding the Vgs and Vgs threshold. It seems to me that mosfet will sound better if the Vgs is high enough to ensure the mosfet to be completely turned on (?).
The problem is, how can we know the precise value of the Vgs threshold?
I have experimented with various mosfet of the same type. I tried to match their Vgs for parallel application. If I'm not mistaken (my paper notes are somewhere), the lower (actual) Vgs sound better. How does this lower actual Vgs correlate with Ciss and gfs? I mean, when the measured Vgs is lower, does it mean that the actual Ciss is lower or the gfs (or gfs per Ciss) is higher?
Yes I have heard about this but I have never been convinced if it is true soundwise, so I have never tried to match an IRF640 with a 9650 for example. I mean, if 540 is better than 640, why downgrade just to match both poles which is never matched anyway? Stochino amp was the last class B amp where I use hexfet, I don't know if I will ever build another amp with these devices to try to match the gfs.
The biggest problem for class AB is the transfer through the zero current crossover region. Those fets suffer from a terrible droop in Gm around this region. Also in relation, there is a larger change in Vgs per G in this region. This is why you would want to bias at a higher level than BJT, to prevent a sudden increase in output Z and consequent distortion. With class A, you don't operate within this region so it is of little issue. An error is produced by transfer though this non-linear zone, from one device to the other, that requires high order frequency components from the fb loop to mitigate or the error will persist into the output signal.
IMHO high frequency components well above 20KHz does have an effect on the 'sound'...ears are naturally tuned for phase differentiations and such distortions. At least for me, lots of crossover distortion is noticable....a little crossover distortion is still noticable. For a switching application as these devices were designed for (exponential transfer), this is not really such an issue. Driving these high order frequencies into the input capacitance requires more current, or charge on demand, whichever way you want to think of it, and thus a more beefy driver stage; since it is important the driver stage is always class A, driver bias for a typical pair might be 10-20mA. Where local EC has its benefits is having a short loop, BW of the error amp can be very high and provide a significant amount of linearization. A side effect of this is very good high frequency damping and PS rejection and the two devices are driven together so the input capacitance is the sum of both. IMO, such a driver stage would certainly be overkill for class A. There may be a few extra parts and complexity in layout but it can be done. Back to the statement,
IMHO high frequency components well above 20KHz does have an effect on the 'sound'...ears are naturally tuned for phase differentiations and such distortions. At least for me, lots of crossover distortion is noticable....a little crossover distortion is still noticable. For a switching application as these devices were designed for (exponential transfer), this is not really such an issue. Driving these high order frequencies into the input capacitance requires more current, or charge on demand, whichever way you want to think of it, and thus a more beefy driver stage; since it is important the driver stage is always class A, driver bias for a typical pair might be 10-20mA. Where local EC has its benefits is having a short loop, BW of the error amp can be very high and provide a significant amount of linearization. A side effect of this is very good high frequency damping and PS rejection and the two devices are driven together so the input capacitance is the sum of both. IMO, such a driver stage would certainly be overkill for class A. There may be a few extra parts and complexity in layout but it can be done. Back to the statement,
"Same" se-circuits at fullrange:
MosFets sound rumbling and a bit unsubtle, with better 3D-stage, without fine curtain; "black".
Silicon bipolars sound 1000 times finer in colours, "more solemn", but with fine curtain (Some bipolars tend to "scratch"); tendency to "grey".
(My) tendencies only,-!!!
PP and, much more, complementer-pp sound ugly, dirty. A or AB. In contrast with se. The more stages the more ugly, dirty. In my mind. May be, the different "characters"! (for ear (not only(-: first different/relative - than wave, swing, string,-) of complementer devices, and topologies, are the reason. The crossover "through the zero current" is not the problem in my mind. Some good, "poor", (not-complementer-) B-amps sound so beautiful - the "distortions" of crossover, switch artefacts give guidance for the ear - to reconstruct a sound-stage, it seems. May be, the fascinating of T- or D-amps is based on a "same" character (and, may be, the reason, why most use pp, complementer-pp, AB, T, D, with some stages/steps (and final: channel separated, step-separated psus), to dominate, to influence complex multiway-speakers,-).
MosFets sound rumbling and a bit unsubtle, with better 3D-stage, without fine curtain; "black".
Silicon bipolars sound 1000 times finer in colours, "more solemn", but with fine curtain (Some bipolars tend to "scratch"); tendency to "grey".
(My) tendencies only,-!!!
PP and, much more, complementer-pp sound ugly, dirty. A or AB. In contrast with se. The more stages the more ugly, dirty. In my mind. May be, the different "characters"! (for ear (not only(-: first different/relative - than wave, swing, string,-) of complementer devices, and topologies, are the reason. The crossover "through the zero current" is not the problem in my mind. Some good, "poor", (not-complementer-) B-amps sound so beautiful - the "distortions" of crossover, switch artefacts give guidance for the ear - to reconstruct a sound-stage, it seems. May be, the fascinating of T- or D-amps is based on a "same" character (and, may be, the reason, why most use pp, complementer-pp, AB, T, D, with some stages/steps (and final: channel separated, step-separated psus), to dominate, to influence complex multiway-speakers,-).
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.