ok, thanks
sitting here looking at Nelson's new BA3 driver curcuit, giving me naughty ideas 🙄
well, my skills are very limited 🙁
sitting here looking at Nelson's new BA3 driver curcuit, giving me naughty ideas 🙄
well, my skills are very limited 🙁
One fundamental problem with the circuit in post 1 & 10, is that in source follower configuration the Vgs excursion beyond the drain voltage that is required to fully turn on the Mosfets. Laterals can require up to 10V beyond Vd to fully turn on. Verticals require less due to their exponential transfer but with a higher Vgsth they may require 5 or 6V beyond Vd to fully turn on. Operating the driver stage from the rails, Vd, prevents the output voltage from attaining full swing. This is just one more reason Mosfet amplifers can be more complicated than BJT's, at least if the goal is to make it proper.🙂 BTW, my favorite output devices are Mosfet, but to use them like they're BJT's is going to yield unsatisfactory results when compared to a similar BJT circuit. They must be used like Mosfets and the issues of Mosfet devices must be taken into account.😉
When designing an amp you decide the design parameters, that is what has gone down here. You don't design for 10 watt wonder if you can bleed out 11 watt, so the statement applies to your design philosophy.
I design with an objective which is pretty normal engineering practice and then decide on the components that is appropriate. I do not think that it is at all necessary to drive anything to saturation, it is linear not logic?
I design with an objective which is pretty normal engineering practice and then decide on the components that is appropriate. I do not think that it is at all necessary to drive anything to saturation, it is linear not logic?
Last edited:
I wouldnt say youre doing something wrong but with two pairs you have doubled the input capacitance which will probably affect what you hear, you could try increasing vas current or using a driver like goldmund.
The capacitance which is different for the N and P fet, was one question besides the current and driver.
Would you suggest some picofarad for better balance of the gates.......datasheets shows 900pico gate capacitance for 2SJ50 and 600pico for 2SK135? Gate resistors in the used circuit is 680R.
How much current without signal for each Hitachi do you suggest? We are running at approx. +-60Vdc and 0R22 source resistors.
We have a lot of the old Motorola BF757/760 in stock, and we used them to drive the fets. Would you suggest a better driver?
We have not looked at "Goldmund"!
No use series gate resistors to equalize the response, for 2SJ162 use a 68 ohm resistor and for the 2SK 1058 a 100 ohm resistor. If you want to use bigger resistors scale them the same 66% for the 2SJ168.
The drive current needed is miniscule because it is a voltage responding device, you can calculate the current needed to raise the voltage across a 900pF capacitor with a standard formula calculating the charge versus time, it requires only a few micro amp.
The drive current needed is miniscule because it is a voltage responding device, you can calculate the current needed to raise the voltage across a 900pF capacitor with a standard formula calculating the charge versus time, it requires only a few micro amp.
The drive current can be quite some on step responses and when the MOSFET is driven closer to its Vds saturation region. When driving a sinewave the current may be neglible, but if you want to have a fast output stage that acts well on low frequency step responses, it's going to take peak current that can be a factor 100x - 1000x higher than its steady state tracking current which normally is in a few uAs. It's not uncommon for FETs to gobble up a peak current of several mA which happens when you want an amp with a high slewrate and have a high power output that requires a big voltage swing.
It's why that National Semiconductor LME offers up to 56mA drive current which allows you to build incredibly fast output stages. Rather than increasing the gate stoppers wich decrease linearity as you now get to deal with a spike voltage over the gate stopper, you're better off lowering the gate stoppers and use compensation Cs to slow down the feedback loop.
P.S. my favourites are the Exicon ECX10N20/EXC10P20 laterals - when driven from a higher than your usual VAS current, they can be driven hard (up to a couple mA peak drive) and it makes them sound very transparant as they have no trouble tracking the VAS.
Last edited:
Are we talking relatively low transition audio signals or nS high speed switching. If it is the latter then I would agree that initial very short duration peak currents of amps at the onset of switching should be a consideration. In audio driving circuits capable of around 10 - 20 mA is quite adequate.
nS region is an extreme end of the scale and with only one set of fets it by far won't be such an extreme. But let's take music. What do you get if you would sum several sinewaves of which the magnitude decreases as frequency increases? A square wave. That's why you want a fast output stage. When you listen to music, it's very normal that musicians time well, causing "synchronous addition" of individual waves that make up the instruments that could theoretically end up to be an edge of a square wave. The chance of how often that's happening I don't know, but it's a mathematically possible scenario. If the amp can't output that sum of waves nicely, you might lose detail in the highs for that brief moment. Especially with electronic instruments whose instrument samples aren't offset, might get into clipping quickly which would not happen if tones would be offset by a fraction. Offsetting instrument tracks in sequenced music is an often applied trick to reduce both clipping and steep edges of waves added together, increasing the crest factor.
