Finally I've got an update to this project, (November 1st. '99), at the bottom of this page. Unfourtunately computers are not what get's me kicking, so I'm rather slow when it comes to updating and fixing bugs on this website, sorry...
PS: This is a single-ended transistor amp, and runs hot. Therefore it is only practical up to maybe 40watts or so. I built it as a 15watter, and have never missed any dBs in the years I've had it. It is also designed with total disregard to efficiency. DIY'ers can do that and get the priorities straight, that is sound quality before such details as efficiency that commercial products must consider.
Concept:
The amp has a balanced single voltage gain stage('VAS') and a balanced single-ended output stage. That means both speaker outputs are hot. When no loop feedback is used, squeezing too much gain from a single transistor is not wise in terms of distortion. The balanced topology gives twice the gain vs single-ended gainstages, enabling 'easy' gain from each half, keeping it all linear, and still giving enuff gain for proper levels. Another important feature of this topology is the distortion cancellation. Each half, or phase, of the circuit will have farely similar nonlinearities. Since the nonlinearity is similar, but opposite phase, the effect is a reduction of the nonlin. The single-ended configuration ensures that mostly even order harmonics are present at each output. (Those harmonics are extremely low by the way.) The balanced circuit cancels much of the even order harmonics, leaving very little distortion even without loop NFB.
Another important (yes there are many) benefit of the balanced design is the draw on the powersupply is greatly eased. Look at where the load gets the current. When the positive half goes up the negative half goes equally down. The constant current sinks will not allow an increase in current no matter how high a signal. In the simplified figure above, the idle current is 2.5amps. The current draw from the supply is 5amps... always! This means there are no problems associated with inductance and resistance in powersupply. This may be one of the reasons for the amp's awesome bass performance, even with the high output impedance this zero loop NFB amp has. The supply rails can be half the voltage of a similar power non-balanced amp as well, allowing larger capacitance for the same size and money.
MOSFETs are great devices even when used without loop NFB. I've seen many claims to the opposite, but must simply conclude those oppinions are made up without giving them a listen. MOSFETs should be driven with high idle current, but that's a given, at least in my opinion. This amp may be built with bipolars of course, but the tripple or quad darlington config required doesn't make sense to me.
Many people claim the low output impedance of amps with no loop NFB gives poor dampening factor and thus sloppy bass. Just like the ones who claim MOSFETs are no good, these claims must come from simple prejudice, and not real listening. I use a single device per outputhalf, and even add source resistors, and I tell ya, the bass is deeeep, detailed, and tight as can be.
The following is my first version of this amp. It sounds great, but the circuit is kinda messy. I've also come up with better gain stages using mirrors that move it another level up. But for now I'll leave the old design here, since it works great. Hell, the main reason for my website is to exchange ideas, so the more variations of various circuits, the better.
Circuit of the VAS: - VAS2:
The input stage (VAS) is configured as a differential pair with another pair of transistors on top forming a cascode.
Notice on this version the inputs are set at a negative voltage and AC-coupled to the preamp. This to enable direct coupling of the output MOSFETs and still have output offset voltages near ground level. Most preamps have offsets anyways, and must be AC-coupled, so nothing lost there. Because of the offset input, the circuit does not give full symmetric output, the negative half will clip at about -20volt, (the positive half will go above +45 volts before clip). With a balanced output, +/- 20 volts is equal to +/- 40 volts across the load, or 100watts/8ohms, so no sweat. I need less than +-10 volts on each side ( for a total +-20 volts to the load) for 15watts/8ohm. This means the VAS will never operate anywhere near it’s extremes, providing low distortion signals all the way to it’s outputstage’s awesome max power output .
Output stage: - BBout2
This is a messy circuit, just like the VAS, the constant current sinks can be done a lot simpler and neater. But it works and I like the cascoded sinker, it ensures constant current no matter how high the output swings, up to the point of saturation of course.
Both outputs from the differential VAS are used, fed into single-ended(SE), not push-pull, power MOSFETs. The amp is therefore fully balanced all the way to the load, the speaker negative is connected to the inverted output, not ground. Since SE driving creates super amounts of heat, I used a single 250watt transistor per side(inverted and non-inverted side).
A Zobel network (10nF + 10ohm resistor to ground) is used, but no coil.
Power Supply:
Each channel draws 4 amps continously and use seperate supplies. I use a 250VA transformer for the positive supply and a 300VA for the negative. The voltage supply for the output stage is +13V and -17V, and each channel has 200,000µF of storage capacitance.
The VAS has its own 50VA transformer and 100,000µF, per channel. No active regulators are used, just brute force regulation with the capacitors, chokes, and resistors. I use many small caps in paralell, since this gives lower ESR and inductance. Voltage supply is +/-50V.
*A little note on the psu* The very high cap reservoir described above creates super amounts of charging currents. After several mods the amp now sports about half the capacitance and I must say I think the sound actually improved. I think this is because the large currents in the previous version created noise and interference in the rest of the system.
Performance:
(This is without loop NFB)
15watts/8ohms, 8watts/4ohms.
Gain about 20dB.
Input impedance: 50kohms (without attenuator).
Output impedance is about 0.5 ohms. (As of fall '99, I've changed output mosfets, so now it's 1ohm! See notes added in the output stage schematic description.) This is all up to the output devices chosen, and the number in parallel, which I don't use.
It’s upper F3 (-3dB) point is just under 270kHz at full power, negative slew a little faster than positive.
Notes and other experiences:
I thought one of the benefits of DIY was that I'd save money...ha-ha...
Squarewave, 23.5kHz, into 8ohm resistor, bottom trace input, top output:
I have now had this amp for a little over a year and have not been able to let it be, all though I was very happy with the original. The main drawback with this simple circuits is the need for bringing the gates of the input MOSFETs down in order to DC-connect the output stage. The problem with that is that ripple on the negative voltage rail is also at the gates since a simple voltage divider is used to bring the gates down. The hum this produced is actually not that much, I've heard many expencive amps with more hum. Besides the hum from one of the transformers is actually louder than the hum at the speaker. But I couldn't leave it alone...
The main reason I did it this way was to DC-connect the output stage, thinking better control of the input capacitance of the power MOSFETs was acheived. Well after finishing the balanced tube preamp, I decided to try and bypass the MOSFET VAS in this amp and connect the preamp directly to the output MOSFETs. The tube preamp is a traditional resistor loaded design and needs a capacitor at the output, so the MOSFET output stage was now AC-coupled and I was excited to hear that the speed and punch was still present.
Now to the point:
I have got a few modifications that may interest those interested(my great english!). The only sonic difference is total lack of noise at speakers.
More or less the same old design, AC coupled to output stage: - new VAS
Just as I thought I was done with transistors and giving it all to tubes, comes an e-mail kick starting renewed interest in this design. I haven't tried it yet, so a few adjustments are likely needed, but none more than with any diy project: - latest version, Nov. 18th.'99.
Here's a suggestion for PCB layout, non-cascode version, VAS only: - VAS PCB.
Here's a suggestion for chassis layout: - Chassis. *BTW. I've found using slightly taller feet greatly increases air circulation. Having used 1cm tall rubber feet, I switched to about 2cm brass spikes, and the entire amp runs noticably cooler. Just a simple detail...
*I know MOSFETs and single-ended designs are kind of a Nelson Pass trademark. He is definately one of our time's great designers, and I've learned a lot from his projects. But this here is my design-I'm not saying it's so great that it can be interpreted as Pass' design-just that I didn't copy from him. Just felt I had to set that straight...