The AD844 is the only CFA I know that has the Tz node brought out. I understand that the reason they do not do it normally is that it adds some 3-4 pF to the Tz node which then becomes the limiting factor on the hf performance.
Possibly a similar reasoning goes for VFAs.
jan
Hi Dave,
I'll see if I can discuss MIC in a bit more detail in the second edition of the book.
When I first did MIC around 1983 for my MOSFET power amp all I really tried to do was compensate the MIC loop for a ULGF of around 10-15 MHz, recognizing that the Miller cap was closer to a short circuit at those frequencies. I did have to take into account the effect of the resistor I used in series with the Miller cap. In essence, at high frequencies the MIC loop began to look like an ordinary closed loop amplifier with some amount of closed loop gain and open-loop gain. I then applied lag compensation across the differential outputs of the input stage.
I did not use SPICE at the time and did quite a bit of fiddling and experimentation to arrive at the values for the lag network.
In later efforts I used SPICE and also saw that in some designs it was desirable to also include a bit of lag-lead shunt compensation at the VAS collector. One way of looking at it was to think about where the current from the output of the VAS was going - how much of it was getting back to the LTP input node to establish the MIC loop gain.
It is important to recognize that the lead resistor in series with the Miller cap, if it is used, plays a role in establishing the effective closed loop gain of the MIC loop at high frequencies, and affects the loop gain of the MIC loop. That resistor is not necessary - I just often like to get some leading phase shift out of the compensation. Without that resistor, the MIC loop at high frequencies tends to move toward looking like a unity gain voltage follower. This will also result in a zero in the forward gain path.
I freely admit that there is a lot of handwaiving here and that the explanation and approach can probably be improved.
That was quite awhile back and even with SPICE I still tweaked around a bit. Nowadays I would do a more formal SPICE analysis of the MIC loop gain.
I have not attempted to do a formal mathematical analysis to establish the optimum values to use.
Cheers,
Bob
Thanks for the helpful reply.
The "ControlCentric" approach implies that it pole splits and that seems to be confirmed by Spice simulations.
Still don't understand the actual mechanism how it occurs.
Any ideas appreciated, can't believe MikeK, Waly, JCX and R.Lee don't have a comment on this between them.
Best wishes
David
I can find very little analysis of this so I may just have to think for myself.
The "ControlCentric" approach implies that it pole splits and that seems to be confirmed by Spice simulations.
Still don't understand the actual mechanism how it occurs.
Any ideas appreciated, can't believe MikeK, Waly, JCX and R.Lee don't have a comment on this between them.
Best wishes
David
Thanks for the helpful reply.
I can find very little analysis of this so I may just have to think for myself.
The "ControlCentric" approach implies that it pole splits and that seems to be confirmed by Spice simulations.
Still don't understand the actual mechanism how it occurs.
Any ideas appreciated, can't believe MikeK, Waly, JCX and R.Lee don't have a comment on this between them.
Best wishes
David
One thing that I usually scratch my head on is indeed the matter of pole splitting. Pole splitting does occur in MIC, but I believe that the two poles that are split are the poles on either side of the Miller capacitor, which in the case of MIC are the poles at the input to the IPS and at the output of the VAS. That leaves the pole in the middle between the IPS and the VAS. That is where we place the lag-lead shunt compensation for the MIC loop.
Cheers,
Bob
LTspice is absolute rubbish at generating publication quality schematics, and I've misplaced or lost my copy of Electronics Workbench Vs 5.12 which I used for the purpose.
LTspice is absolute rubbish at generating publication quality schematics, and I've misplaced or lost my copy of Electronics Workbench Vs 5.12 which I used for the purpose.
Simetrix is nice , with very readable schematics...provided
there s a minimum care in drawing the circuits.
LTspice is absolute rubbish at generating publication quality schematics, and I've misplaced or lost my copy of Electronics Workbench Vs 5.12 which I used for the purpose.
