A tale of two op-amps: A Hitch-Hikers guide to op-amp specs

A few weeks ago we covered the Touchstone Semi TS1001 ultra low-power and low-voltage op-amp. Ultrasounder commented:

Looking at the datasheet, the key [specs] reveal a very lousy, CMOS input-output opamp which is not speced for anything beyond audio. The GBW is 4KHz which means that the max frequency this thing can reliably amplify without clipping/distortion is right around 2KHz. This was my gripe when I had a phone chat with the marketing guy also running this competition…

We were intrigued and asked for a guest post with more info. What we got back was a brass-tacks guide to evaluating op-amp properties from a datasheet. Find out what it means to have a lousy, CMOS op-amp, and when it’s useful. Thanks for guest posting Ultrasounder!

Stealth mode startup Touchstone semi made waves by announcing a “Nano Watt Analog” that claimed the industry’s lowest active power consumption. The supply rail can be as low as 0.85V, and consuming a measly 5uA over its entire supply range, which tops out at 2.5V. The idea of operating an Op-Amp at 0.85V is intriguing enough and solves an age old problem of an Op Amp not having enough headroom when the CR2031 button cell is drained down to a 1V.

Now, how is that the TS1001 is able to operate down to 0.85V? The answer lies deep in the fabrication process technology used in the Op-Amp. One could only, guess that the Op Amp shares the same sub-nano meter technologies that is used in many of the latest generation Digital ICs, such as flash memories and MCUs that are able to operate down to a 1V at sub micro amp currents. Lower the gate threshold voltage, lower the supply voltage. On the flipside, the upper limit on the supply for the TS1001 is 2.5V.

Now how does the TS1001 really rank up to its immediate cousin its traits? Let’s do quick apples-apples comparison with another Ultra Low power Rail-Rail Op Amp from TI. The TLV2401 operates from a single supply from 2.5V to 15V.

DC Characteristics

Input Offset-Voltage

  • TS1001      0.5mV
  • TLV2401 (2.5V) 0.3mV

The input offset voltage is created due to the bias current mismatches in the input differential pair. All transistors are not made the same and any asymmetry in the transistor devices (Channel length for instance) is going to create imbalance in the bias currents mismatches which appear as offset voltages on the inputs.

How to test this? Short the input terminals (+ and -) and you should not see any significant DC voltage on the output. If at all you see something that is the input offset voltage.

What it means to you. Well, if you are measuring a low level DC signal any significant offset voltage can completely swamp out the measured signal and that doesn’t do any good. So this is a significant DC parameter to keep in mind. And the TS1001 has a higher offset voltage than its closest neighbor. So, remedy would be use very high impedance on the inputs for gain setting resistors. The higher the precision of the resistors, the better (0.1%).

Input offset voltage drift

  • TS1001    20uV/C
  • TLV2401 3uV/C

This is the overall shift in the offset voltage over temperature. If your application needs good temperature stability, then the TS1001 has a 20uV/C. The TS1001 is fully specified at 85C, so by that time the input offset would be 1.2mV.

Input Bias Current

  • TS1001 (25C) 10pA
  • TLV2401       100PA

The input diff pairs and other circuitry needs current to be biased in saturation region. This bias current is supply independent and is generated within the chip by constant current sources. But if the Diff pairs are not symmetrical due to inherent process related mismatches, there could be currents flowing from the supply which contribute to the Input offset voltages that we talked about. Because, of the small geometry CMOS transistors used in the TS1001, it has lower input bias current which is a good thing.

Where low-noise is important, good old Bi-Polar input Op-Amps are preferred. CMOS transistors, like those in the TS1001, tend to be noisier than Bi-Polar transistors. Why have companies started using CMOS technology instead of Bi-Polar for the input stages?

Bi-Polar transistors require base current to be in the linear region and that translates to higher bias currents. Not great for low-power devices. JFET input based Op Amps have even smaller input bias currents but require higher supply voltages. Of the available technologies, CMOS transistors have the lowest power and supply voltage requirements. That makes it the obvious choice for a part like the TS1001.

While CMOS transistors have some benefits for low power applications, they tend to be noisier than Bi-Polar transistors. And where noise performance is of paramount importance, Bi-Polar input Op-Amps are a better choice.

Input Supply Current

  • TS1001 (0.6uA)
  • TLV2401 (0.88uA)

The TS1001 is a clear winner here for the obvious reason that it consumes 0.6uA at 0.85V (according to the datasheet). Very suitable for low power embedded applications running off of a single cell CR2032 battery. A worthy addition to a MSP430 Value line series as the Value line lacks an on chip Op Amp.

