John Curl's Blowtorch preamplifier part II

[janneman]

Originally Posted by godfrey

Yes, that is another way to do it, but you would still need two servos - one each for the outputs.

With JC's approach, one servo compares the average of the two outputs with ground and applies a common-mode adjustment while the second servo compares the two outputs with each other and applies a differential-mode adjustment.

Yes I understand, I was trying to see which method had an advantage over the other. Can't see it at the moment except the point Stuart made about the accuracy.


Also, JC stated that with separate servo's at each output you would lose the differential correction which I don't see either; if both outputs are controlled to ground, the differential offset is automatically controlled to zero.


jd

[AndrewT]

to avoid adding a cascode to reduce dissipation?

[john curl]

I don't recall saying anything is wrong about separate servos, I just don't do it that way. Whether one way is better, I am not sure. I am TRYING to explain the servos on the schematic.

[janneman]

Originally Posted by john curl

I don't recall saying anything is wrong about separate servos, I just don't do it that way. Whether one way is better, I am not sure. I am TRYING to explain the servos on the schematic.

No, but you said:

Originally Posted by john curl

[snip] When you are comparing the two outputs to ground, you lose the difference between the two outputs. [snip].

... which is not correct. When you are comparing (nulling) both outputs to ground, you are also ensuring that they are equal, ie no differential offset between the outputs.

jd

[john curl]

Jan, I was talking about one of the servos and how it is wired at the input where it floats and has no DC ground connection, except through the outputs themselves.

[john curl]

This is a situation where actual operation is very useful, like in a lab. If you remove one servo, the output will change in one way. If you change to the other servo and remove it instead, you see a different output change, or error. One is called a common mode error, the other is called a differential mode error. They are different in character, but they are both errors.

[scott wurcer]

Originally Posted by PMA

Yes. And you work in a most linear portion of the transfer characteristics and get lowest noise as well. Interestingly enough, most even well known designers run the JFETs at low idle in audio preamps. Like 2mA for a 10mA Idss device. Any good reason to make it so? I can guess the only one - part selection not needed then.

The gm is the same at 2ma (plus or minus a little) and noise is actually a fairly weak function of drain current (gm -> square root, noise -> fourth root). I can certainly see designs where input gm figures into the BW and stability issues. Not all circuits are open-loop.

[PMA]

But linearity of transfer characteristics is much better near to Idss. And even for FB designs there are no stability issues - depends on overall design. If you compare high and low idle, you get both lower distortion and better sound ()at high idle.

[john curl]

However, I use parts with Idss OVER 15 ma for the CTC Blowtorch. See everyone how difficult that is to copy?

[Terry Demol]

Originally Posted by john curl

However, I use parts with Idss OVER 15 ma for the CTC Blowtorch. See everyone how difficult that is to copy?

Only for the reason that 15mA is very high for a BL so very few would land
in that zone and V's are very hard to find - for us mere mortals

Having said that, I think 10mA BL's are fine - you will find the IP device
transfer curve is the least of your worries in the Blowtorch design. Other
non linearities will take precedence over the difference between a few mA
Idss IP devices, obviously provided they are all matched.

T