New Video: Basics of the Cascode transistor amplifier and the Miller Effect

[w2aew]

Just a short video on what a Cascode amplifier is, what the Miller Effect is, and how the cascode helps alleviate bandwidth limitations caused by the miller effect:

[HighVoltage]

Thanks for another great video.



 

[dentaku]

Cascode always seemed like a strange word to me and I never looked into what one was. Even though I will probably never use this because it's really only useful at high frequencies I'm glad I watched the video and learned something new.
Most of your videos have that "why didn't they teach it that way in school" quality to them. 

[SeanB]

Usable at low frequency as well, even at audio with a high impedance source, where the roll off might be a problem at even 100Hz.

[mswhin63]

I remember learning the cascode amplifier at Uni 3rd Year basic subject. Unfortunately it wasn't presented as well as yours. I was a pleasure to recap on the subject. 

[dentaku]

Usable at low frequency as well, even at audio with a high impedance source, where the roll off might be a problem at even 100Hz.

Yup, Alan mentioned this in the Youtube comments.

"The Miller effect is generally not an issue for audio amplifiers unless the source impedance is very high, so the cascode isn't used much for audio."

[Zero999]

It's good video.

I have one question about the schematic:
http://marcusjenkins.com/wp-content/uploads/2015/05/35dBAmp.jpg

Is there any reason why the input to Q3 is AC coupled? I've just plugged it into LTSpice with Q3 as DC coupled and it simulates fine and means R7 to R9 and C5 can be eliminated.

[w2aew]

It's good video.

I have one question about the schematic:
http://marcusjenkins.com/wp-content/uploads/2015/05/35dBAmp.jpg

Is there any reason why the input to Q3 is AC coupled? I've just plugged it into LTSpice with Q3 as DC coupled and it simulates fine and means R7 to R9 and C5 can be eliminated.

I'm not sure why Marcus chose to AC couple the emitter follower. Normally this is done to allow a more favorable bias point in the following stage, which seems unnecessary in this case.

[Zero999]

Here's the circuit I simulated. I'll try building it.

[T3sl4co1l]

I'm not sure why Marcus chose to AC couple the emitter follower. Normally this is done to allow a more favorable bias point in the following stage, which seems unnecessary in this case.

It gets 2/3 rather than 1/3 of the available supply, which will improve Ccb, therefore loading the gain stage less at high frequencies.  But yeah, over the apparent bandwidth this device is intended for, doesn't much matter.

Tim

[Jony130]

It's good video.

I have one question about the schematic:
http://marcusjenkins.com/wp-content/uploads/2015/05/35dBAmp.jpg

Is there any reason why the input to Q3 is AC coupled? I've just plugged it into LTSpice with Q3 as DC coupled and it simulates fine and means R7 to R9 and C5 can be eliminated.

Maybe because R7 to R9 and C5 form a bootstrap circuit to increase the follower input impedance.   

[G0HZU]

I watched the video and I think W2AEW/Alan's  technical description about Miller effect was extremely good

But I'm not very impressed with Marcus' application circuit that Alan was using as a demo for Miller effect. The best advice I can offer is to watch Alan's paper study again and again because it is very good indeed. But don't take the application circuit seriously as a broadband HF amplifier. i.e. this isn't a circuit I would recommend anyone to use as Marcus intends. It really isn't going to be much use as a general gain block with 50R output impedance even with the benefits of the cascode section.

It served its purpose to allow Alan to show the benefits of the cascode circuit wrt Miller effect but I don't think it's much use for anything else across the HF bands.

[Zero999]

It's good video.

I have one question about the schematic:
http://marcusjenkins.com/wp-content/uploads/2015/05/35dBAmp.jpg

Is there any reason why the input to Q3 is AC coupled? I've just plugged it into LTSpice with Q3 as DC coupled and it simulates fine and means R7 to R9 and C5 can be eliminated.

Maybe because R7 to R9 and C5 form a bootstrap circuit to increase the follower input impedance.

Bootstrapping only reduces the loading due to the biasing resistors. It does not increase the impedance above what it would be without them.

