JFET input, MOSFET VAS, LATERAL output = Perfect!!

[danspy]

Clearly not a finished AMP yet,

I think it needs a DC Srevo because it drift,s with Temperature and with Voltage, sometimes up to +-100mV

[mikelm]

I've never been that worried about up to 100mV of DC offset - based on power calculations

100mV / 6ohms = 16mA current

16mA x 100mV = 0.0016Watts

. . . any reason why I should be worried about this ?

mike

[WuYit]

Hugh,
agreed.

mikelm,
so I would aim for single-ended cascoded input stage (JFET/bipolar), no VAS, common source output stage and DC feedback (no IC servo).

[AKSA]

Lumba,

The cascode has disadvantages:

1. While it lowers distortion, it does NOT improve the sound (I've verified this empirically).
2. It is not needed with a rail voltage of just 25, since Vds is rated well above this voltage.
3. Zout, and thus drive, is not as good as a resistively connected drain.
4. It adds one further active device, and attendant bias networks.

In fact, if we were to add a cascode, it should be constant power cascode, where the cascode hikes up and down with the input signal, keep Vds constant across the input/feedback device.

Lastly, if we use a SE cascoded input stage as you suggest, then it would have to be cap coupled to drive a common source output stage. Cap coupling within a SS amplifier is less than desirable. There are ways around this, but they are wacky too.

No, I feel the amp should be left pretty much as is, though with the CCS below the jfet source omitted.

I support Mike's comments about offset. People are very concerned about it, but even at 200mV of offset, a high figure by the exacting standards of folks here, the voice coil DC dissipation is a miserable 5.8mW. Most speakers are designed to dissipate at least 25W continuously, so this is a walk in the park in broad daylight..... no, what is more important is the displacement of the cone with a bit of standing DC voice coil current. What does it do? Increase even, low order distortion, that's all - you wouldn't notice it in a million years of listening.

Kapitz?

Hugh

[tinitus]

Originally Posted by AKSA

Most speakers are designed to dissipate at least 25W continuously, so this is a walk in the park in broad daylight.....

yes
funny tho that the low voltage single jfet was my immediate worry too
it won't be long until people ask for double outputs and more power

[GregH2]

All,

Edit: The schematic has an error. I am currently using no output zobel or snubber of any kind. The amp seems stable enough without it.

Taking some of master Hugh's advice on board, I have created the best incarnation of this amplifier I have heard yet. It appears stable, and DC offset appears predictable and manageable. Better yet we have dropped several active devices to do it. Note the minuscule distortion figures at 16V p2p into 8R. Hugh will also be happy that the low order harmonics taper off in ascending order

The modifications are all things Hugh quietly suggested way back in the thread.

As my JFETs have not yet arrived I am stuck with the mosfet front end and parts I had lying around. The bias scheme shown may have to be modified to allow a negative gate bias if a jfet is used. If there is enough interest I can post a schematic for biasing a jfet.

I have used resistor values I had at hand and the cheapest of cheap transistors. I think the circuit could be significantly improved and tweaked, but here is a good starting point. The jfet front end will work wonders, too.

The amp as it stands sounds very different to how it did before. The warmth that was in my initial version with the bootstrap VAS has come back. If you like valves this is the version to build. If you want detail, the last version I posted with the fet vas and bjt referenced CCS is the way to go. That said, my VAS may be getting let down by the poor parts I have used.

The amp seems stable and sounds very very nice. If you listen to easy going vocals you're in for a treat. A Dave Gilmour solo can sound pretty good too

[lineup]

Nice and simple circuit. Clean and easy to understand.

THD 0.005 is low and typical for a well built version.
Hard to make it lower using so simple amplifier.

I wait for you to get JFET.
Will be interesting to see what you will come up with.

And I would use a Zobel: 47nF + 10 Ohm
Just in case.
Does not hurt or effect any sound.

[GregH2]

Originally Posted by lineup

Nice and simple circuit. Clean and easy to understand.

THD 0.005 is low and typical for a well built version.
Hard to make it lower using so simple amplifier.

I wait for you to get JFET.
Will be interesting to see what you will come up with.

And I would use a Zobel: 47nF + 10 Ohm
Just in case.
Does not hurt or effect any sound.

Thank you lineup. Yes you're correct, some sort of rc zobel would be prudent. I just wanted to demonstrate how stable the amplifier appears to be. This is a far cry from my initial build which I could not get to settle down.

I am very happy with it in this form. Sounds good and square wave response is spectacular as the higher dissipation of the bd transistors has allowed me to add another 5ma of vas current (total 15ma or so). On first impression dc offset is much more stable too. All in all there's a lot of bang for your buck - enviable performance for only a few $$ worth of parts. I built this current design on a perfboard in under an hour.

[lineup]

Hope the JFET version will be stable, too.
I note you did not even needed a compensation cap in this version

[GregH2]

All,

DC offset results for the new circuit are in, and you're going to be very happy. These are a worst case scenario. They were taken using an unprotected circuit on a perfboard in a drafty room. Temperatures inside were warm and cold gusts were blowing in the window throughout the test. Temperature variation of the components is expected to be significant.

Hugh suggested removing the input ccs and biasing the fet gate with LEDs when the thread was in its infancy, but I didn't try it at the time as I didn't think such a simple circuit could work so well. I was wrong.

Turn on "thump" is only 120mV or so, much better than the 500mV of the original circuit. Offset drops to under 100mV almost instantly, and to within 20mV in a couple of minutes. It then easily stays within 15mV for the remainder of the test (total test time = 50min). The small scale deviations are higher in frequency than the original circuit, but the overall long term stability is better....and this is with two less active devices.