[SOLVED] Constant Gm vs Constant Gain

[pooh_bear]

Hi, I was wondering if designing a circuit for constant, flat, gm means it has to be designed as well for constant,flat, gain. And if the isn't kept constant rail to rail, then what is the industry set standard on the quality of the changing open loop gain? (THD+N?). Also are there any sort of data in the data sheet of opamps that at least indirectly tell us the how much the open loop gain of the amplifier changes with input common mode level? I can't seem to find graphs like output voltage accuracy(which would reveal gain) vs. input common mode level in the analog devices data sheet. Anyway, any help is greatly appreciated.

 

[saro_k_82]

Hi, I was wondering if designing a circuit for constant, flat, gm means it has to be designed as well for constant,flat, gain.

Not exactly. By having constant-Gm, a constant closed-loop BW is targeted. This keeps the transient response unchanged with common-mode / temperature. This relaxes the design for compensation as well as the timing requirements.
Constant open-loop gain is never a highly controllable spec and constant closed-loop gain is accurately achieved with precision passive components.
THD is a tricky spec here. If the input frequency is well withing UGB, then the loop-gain would be too high to suppress it's own variations, and if the frequency is close to UGB, the higher harmonics would be filtered automatically. Aol variation with ICMR would be available as well as Aol variation with frequency. You could work with these to get these numbers for your application.

 

[pooh_bear]

Not exactly. By having constant-Gm, a constant closed-loop BW is targeted. This keeps the transient response unchanged with common-mode / temperature. This relaxes the design for compensation as well as the timing requirements.
Constant open-loop gain is never a highly controllable spec and constant closed-loop gain is accurately achieved with precision passive components.
THD is a tricky spec here. If the input frequency is well withing UGB, then the loop-gain would be too high to suppress it's own variations, and if the frequency is close to UGB, the higher harmonics would be filtered automatically. Aol variation with ICMR would be available as well as Aol variation with frequency. You could work with these to get these numbers for your application.

Oh thank you so much for the explanation on constant gm and THD. I was frustrated trying to get the gain to be extremely flat for a rail to rail input amplifer but always had like 6 db fluctuation towards the rails.

 

[Junus2012]

for the closed loop amplifier, I think that there is no problem to work wit variable gm if we assuring the stability but this will be in the cost of the higher compensation capacitor so less GBW and less S.R or we need more power to maintain it. However the gain of the closed loop amplifier is dependent upon the passive f.B network that would be a constant value not changing as gm changes.
But the accuracy of the closed loop amplifier is depending on the open loop gain which is in your case varying with common mode voltage, it will be more accurate when the gm is maximum. Since we are assuring the accuracy with minimum gm then the closed loop output is not going to be effected.

 

[saro_k_82]

Oh thank you so much for the explanation on constant gm and THD. I was frustrated trying to get the gain to be extremely flat for a rail to rail input amplifer but always had like 6 db fluctuation towards the rails.

6dB variation in gain with Common mode is expected naturally in R2R ip stages. If you want to make the gm constant with CM you have ways to do it, but first convince yourself that you cant do without it as it is going to cost you on other specs and effort.

 

[Junus2012]

Saro_k is right, because the variation of 6db (double gm) is not critical in the closed loop, the circuit that make it constant is not easy easy or even full accurate

 

[pooh_bear]

Oh thank you guys. So I think I get it. Open loop gain can and will fluctuate a lot(process variation, change in output resistance), but as long as my gain is high enough for desired dc voltage accuracy, and if the constant gm circuit can maintain constant gm to produce a constant gain bandwidth because GBW = Gm/Cl, then I shouldn't care how much my gain fluctuates with input common mode.

@saro_k, It's funny how you said "If you want to make the gm constant with CM you have ways to do it, but first convince yourself that you cant do without it". Constant gm is so cool, you can use the entire power supply : ). J/K, You're right. It's probably not applicable or worth it to do RTR constant gm amplifiers to put in applications like ADCs , but I wanted to make one for practice and understanding.