Cascodes- the truth is out there...

A while ago I showed a bootstrapped cascode of the input stage:

http://www.diyaudio.com/forums/showthread.php?postid=209401#post209401

and somewhat relevant discussion followed over several pages

by bootstrapping the diff pair cascode reference to follow the input common mode and maintain a constant Vcb on the input Qs some additional distortion reduction can be seen in sim – the effect is much more pronounced with the higher nonlinear Cgd of low noise fets as input Qs

given the wide separation between parasitic oscillation and audio frequencies I’m sure some R-C damping could control rf parasitics without eliminating the audio frequency improvement
 
JCARR
"frequently referencing the biasing to a fixed point is easier to implement and provides the majority of the benefits (subjective listening as well as objective measurements)."

As the proportion of modulation to voltage Vce is higher when the bias is fixed than with no cascodes, I have doubts about a reduction in distorsion but there is a reduction in variations of dissipated power. Many peopple disagree about listening improvements.

Douglas Self says input stage cascodes do nothing to reduce input stage non-linearity. However, I think he refers only to fixed bias schemes.

~~~~~~~~ Forr

§§§
 
Ditto what JCarr has said:

Putting a cap on the base of the cascode element is usually trouble. You don't need fancy equipment to sort this out, though. A simple square wave will show little aberations on the rising and falling edges. A decent 100 MHz 'scope will suffice for this.

One way around this problem, favoured by a former member here, is to use a MOSFET for the top element. You can RC filter the reference, and put a 1K or so resistor in the gate lead.

Jocko
 
Forr:

I did the experiment years ago, building both fixed-bias cascode and bootstrapped cascode versions of the same input stage and comparing the measurements. There was a distortion reduction with both, but greater with the bootstrapped version. And as jcx mentioned, to some extent the results depend on the choice of input device. Today, I use bootstrapped cascodes when I can, but won't lose any sleep if I have to resort to fixed-bias.

Jocko, I've had cascodes oscillating at 250~300Mhz. Sometimes a 100MHz scope is sufficient, other times it isn't.

Greg, I've likewise experimented with biasing the cascode with the output of the driver stage, and also using the output stage signal. This was for a 50W amp, and my goal was to make the distortion curves as uniformly low as possible for 1kHz, 10kHz, and 100kHz. I found it useful to add an adjustment pot so that you could monitor the distortion and dial in the optimum bias point. With adjustment, 1kHz, 10khz and 100khz all were reduced to below 0.0002% at any output level up to clipping. Without the adjustment, the distortion levels at the respective test frequencies didn't line up nearly so well.

hth, jonathan carr
 
Hi Hartmut: Long time no speak!

>When I tried to built up a Connoisseur style preamp many years ago, with lots of helpful input from your side, this might be one of the reasons I failed.<

Could be. I wouldn't expect any circuit to sound at its best if it were oscillating! 😀

>The other was using resistive loaded emitters in the folded cascode instead of using current sources.<

Ideally it should have been operating as a current see-saw, but you don't get that behavior if resistors are used instead of current sources.

>While I know your feelings about Kaneda circuits, I think that these are less demanding for the DIY builder, when it comes to hit or miss.<

Except for Kaneda's propensity for using unobtanium semiconductors, you may be right. 🙂

best, jonathan carr
 
HI JCarr,

Do you have some numbers to compare distorsion when using no, fixed and bootsrapped cascodes ? I agree it may depend on input devices, but also on other parameters like frequency (a cascode scheme may better at some frequenices and worse at others) and bias voltage.

Self says that the linearity of a cascode can be improved by having different current in the CE and the CB devices (I suppose he means more current in the CB one).

Another scheme worth of interest but rarely viewed is the parallel cascode where the CE and CB are of different polarity. It needs a resistor or CCS load for the CE where the current is sensed by the emitter of the CB. I wonder if this scheme is as good as the series cascode.

~~~~~~~ Forr

§§§
 
good to see more refs online, but Hawksford is a little late for priority - Baxendall seems to have beat him to it by 20+ yrs:

http://www.diyaudio.com/forums/showthread.php?postid=159617#post159617

so I prefer to call the topologies in the Hawksford paper Baxendall Super-Pairs after the reference in Heinlein and Holmes "Active Filters for Integrated Circuits"

(the Heinlein book should be subtitled "prior art" - it has 1326 references covering op amp internals and applications up to its publication date of 1974)