Up to a mA with only one pair going 100 watts/8Ohm with few voltage left on the PSU rails isn't uncommon, simulate it I'd say 🙂 When you start paralleling pairs you're going to need a strong driver if you want an output fast enough to handle thos exceptional transients that can occur in music. Some would describe this as PRAT.
All I'm saying, keep in mind that MOSFETs can and will draw current in certain circumstances. When we're actually talking high-speed switching, a gate current of 1mA is nothing.
Up to a mA with only one pair going 100 watts/8Ohm with few voltage left on the PSU rails isn't uncommon, simulate it I'd say 🙂 When you start paralleling pairs you're going to need a strong driver if you want an output fast enough to handle thos exceptional transients that can occur in music. Some would describe this as PRAT.
All I'm saying, keep in mind that MOSFETs can and will draw current in certain circumstances. When we're actually talking high-speed switching, a gate current of 1mA is nothing.
Last edited:
I agree. High frequency components do affect the sound and sound stage. For the fast transients in a music signal and all its combined components as you point out may require a large change in conductance from the Mosfet in a short period of time which requires charge. It is not hard to make a good Mosfet amp but there are fundamental things that must be addressed, such as gate voltage drive above Vd. Also it is prudent to dampen the high frequency oscillator that is formed by the internal capacitance and lead/bonding wire inductance. Placing a zero between the gate and drain pin along with local decoupling will accomplish this. A very simple easy way to prevent local oscillation around the device. (I use COG SMD caps for this filter.) Also as you point out a larger gate stopper reduces the frequency of Fc in the LP filter that is the gate stopper/input capacitance. By including the simple local dampening, a smaller gate stopper can be used to create a very fast output stage. Of course the driver stage bias should be adequate at several mA to ensure it's always class A bias.
I am not sure what you are summing or combining. A circuit respond to only one scalar stimulus at any one instance.
Last edited:
I was summing a hypothetical number of sine waves of different frequencies and magnitudes to illustrate music is far from your gorgeous looking smooth sinewave and can be comprised of steep edges of square waves (not that they end up to be nice square waves, just the edge part). Besides designing amps I also compose music and actually see the waves of my instrument tracks visually. The master channel obviously displays the addition of waves of the individual channels.
Do not confuse what you see on a scope with reality, first find out how it works. Why not show us one of your hypothetical amp designs.
Err? Why should a wave form not be what it is?[Edit: I'm speaking of digital mixing/tracking software, not an analogue scope] And what exactly are you trying to tell by finding out how what works? You are reading things I did not write at all. I'm not talking a hypothetical amp design, I was talking a hypothetical wave form that's the sum of several sines.
Last edited:
Sorry I was unnecessary nasty. The question was regarding instantaneous peak drive current to charge the gate capacitance of a lateral mosfet. Let's take it from there.
Homemodder is correct in saying that it is possible to achieve higher switching speed by using a separate driver. However, I tried to indicate that a VAS of a few mA is adequate to drive a Lateral MOSFET output stage directly far beyond audio frequencies quite successfully. It depends upon your design approach and your objectives. We were discussing a very simple audio amplifier.
Last edited:
In summary:
Higher slew rates (wider bandwidth) tend to enhance the subjective terms "speed" and "transparency" of an amp. It is easily demonstrated by the Simple Symmetrical Amplifier of Lazy Cat.
Mosfets are fast devices but the ultimate "speed" of the amp cannot only be attributed to the output devices alone but the sum total of the design approach which may take different forms altogether or topologies depending on the designers objective. No good stable design are better or worse sounding, only different and every good design would find appeal amongst listeners and music tastes.
A final note; it is a lot more tricky to drive MOSFETS at high switching speeds than it is for relatively slow audio, this I only learned about when I was tasked to develop PWM power management systems.
Higher slew rates (wider bandwidth) tend to enhance the subjective terms "speed" and "transparency" of an amp. It is easily demonstrated by the Simple Symmetrical Amplifier of Lazy Cat.
Mosfets are fast devices but the ultimate "speed" of the amp cannot only be attributed to the output devices alone but the sum total of the design approach which may take different forms altogether or topologies depending on the designers objective. No good stable design are better or worse sounding, only different and every good design would find appeal amongst listeners and music tastes.
A final note; it is a lot more tricky to drive MOSFETS at high switching speeds than it is for relatively slow audio, this I only learned about when I was tasked to develop PWM power management systems.
Last edited:
I would think that this complex tool was specifically invented to allow technicians to "see" more of the reality in electrical waveforms.
Old scopes used persistence to display a series of points to make up a wave form while new ones sample points to display a wave form built up over time so you can observe/measure. In other words it displayes a series of samples of "instantaneous" points with respect to time as the signal serially passes through the test specimen at sample time intervals. It is not a complex tool at all.
- Status
- Not open for further replies.