The sound of cascodes

I have tried cascodes. I have tried them in input differentials, in driver stages, in line stages, in powerstages. All with the same result - an initial luring illusion that the sound is better, cleaner and more focused because the resolution seems better. After a week of listening things have always changed. The problem is that the amp is not silent when it should be. The sound of a cymbal clash has two distinct phases - the initial strong clash transient followed by the decay phase with its very rich harmonic content. With cascodes the clash transient becomes exaggerated and extends well into the decay phase, masking the instrument´s natural timbre. The sonic signature of this transient behaviour is very fatiguing, a bit like TIM.

I´ve spent hours thinking on why the sonics does not correlate with the theory, it should be better but it is not. From my experience, good interstage linearity always makes good audio performance. The cascode is a compound circuit build with TWO transistors. The question is if something about the interstage linearity affects the sound. I am pretty sure no one would seriously use a stand-alone CB stage as a gain element in a power amp front-end because the local overload margin would be next to nil (due to the high gain). But we hook it up with a CE-stage collector and figures are now great. Question to the board: what about the interstage linearity ?

I once tried to "improve" a very good power amp by cascoding the input differential but the audio magic was now gone and after just 10 minutes of listening the stuff was lifted out again. It just doesn´t work for audio AND I DON´T KNOW WHY. It is annoying that I cannot explain the reason in technical terms.

"You are wrong - cascodes are better" is not the reply I seek. I have heard those words many times already but the sound still does not change to the better. It would be much interesting to hear someone explain technically why transients seems to be mangled in a cascode stage.

Kind Regards

Syl
 
Cascode Quandry!

Have you tried running more current through the cascoded stages?

How did you bias the cascodes? Did you use resistors, zeners or current sources. Did you put a capacitor from the base of the cascode device to ground?

Did you reference the cascodes to ground or the emitter of the cascoded device?

You raise some good questions but raise a few others and maybe we can come up with some answers.

Jam
 
Should have known Jam would get here before I did...
I am suspicious of cascodes and use them only when their use allows me to dispense with another gain stage later. Yes, they are better on paper, but that doesn't always pan out in the listening room.
Jam's points are valid, however. There are a million variables that can be toyed with to improve the performance of a cascode. By all means, jiggle the parameters back and forth.
But do you really need all that gain? You're just going to have to burn it off later with NFB or whatever. There are times when a cascode is just the ticket, but they aren't as common as many people suppose.

Grey

(Jam and I can jog each others' elbows about using resistors vs. current sources 'till the cows come home. I prefer resistors [in most cases--current sources do have their uses] and he prefers current sources. I feel pretty much the same way about current sources that I do about cascodes...use only when absolutely necessary...and that ain't often.)
 
Cascodes

Syl,

Interesting comments on the sound. I followed it clearly; decay is a big issue, it's one of the final frontiers to get it right, but it should not be the main event.

There are four reasons I don't go for cascodes. One is complexity/tube count, two is floating the upper filament, which often creates noise, three is the very high impedance output, and four is lousy PSRR. If you want more gain, use a different tube, or cascade two medium mu tubes, preferably with direct coupling - and much lower Zout, which at these levels of signal is always required to drive output tubes anyway. And I have to say that I don't like mu followers either (although Zout and PSRR are not issues with this topology).

I too have often wondered why cascodes are not always good performers. Perhaps, like cathode followers, they are hypersensitive to operating point. But one thing has always worried me; the huge impedance mismatch between the drive of the bottom tube and the input of the upper tube. Impedance mismatches are often the source of many problems, and this could be it.

It's strange that most good designs are very simple, with plate loaded triodes and the odd cathode follower. :D


Great post!

Cheers,

Hugh R. Dean

www.printedelectronics.com
 
Syl and all,

1st,
thanxalot for the great input ! :)

2nd,
had the opportunity to listen for an extended period to two preamps suffed with cascodes. One was single-ended, the other was fully differntial. Cascodes were hybride: bottom MOSFET. top tube.

While those preamps, designed by a dedicated cascode lover, were the best HIFI i ever had in some respects, it was not music to my body and soul. I suffered long-term listening fatigue with both preamps, it went that far that i preferred to have no music playing in the background (remember my last posts in the "Going Pro" thread ? :) ). I was considering building that differntial preamp for me but to incorporate some own ideas and i wondered how close i should stick with the original design; so i discussed a lot of toplogogy issues with my friends and cascode-or-not was the most important (2nd being fancy output CF or OPT).

Both friends, one a tube wizard, the other a SS guru with tube experinece, complained about "spiritual" nastinesses of cascodes i also felt uncomfortable with (and it turned out i could describe their unease better thant hey themselves could). We agreed sonics were sterile, not involving, artifcial, excellent for the brain but not for the heart. It was very interesting for me to find the SS guru (hifidaddy) to complain about the same in SS circuits i had to complain about in hybride in pure tube cascodes.

