FETish: ultra-simple transconductance amp

[Mr Evil]

Some threads around here recently got me thinking about a few things, namely extremely simple designs, class-AB input stages, and transconductance amplifiers (current drive for the speaker) as a means of reducing speaker distortion. It just so happens that the three are somewhat complementary, as you can see in the attached schematic.

Starting with the input stage, it uses JFETs (2SK170/2SJ74) for the self-biasing, which reduces component count, and it has a class-AB "current feedback" topology (low-impedance feedback node), which also means fewer components compared to an LTP. Both of these cause poor DC-offset, so a potentiometer is required for adjustment. The exact value of pot. needed depends on the Idss of the JFETs and the desired quiescent current; it is 100R in this case, for a quiescent current of ~2mA.

Normally, an audio amplifier has a VAS followed by a current amplifier stage to reduce output impedance, but a transconductance amplifier needs a high output impedance, so I just did away with that and made the VAS the output stage. It uses lateral MOSFETs (ECF20N16/ECF20P16) because their negative temperature coefficient lets me get away with using just the pair of resistors (270R gives ~100mA quiescent current) for biasing and no source resistors.

Current feedback is provided by R3. A value of 1R0 leads to a rather low gain, but I don't have any smaller resistors that will fit in a breadboard...

And that's it. A mere 8 components, not counting supply bypassing. It's not going to win any prizes for technical performance (sims say 0.5-2% distortion depending on device matching and only 800R output impedance, but high bandwidth and slew rate), but the idea is that because current drive for loudspeakers results in lower distortion from them, it should be possible to get away with much higher distortion from the amplifier and still have low overall distortion.

It's only breadboarded at the moment and I haven't performed any measurements, but I connected it up to an 8" full range speaker I have lying around and it actually sounds pretty good. The lack of electrical damping due to the amplifier's high output impedance increases bass output at resonance. This means that only speakers with high mechanical damping are likely to be suitable unless you like high Q. Current drive also increases output at higher frequencies where the speaker's inductance would normally start reducing output. Again, this is only going to work with some speakers, others will become too bright.

[Mr Evil]

Here's a photo of the breadboard, which gives a good idea of just how simple this circuit really is - there is more space taken up by the wires than by actual components.

One of the uses I intend to put this to is in a motional feedback system (current drive makes sense here). I have tried this before using piezo transducers as the feedback element, but failed due to stability problems at worthwhile feedback levels. I want to try again using a dual voice coil woofer for feedback.

[454Casull]

http://www.essex.ac.uk/dces/research...nt%20drive.pdf

They use a coil to sense velocity.

What do the sims say about the distortion spectrum? Hopefully it's mainly 2nd and 3rd-order.

[Mr Evil]

Originally posted by 454Casull
http://www.essex.ac.uk/dces/research...nt%20drive.pdf

They use a coil to sense velocity...

Interesting. I just skimmed it, but they say they wound a coil over the voice coil, so using a dual voice coil woofer should have the same result.

Originally posted by 454Casull
...What do the sims say about the distortion spectrum? Hopefully it's mainly 2nd and 3rd-order.

I just this minute finished doing a measurement of the physical circuit, so attached is the distortion spectrum, 1kHz, 1W into an 8R resistive load. THD is about 2.2%, mostly 2nd harmonic at about -33dB. 3rd and 6th harmonics are very low for some reason.

[Mr Evil]

And here's the simulated distortion spectrum for comparison. It is more optimistic about 2nd harmonic, but shows higher harmonics too high (maybe I need to turn the resolution up). It fails to show the low 3rd and 6th.

EDIT: I forgot to mention above: I can only make single-ended measurements for distortion, which is not possible for this circuit due to the series feedback resistor. I therefore measured the voltage across the feedback resistor instead, and assumed that the current drawn by the inverting input of the amp is insignificant.

[Nelson Pass]

I like the name.

[Wavebourn]

Similar approach:

http://www.diyaudio.com/forums/attac...amp=1222978630

[CBS240]

Hi

Question: Do the J-fet's have enough conductance to drive Ciss of the output mosfets effectively?
After all, with the common drain configuration, you don't get the advantage of bootstrapping the input capacitance as you do with common source.

[Wavebourn]

Originally posted by CBS240
Hi

Question: Do the J-fet's have enough conductance to drive Ciss of the output mosfets effectively?
After all, with the common drain configuration, you don't get the advantage of bootstrapping the input capacitance as you do with common source.

Let's try on a napkin:

10 mA
-------- = 10 V / uS
1 nF


If S=10A/V for 5A we have 20 V/ uS.

I'm not familiar with Laterals, so may be wrong.

[Mr Evil]

Originally posted by Nelson Pass
I like the name.

I thought it appropriate because I do like to use FETs everywhere, probably more than is appropriate really, but I'm sure you understand

Originally posted by Wavebourn
Similar approach:

http://www.diyaudio.com/forums/attac...amp=1222978630

I've not built an amp with an output transformer before. Is there any particular advantage? Also, is that negative current feedback but positive voltage feedback you have there?

Originally posted by CBS240
Hi

Question: Do the J-fet's have enough conductance to drive Ciss of the output mosfets effectively?
After all, with the common drain configuration, you don't get the advantage of bootstrapping the input capacitance as you do with common source.

Let's see... assuming 15V peak output voltage at 20kHz gives a required slew rate of 2 * pi * 20e3*15 = 1.88e6V/s. Assume capacitance is 1nF (I can't find a datasheet for these particular MOSFETs, but it's almost certainly less than that) means the maximum current required is 1e-9 * 1.88e6 = 1.88mA That's a peak current of less than the quiescent current required to charge 1nF@20kHz.

Yfs for a 2SJ74 is 22mS, so the input voltage for that is 1.88e-3 / 22e-3 = 85mV

Not only is that not going to be a problem, but the sims say that the slew rate of the amp as a whole is about 200V/uS or something silly like that. I'm not sure what the real amp performs like in that respect because I can't be bothered to set up a decent square wave generator (I suppose I should build a permanent one sometime), but simulated slew rate is usually accurate in my experience.