Does Anyone Here Like Class A amps?

[Joe Berry]

I had orginally thought of using mosfets for drivers but according to several people that was not a good idea, includeing Charles. I suppose I could get enough current thru the second stage (20-40 mls.) I could drive a whole bank of mosfets.

Actually, you can get a really nice sound without all that current. In my (simplified A75) case, setting up the second stage as a pure folded cascode, with +/- 10mA peak driving the output stage gates, gave the clearest sound of all, despite using a global NFB loop. Unfortunately, technical limitations (audible hum, limited output swing at HF) ruled this out for me. You might avoid these problems by using a tiered power supply as in the original A75, and as suggested by John Curl above.

 

[vt4c]

Would be interested to know..Has anyone build the Kaneda 50W class A amplifier before and how it sounds compared to other class A amplifier

 

[jam]

Joe,

Have you tried leaving the outputs outside the feedback loop, If so what were the results?

Charles,

I dug this up somewhere. I has a Sziklai pair for the outputs. Any comments, and choices for input fets?

Regards,
Jam

 

[Joe Berry]

Joe, Have you tried leaving the outputs outside the feedback loop, If so what were the results?

Yes, and I preferred the sound with the outputs outside the loop. In general, this gave a looser but more natural sound, with more dynamic contrast. Now, this was with the second stage set up to provide some degree of current gain. The only way I ever tried the pure folded cascode second stage was with the outputs included in the NFB loop, but (hum aside) the amp sounded quite good to me set up that way.

 

[jam]

Joe,

When you said that you simplified the A75, did you remove one of the mosfets from the second stage?

Regards,
Jam

 

[Joe Berry]

Joe, When you said that you simplified the A75, did you remove one of the mosfets from the second stage?

Yes, I removed the MOSFETs that cascode the folded cascodes, if that makes sense. This was in order to run the front end from the same power supply as the output stage without losing too much output voltage swing. The point was to retrofit the topology into an existing amplifier design (a Forte 1a power amp) as simply as possible. I don't know if you subscribe, but there is a full account of the project in the May & June 2003 issues of Audioxpress magazine.

 

[john curl]

Jam's latest schematic is not optimum. I have made circuits like this 30 years ago, but they are too hard to bias. The feedback pair will just slow the amp.
The bias control is primitive, and is a step backwards from the one on my schematic. I hope this is enough for contemplation.

 

[Steve Eddy]

The feedback pair will just slow the amp.

Why would the Sziklai feedback pair slow things down any more than a Darlington feedback pair?

A Darlington pair slows things down too unless you add a resistor across the base/emitter of the output device. Seems the same function is served by the resistor across the base/emitter of the output device of the Sziklai pair.

se

 

[bocka]

Jam's latest schematic is not optimum. I have made circuits like this 30 years ago, but they are too hard to bias. The feedback pair will just slow the amp.

The bias control is primitive, and is a step backwards from the one on my schematic. I hope this is enough for contemplation.

Completely right. Bias control in a CFP is not easy to build. I needed some month only to get the bias stable in a CFP. And I do not think this simple Vbe multiplier will work correctly.

CFP has other disadvantages: The distortion distribution is worse than a simple emitter follower. Reagardless which toplogy you are using a triple output stage is necessary, when using bipolars in the outputs. MOSFETs require a driver stage (emitter follower) because of the very high input capacitance (some nF) when paralleling them.

To my opinion there are four output configurations possible:

1. classical triple emitter follower

2. CFP or (darlington) emitter follower with SE emitter follower as first stage (prefered)

3. MOSFET output stage with emitter follower as first stage to reduce Cgd

4. single bipolar output with MOSFET driver (prefered)

I've tested configuration 4 with a IRF510/9610 and MJL1302/3281. Works really fine as expected. Biasing can be done with a IRF510 (Vgs mulitplier).

 

[mikelm]

does anyone have a better image of john curls amp. I would like to have a closer look at it and the one posted earlier is really hard to see clearly

thanks

mike

 

[Charles Hansen]

Regardless which toplogy you are using a triple output stage is necessary, when using bipolars in the outputs.

4. single bipolar output with MOSFET driver (prefered)

I've tested configuration 4 with a IRF510/9610 and MJL1302/3281. Works really fine as expected. Biasing can be done with a IRF510 (Vgs mulitplier).

Hello Bocka,

I agree with your first point 100%. A two-stage bipolar output loads the circuit far too much, especially when driving low impedance loads. For more on this, please refer to:

http://users.ece.gatech.edu/~mleach/lowtim/output.html

I've also used your configuration #4 with good results. Sometimes this is called a "FET-lington". However, I would definitely avoid using the vertical MOSFETs you've chosen. The input capacitance of these devices is quite high, but even worse, extremely non-linear. This is true of all vertical MOSFETs. A much better choice are the lateral MOSFETs specifically designed for audio, such as the BUZ900P and BUZ905P. These are improved versions of the original Hitachi lateral MOSFETs.

Good luck,
Charles Hansen

 

[bocka]

Hi Charles,

I've used the IRF610/9510 - not the IRF510/9610 as posted before - because they're layin around and I want to see if this topology works as expected. I'm really astonished how small the reverse capacitance of the BUZ900/BUZ905 is. First idea was to use the Renesas 2SJ76/2SK213 but they are suffering from a too small power dissipation. I'm biasing the MOSFETs quite high about 150mA @ 40V. Thanks a lot for the Magnatecs, I didn't take a look at the Cgd before although the datasheets are on my server...