I do not condemn cascodes at all, I see it like Nelson Pass, they are another tool in the box.. They just seem not to sing for me. And my preamp design ended up in having neither cascode nor CF (but a lot of iron, personal taste). And i found it quite a challenge to get the gain low enuff and still maintain the count of stages needed for a proper split phono EQ. Two choke-loaded long-tailed-pair gain stages together withan 1:12 MC input transformer too easily have the total gain i need :) .... a cascode would be the last i need here.
 
Thank´s for all replies.

It seems like some people use them only if they have to. I too use one in my 100W power amp. I use fet´s for input and they cannot cope with 53 Volts. This amps drives my woofer so only bass matters.

Jam,
The way I use them I want to maximize the local feedback of the CB stage, so I use non-bypassed resistors connected to the base. The reference is ground, not the emitter of the CE-stage. In your view, what is the sonic difference between referencing to ground and referencing to the emitters of the CE-stage ?


Regards

Syl
 
Re: Cascodes

Hi Hugh.. You wrote "There are four reasons I don't go for cascodes. One is complexity/tube count, two is floating the upper filament, which often creates noise

three is the very high impedance output, and four is lousy PSRR. "

I think that this is not a drawback. High impedance output => Less interference from the the BJT/FET on the signal. It is only a "linear" resistor who is the output impedance not a "nonlinear" transistor.

Lousy PSRR .. I think this comes from a pour base/gate drive of the Cascode BJT/FET ?? I cannot see another way of entering the circuit.

As JAM writes "How did you bias the cascodes? Did you use resistors, zeners or current sources. Did you put a capacitor from the base of the cascode device to ground? "
This could possible effect the PSRR!?

I know you where talking about tubes, which i do not know anything about. ;)

I would as the second/third person in this thread agree with Nelson "It has been my experience that cascoding does not always improve the sound. It's just another tool in the box."

Sonny
 
Sly,

Non-bypassed resistors could lead to 'bounce' of the cascode base voltage and possibly modulate the signal with noise from the supply rails, assuming you are feeding the voltage divider from the supply lines.

I find that if you reference the cascode to the emitter the sound seems to be dynamically compressed, this is most apparent when you cascode the output stage, though this method measures the best. ( Grey's rule of what you measure is not you hear )

I prefer to use a zener and current source to derive my cascode voltage and bypass the zener with a small capacitor to squelch any zener noise (Too large a capacitor and you might hurt the sound again). Another way is derive your cascode voltage from a low noise three terminal regulator.

If you use mosfets as your cascode device don't forget your gate resistor even if you don't have any instability, a base resistor sometimes helps when using high bandwith bipolars. I had to mention this to please the TDR twins Harry and Jocko and not summon their wrath........heh, heh.

Jam

P.S. I see Grey has come out of hiding again!
 
Here's a bone!

Syl,

Thank you for your comments; yes, I was talking of tubes, certainly, but I have generally always taken care to tie down the cascode base pretty firmly.

I might ask you; what sort of compensation do you use on your cascodes? How many pF, and where?

After hammering the SS cascodes :D let me throw an unusual one into the ring........

In 1964, in Britain, Christopher Rush did some pioneering work on a fascinating cascode which later came to bear his name but which no-one ever seems to talk about these days.

In Rush's cascode, the signal enters a lower PNP base, collector is grounded. Emitter of this transistor is connected to the emitter of an NPN, and the base of this second transistor, the upper, cascoding device, is held at a constant voltage above ground, usually around 2.2 volts. A CCS feeds the upper transistor collector, from whence the signal is extracted.

Five years ago, I experimented with this configuration, and built an output stage for a CD player using it. It ran at 8mA, and features a Zout of 3 ohms! (I'm listening to it now, actually!) Of course, it uses very heavy feedback; the Rush Cascode is used in a conventional Bailey topology in place of the voltage amplifier, and the circuit is single ended throughout.

The standouts of this topology are meteoric OLG, and DC to light frequency response. Of course, stability is a big issue, and takes a lot of work to guarantee.

However, thought you guys might be interested!

Hugh

www.printedelectronics.com
 
Could it me that throwing a cascode into any amplifier is not such a good idea. Its likely to change the open-loop response and hence feedback levels. If you cascode a common emitter stage the miller effect will be eliminated and this could easily upset the feedback levels and operation of the amplifier.
 
Re: 3020 phono stage

For reference here is the NAD 3020 phono section map... perhaps someone can explain the assymmetrical rails.

dave

Schema_3020_phono.gif
 
Rush cascode????????

"In Rush's cascode, the signal enters a lower PNP base, collector is grounded. Emitter of this transistor is connected to the emitter of an NPN, and the base of this second transistor, the upper, cascoding device, is held at a constant voltage above ground, usually around 2.2 volts. A CCS feeds the upper transistor collector, from whence the signal is extracted."

Sounds like a diff pair with complementary devices instead of same type (as shown in NAD schematic). Neither one of which is a cascode.....which is what this thread was actually about at one point.

Jocko... it is RMFM in this case. (read my freaking mind)

Cheers,

H.H.