Are you sure you know how to draw schematic in LTspice, it's very simple and user fiendly?
Or use ExpressSCH, it's free, but not so user friendly.
I am not referring to drawing schematics; I am referring to creating bitmap images for publication of those schematics
Last edited:
@ michaelkiwanuka
Electronics Workbench is now Multisim, as i expect you know I also started out with a free version of EW that was given away on the front cover of an electronics magazine, back in the day, on a Floppy !
You must have missed this i posted a year ago Multisim 4 Free ! http://www.diyaudio.com/forums/software-tools/218542-multisim-4-free.html
Electronics Workbench is now Multisim, as i expect you know
I also started out with a free version of EW that was given away on the front cover of an electronics magazine, back in the day, on a Floppy !You must have missed this i posted a year ago Multisim 4 Free ! http://www.diyaudio.com/forums/software-tools/218542-multisim-4-free.html
@ michaelkiwanuka
The link i gave is for version Multisim 12.0
I've posted quite a number of screenies on here with version 2001 & they all look fine to me However, i did NOT use any MSim inbuilt image capture to do those, instead i used a seperate screen shot software to do them. You could too, & save them in .BMP format
The link i gave is for version Multisim 12.0
I've posted quite a number of screenies on here with version 2001 & they all look fine to me
However, i did NOT use any MSim inbuilt image capture to do those, instead i used a seperate screen shot software to do them. You could too, & save them in .BMP format In really Jurassic times, when I did all this Transfer Function stuff with complex algebra, I might have pontificated.I can find very little analysis of this so I may just have to think for myself.
The "ControlCentric" approach implies that it pole splits and that seems to be confirmed by Spice simulations.
Still don't understand the actual mechanism how it occurs.
Any ideas appreciated, can't believe MikeK, Waly, JCX and R.Lee don't have a comment on this between them.
But since I wrote my own linear circuit analysis package in the early 80's to help me with this (checking 5+ pages of algebra is not my idea of fun) I've always used to sims to get better insights into what's happening rather than analyse.
The only lapse from this was Cherry and my remaining brain cell has problems repeating this.[*]
Since then, I'm been very much like Bob, doing in sim what we did in 'real life'.
There's a lot to be learnt doing sims .. especially if you are lucky enough to have those 'Ah Ha!' moments when a sim suddenly explains something 'real life' you have been struggling with. It's then when you can improve your model and have slightly greater confidence that it reflects 'real life' .. but a healthy pinch of salt is always required.
Presently, I'm trying to find some time to look into these evil non-minimum phase bits that you and others have unearthed with a view to getting closer to them instead of banishing them to zillion GHz as has been my previous wont. SPICE world only ..
Dunno if this counts as your analysis.
To add something useful to the above garbage, I'll suggest that simple (?) analysis is 'accurate' when you have 1st & may...be 2nd order stuff but more than that, a 'full' SPICE model may be more useful.
[*] I think I did it in Lotus 123 but it was nearly 30 yrs ago. There's a problem doing it in Excel cos the way it iterates.
I want to work at a level that is as accurate as possible whilst still comprehensible.
Comprehension can reveal best areas to explore with Spice.
I really like R.D. Middlebrook's explanation of this. Definitely read it if you haven't already, then you can help me understand it better.
Best wishes
David
No publisher in his right mind would accept bitmaps for illustrations, unless it is the only option and it is paramount to print them.
Getting good pub quality illustrations is the single greatest time sink in publishing Linear Audio.
If you can actually print your schematics, graphs etc, use a virtual PDF printer to generate vector graphics. I use and recomment the free Bullzip PDF printer.
For graphs, use the data export facility if your program has it, to generate data tables (X, Y, Y etc), as ascii tables or excell compatible tables. There are many 3rd party graphing programs that accept these tables to produce crisp, pub quality graphs. I use Dplot which is relatively costly ($ 200) but is a publishers delight.
jan