2
\$\begingroup\$

Razavi's textbook Design of Analog CMOS Integrated Circuits (2nd edition) includes a possible option of improving the slew rate of a folded-cascode amplifier in Figure 9.77, included below (along with the rest of the amplifier schematic). As you can see in 9.77(b), one option is to have two diode-connected transistors with one end tied to VDD.

The reference cited in Razavi's text for this is another well-known book, Analog Integrated Circuit Design by Carusone, Johns, and Martin. Carusone's book illustrates these clamp transistors in Fig. 6.20 (below) and describes their operation in Section 6.4.2. A difference here is that in Carusone's book the diode-connected clamping devices have their drains attached to the gates of the bias current sources rather than to VDD like in Razavi's book.

A couple things I'm wondering about:

  1. Are the approaches equivalent (diode device drains at gate/VDD), or is one an error?
  2. How common is the approach of having these clamp devices in general? I've been trying to find a published work which describes designing a folded cascode OTA with these clamp devices included but I haven't been able to. Interestingly, the reference cited in Carusone's text for these devices isn't a paper but actually a "private conversation" with an individual "S. Law" at Xerox Corporation in 1983.

folded cascode razavi clamp circuit carusone clamping

\$\endgroup\$

2 Answers 2

2
\$\begingroup\$
  1. It seems to me that both options of connecting the diode to VDD and to the bias line are possible.

    When connecting to VDD, the clamping voltage on VX, VY will be around VDD-VTHN.
    When connecting to the bias line, the clamping voltage on VX, VY will be around VBIAS-VTHN.

    So, the clamping voltage is different in both cases. You could choose one that works for you. Basically, you need to ensure that there is no current in the diode at the DC operating point and current has to flow through the diode only during slewing to limit VX/VY to the value you want. When connecting the diode to bias line, the bias line gets disturbed and current Ip increases. I don't see an issue with this but, if Ibias1 is too low, it could cause a settling issue.

  2. Diode clamps are very commonly used for voltage limiting but in the specific context of a folded cascode, I have not come across the same so far.

\$\endgroup\$
2
  • \$\begingroup\$ Thanks a lot for the response. On the first point, it seems that if you set up the diodes so that the clamping voltage is VBIAS-VTHN, the current through the diodes has to come through the current mirror device Q11 (Fig. 6.20) rather than directly through the power supply. Are there any potential issues with this? It seems that the Q11 current (and its VGS) are mostly fixed by the current source Ibias1. \$\endgroup\$
    – Halleff
    Oct 19, 2023 at 18:29
  • \$\begingroup\$ Yes, like I mentioned, since the current through the diodes will come from Q11, it will disturb the VGS and hence increase the current IP. Once this transient event is over, for VGS to come back to its original value, it might take sometime particularly if Ibias1 is a very low and if there is a high capacitance on the bias line. You'll have to see that in your simulation and see if that is acceptable. \$\endgroup\$
    – sai
    Oct 20, 2023 at 5:31
1
\$\begingroup\$

How common is the approach of having these clamp devices in general?

the reference cited in Carusone's text for these devices isn't a paper but actually a "private conversation" with an individual "S. Law" at Xerox Corporation in 1983.

These clamps may have helped in older, less-controlled CMOS processes of the time. Today, the capacitance they add may well be detrimental overall. This is process-dependent. Unless someone with lots of IC design experience can chime in with more details, I'd say such clamps are a tool in the design toolbox, and can be used where they help, but they are not a universal approach to speeding things up.

A sufficiently fast op-amp will just keep up with the inputs, without need for clamping internal nodes. A slower op-amp may perform better by adding some slew rate limiting - perhaps as simple as a series resistor - on the input rather than having internal clamps.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.