 

[Charles Hansen]

Hello -

In addition to the small Cgd of the lateral MOSFETs, they have another huge advantage. If you look at the data sheets, Cgd is typically specified at 10 Vds. But what isn't shown on the data sheets is how the value of Cgd varies as Vds varies.

As Vds goes to zero (the amplifier approaches clipping), Cgd of a lateral MOSFET increase by a factor of 2 or 3 which isn't so great. But compared with a vertical MOSFET, it is wonderful. A VFET increases by a factor of 20 or 30! So stay away from the vertical MOSFETs if possible. They were designed as switches, and not linear amplifiers.

Best regards,
Charles Hansen

 

[jam]

Hi Charles,

Thanks for the information, you and Mr.Curl have given me a lot to think about.

This is what I have so far.

a) J-fet differentials .

b) Folded cascode like Joe describes.

c) Vbe or Vgs multiplier (how about a diode string and trimpot?...a non-feedback bias circuit, keeping with Charles's theme.

d) Triple darlington output or option 4 with lateral mosfets. (If this is to be a non-global feedback design I probably would stick with the triple.

e) We would have to degenerate the diffs. and folded cascode to obtain the right amount of gain.

f) Front end to be driven by it's own regulated (non-feedback) supply.


Questions.

a) Can we get by without a servo to keep dc offset to a minimum?

b) Any objections to using mosfets for the diffs. to use higher fails or cascodeing the diff. pairs to attain the same goal.

c) Should we drop the differentials and use an input stage like above.

d) Who is going to layout the pcb once the proto is working.........




Regards,
Jam

P.S. more food for thought!

 

[bocka]

Hi Jam,

using JFETs in the diff-stage is a good design approach. Slightly
degeneration helps to prevent this stage from unstable operation.

When using lateral MOSFETs a Vbe-multiplier is a must. Vgs only works
with vertical MOSFETs. To my opinion a diode string and trimpot is not
a good solution, because Vbias depends much more from the tail current
than a Vbe multiplier does.

Option 4 has only two transisors in the signal path, a triple emitter
obviously 3... So I'd prefer the FET & bipolar output stage. The
output stage topology you've posted is very close to that what I'm
using. But I would never use bipolars without the emitter resistors. Thermal
stability is strongly dependant from the emitter resistors. Many
(especially the DIY) designs blow off because of thermal runaway.

Any suggestions about the current in the (MOSFET) driver stage? Currently
I'm using 150ma. This should be enough for up to 15 Amp ouput current

I like the differential input stage. You can obtain balanced operation
easily.

 

[jam]

Boca,

Thanks for your input. I think you need at least 100 ma. for the mosfets to sound good. I would like to know if you ran the bias past 200ma. if there would be any sonic benefit.

Bi-polars do not need such a high bias current I believe. It might be useful to compare the mosfet driver to the triple darlington. I suspect that Charles has already tried it and prefers the latter.

Joe,

I read your article and concluded that the A75 was a bit more complex for it's own good. I have built the A75 and got an improvement by removing half the outputs and increasing the current in the second stage. The A75 still hollds its own against the competetion and probably a lot of diyer's stayed away because of it's complexity. Your approach has a lot of merit but .....but what if you replaced the diffs. with TO220 devices upped the current on the circuit and ranit at 25 ma. as a true folded cascode.

Regards,
Jam

 

[Bricolo]

Hi Charles,

I've used the IRF610/9510 - not the IRF510/9610 as posted before - because they're layin around and I want to see if this topology works as expected. I'm really astonished how small the reverse capacitance of the BUZ900/BUZ905 is. First idea was to use the Renesas 2SJ76/2SK213 but they are suffering from a too small power dissipation. I'm biasing the MOSFETs quite high about 150mA @ 40V. Thanks a lot for the Magnatecs, I didn't take a look at the Cgd before although the datasheets are on my server...

I'm using a 2sk216 (same as 213, just higher voltage) in my szekeres headphone amp
it's biased in class a, at 125mA, no problem.

 

[Joe Berry]

Your approach has a lot of merit but .....but what if you replaced the diffs. with TO220 devices upped the current on the circuit and ranit at 25 ma. as a true folded cascode.

I suspect that it would work well, but I didn't explore it in the context of the article as I was trying to stay as close as possible to the original A75.

Today, I would also want to try the 2SK398/2SJ109 diff section that was suggested earlier, driving a ZVN3310/ZVP3310 folded cascode WITH current gain, with a NFB loop around just these two stages. The ZVs don't suffer too badly from capacitive nonlinearity, so they may work better here than the IR parts. With this setup, you could probably even get away with a single split supply for all three stages. I haven't built it (yet), so please treat this as just more grist for the mill.

 

[bocka]

Hi Bricolo,

if you are running a 2SJ79/2SK216 @ +/-30V and 150ma this will result in a worst case power dissipation of max 60V * 150ma = 9 W.

In the datasheet you can find a max power dissipation of 30W @25°C. If you get an case teperatur of about 90°C the max power dissipation must reduce to about 15W. Well with a small safety margin this should be enough although I would change to the larger MOSFETS like BUZ900P/905P or 2SJ162/2SK1058. This would be much more reliable.

 

[Bricolo]

Hi Bricolo,

if you are running a 2SJ79/2SK216 @ +/-30V and 150ma this will result in a worst case power dissipation of max 60V * 150ma = 9 W.

In the datasheet you can find a max power dissipation of 30W @25°C. If you get an case teperatur of about 90°C the max power dissipation must reduce to about 15W. Well with a small safety margin this should be enough although I would change to the larger MOSFETS like BUZ900P/905P or 2SJ162/2SK1058. This would be much more reliable.

My 2SK216 runs at 0-12V