[commie]

Do you really want to know why cascode bjt blocks came about?, well the gm of a valve is about 1 to 5mA/V this similar to FET's , a common bjt such as the 2n3904 has a transconductance(gm) in the region of 80mA/V.Clearly, for radio front ends and I.F.'s BJT's produce too much gain bandwidth product so in an attempt to reduce gain radio designers(1960's) needed a way to reduce gain in BJT's and so the BJT cascode was devised.

If you try and use BJT's on your front end radio design, you will end up with an oscillating mess.

Cheers
Commie

[w2aew]

I watched the video and I think W2AEW/Alan's  technical description about Miller effect was extremely good

But I'm not very impressed with Marcus' application circuit that Alan was using as a demo for Miller effect. The best advice I can offer is to watch Alan's paper study again and again because it is very good indeed. But don't take the application circuit seriously as a broadband HF amplifier. i.e. this isn't a circuit I would recommend anyone to use as Marcus intends. It really isn't going to be much use as a general gain block with 50R output impedance even with the benefits of the cascode section.

It served its purpose to allow Alan to show the benefits of the cascode circuit wrt Miller effect but I don't think it's much use for anything else across the HF bands.

Thanks for the nice comments.  And, I agree.  Marcus' circuit may suit his particular application just fine, but it is not a good example of the general purpose RF/HF gain block - and it won't drive 50 ohms (you'll note I used the 10x probe to view the output).  I guess I was just a bit "lazy" and decided to use his application-specific circuit as an example for a more general purpose issue, rather than build up a breadboarded circuit myself.

[slateraptor]

Excellent content, as usual. Thanks for the refresher, Alan.

[T3sl4co1l]

Do you really want to know why cascode bjt blocks came about?, well the gm of a valve is about 1 to 5mA/V this similar to FET's , a common bjt such as the 2n3904 has a transconductance(gm) in the region of 80mA/V.Clearly, for radio front ends and I.F.'s BJT's produce too much gain bandwidth product so in an attempt to reduce gain radio designers(1960's) needed a way to reduce gain in BJT's and so the BJT cascode was devised.

If you try and use BJT's on your front end radio design, you will end up with an oscillating mess.

Um, increase stable gain*?

A cascode also removes one layer of Early effect, which isn't a big deal for BJTs (or the same thing in FETs, channel length modulation), but was of key importance for triodes, back in the day.

The removal of feedback capacitance is critical to improving the stability at high gain.  If the reverse-transfer coupling is greater than reciprocal gain (s12 > 1 / s21), you might not necessarily get an oscillator, but under practical conditions, with a tuned and matched source and load... yep, it's going nutzo.

You can get something like 30dB of gain from a 2N3904 at wideband frequencies (say 1MHz or so), but if the feedback capacitance is effectively taking the output and dividing it by, say, 24dB, then the extra 6dB goes back through the amplifier and oscillates.  Contingent on its not being absorbed by the input source, or phase shifted away, of course.

The most common way to address it is with neutralization: connecting an inductor in parallel with C-B.  Obviously, with the help of a coupling capacitor (Xc << Z), and usually with a generous resistance (R ~= Z) to stretch bandwidth, at the expense of gain.

Back in the day, feedback capacitance was addressed by placing extra grids between grid and plate, shielding their electric fields.  This went to quite a meticulous end, as tubes like 6AU6 achieved extremely low feedback capacitances: on the order of 0.002pF (relative to input/output capacitances on the order of 10 and 5pF).  This allowed extremely high gain per stage (mainly limited by the impedance achievable on the tuning coils), 40dB being easily achieved.

I'm not sure if any semiconductor devices have yet come along which match the feedback performance of pentodes, but MOSFETs of similar construction (namely, dual-gate FETs with the upper gate grounded, effectively making a monolithic cascode!), and many RF-voodoo parts (PHEMTs, HBTs, etc.), do achieve very low feedback capacitances, which allows them to provide very good gain, at much higher frequencies (like, 30dB, from a single component, at 5GHz!).

Tim