how to mantain a constant gm in a resistive loaded diff pair

[jayanth.neo]

Hi guys...

I would like to know about the variation of gm with temperature and ways to compensate for it. In a resistive loaded diff pair, if the gm reduces with temperature, is there a circuit technique that can sense it and maintain the gm fairly constant?

Suggestions on this would be very much appreciated.

Another suggestion would be on how we can increase the temperature coefficient of a resistor. For eg: if we have a resistor with a temp co of +1.5, what method will increase the temp co to some other desired value greater than 1.5?

 

[LvW]

Re: how to mantain a constant gm in a resistive loaded diff

Hi guys...
I would like to know about the variation of gm with temperature and ways to compensate for it. In a resistive loaded diff pair, if the gm reduces with temperature, is there a circuit technique that can sense it and maintain the gm fairly constant?

In general, this should be possible - of course within certain limits and with limited accuracy - but this is true for each regulated device.
If you like to hold the gm at a constant level, that means that the gain should be constant. Thus, it is in general a task which can be solved with an automatic gain control circuitry. As you can sense everything you also can regulate everything (nearly ?) with an appropriate control loop.
But the question arises: Makes it sense ? Is it economical ?
I really have some doubts.

Perhaps the best idea would be to establish a temperature control loop to hold the temp at a constant level ?

 

[jayanth.neo]

Re: how to mantain a constant gm in a resistive loaded diff

Hi guys...
I would like to know about the variation of gm with temperature and ways to compensate for it. In a resistive loaded diff pair, if the gm reduces with temperature, is there a circuit technique that can sense it and maintain the gm fairly constant?

In general, this should be possible - of course within certain limits and with limited accuracy - but this is true for each regulated device.
If you like to hold the gm at a constant level, that means that the gain should be constant. Thus, it is in general a task which can be solved with an automatic gain control circuitry. As you can sense everything you also can regulate everything (nearly ?) with an appropriate control loop.
But the question arises: Makes it sense ? Is it economical ?
I really have some doubts.

Perhaps the best idea would be to establish a temperature control loop to hold the temp at a constant level ?

Hi LvW,

Thanks a lot for the reply. I perfectly understand what you say. We can indeed sense everything and regulate other parameters with ofcourse a few drawbacks. There is no free lunch.

I believe I should explain the reason why I am asking such a question.

I have a 3 stage equalizer, each being a resistive loaded diff pair with 7 set variable gain controlled by source degenerated resistors and switches.

The problem is that the gain of each stage and hence the total gain reduces with temperature from( temperature swept from - 25 to 125). Temperature control loop can definitely keep that temp at constant level, but this is more like a PVT variation issue and hence must be addressed.

In my diff amp, I have a positive temp co resistor and the gm has a negative temp co. . The gain is more influenced by the gm and hence reduces with temperature.

So from a crude observation, I see that I either have to control the gm to a constant value or equate it to cancel out the temp co of the resistor and then keep my gain fairly constant over temperature or to further increase the temp co of the resistor to cancel out the effect of gm.

either way my ultimate aim is to maintain the gain of each stage and hence the total gain constant over the temp.

Hope you get to understand the situation.

Now, do you see some way to do this?

Thanks anyway for being the first to reply.
I really appreciate it.

 

[nowatt]

Re: how to mantain a constant gm in a resistive loaded diff

You should be able to find 'constant gm bias' in most
MOS analog design text books, or do a web search.
The technique varies the bias current through the
transistor to keep the Gm same as 1/R (where R is
the same type as your load/degen. It is a simple circuit and works quite well. Downside is that DC bias points can change quite a bit over PVT and give headroom issues . (Also body effect can mess it up a bit)

Added after 5 minutes:

Looking further i see the schematic on this board posted 11 jan 2009
revoh Constant Gm-R topology comparison

 

[FvM]

Re: how to mantain a constant gm in a resistive loaded diff

I think, the most simple solution for an overall temerature compensation of diff pairs, if thermostatic control isn't applicable, would be a variable amplifier bias current. For a DC coupled cascade, a compensation of the shifted output common mode level may be necessary. As a major advantage, a single voltage can control the gain of many amplifier stages.

If this suggestion doesn't fit, an individual gain adjustment by PTC is another option. Pure metal PTCs (e.g. Pt100 or Pt1000, Ni, Cu resistors) have a rather exact ~T characteristics, that perfectly matches BJT Vt. Silicon PTCs have a higher TC, but can be adjusted with series/parallel resistors to achieve an arbitrary linear TC over a wide temperature range.