CMRR

  • TS1001     74dB
  • TLV2401 120dB

This is a very significant spec to me when I am trying to design in an Op-Amp. This is the ability of an Op Amp to distinguish between any signals that are common (noise and offsets) to both its inputs and only amplify the difference. In other words, if there is a signal that is common to both the inputs, if the CMRR is higher then, that noise gets rejected and the difference between the two terminals is amplified faithfully based on the closed loop gain. High output impedance sensors, such as Piezo elements tend to have a low signal to noise ratio and hence need a higher CMRR Op-Amps to improve the dynamic range of the measured signal. This is very much application dependent and the TS1001’s CMRR is inconveniently low, but with proper buffering can still be used with high impedance sensor nodes.

PSRR

  • TS1001 74dB
  • TLV2401 120dB

This is the ability of an Op Amp to reject any noise associated with its supply. If any high frequency ripple is injected in to the Op Amp via its supply rails, which will modulate the output with this high frequency envelope thereby degrading the performance of the Op Amp. Bottom-line, please do yourselves a favor by avoiding switchers for Op Amp supplies and always use a Low Noise LDO. That definitely improves Op Amp noise performance. Also, it always a good practice to provide high frequency bypassing to the Op Amp supply rails right on the pins.

AC Characteristics

UGBW

  • TS1001 (RL= 100KOgm, Cl=20pF) 4 KHz
  • TLV2401 (RL=500KOhm, CL100pF) 5.5 KHz

This refers to the bandwidth over which the Op Amp continues to be an amplifier and is faithfully able to amplify input signals. The Op Amp exhibits a dominant pole response, very similar to the frequency response of a low pass filter. A DC gain that starts out higher and gently rolls off at 20dB/dec finally crossing 0dB at a particular point in frequency.

What it means to you- It is a rule of the thumb that the Gain Bandwidth of the Op-Amp so chosen for a particular application should be at least 10 times the highest frequency component of the input signal.  Let’s get real. Let’s say you are building your next guitar effects processor for your electric guitar. If you wanted to sample the raw guitar signal at Nyquist rate, then the effective sampling frequency would be twice the max Audio bandwidth of 20 KHz which works out to be the magic number 44.1KHz. In order to keep the high frequency hash from folding over in to your pass band you need an “Anti-Aliasing” filter, usually a Low pass filter with a very sharp roll off from pass band to stop band with minimum pass band attenuation.  The cut-off frequency for the filter is 20KHz. Would the TS1001 work in this case? Of course not, because of its Gain Bandwidth is 4KHz.  This is an important number to overlook in the datasheet and could make or break your design.

Phase Margin

  • TS1001 70
  • TLV2401 60

Without getting in too much of control theory math, let me just surmise this spec by saying that for an Op-Amp to be stable in closed loop (you wouldn’t want to use one in Open loop anyways!!), its Phase Margin should be sufficiently high at the instance when its gain is one. In fact, typically this is tested by wiring up the Op Amp in a unity gain follower configuration. How high is high. Conventional wisdom says, 60 Degrees. Our TS1001 is golden as it has 70 degrees.

Slew Rate

  • TS1001 (RL=100KOhm) 1.5V/ms
  • TLV2401 (RL= 500KOhm, CL=100pF) 2.5V/ms

The rate at which the Op Amp’s output tracks the input. Any delay in doing so, would induce phase shift in the signal amplified. It is typically specified at a Resistive/capacitive load combo with a step input.  High speed Op- Amps optimized for video line drivers have slew rates in the hundreds/USecs. It all comes down to how much of capacitance your Op Amp is driving. Higher the cap, the more current it needs to charge the cap and hence higher drive strength required of its output drivers. Our TS1001 is more than sufficient for dealing with slow moving signals or even better dealing with DC. It’s not even rated for a capacitive load.

[1] Design with Operation Amplifiers and Analog Integrated Circuits- Sergio Franco

[2] CMOS Circuit Design, Layout, and Simulation- Jacob Baker

Join the Conversation

9 Comments

  1. Interesting. So now what would be the best way to drive the TS1001 from a CR2032? Because the initial 3V is too high, but how do you get that down to 2.5, but allow you to go all the way down to 1V?

    A 2.5V LDO would do alright, but then you lose a couple hundred mV at the end, and have the additional current draw of the regulator. Is there a better way?

  2. You could power the Op-Amp from a single 1.5V button cell battery. It is shown on their application sections on Page 8. Any other approach would involve a CR2032 and a post regulator like you observed.

  3. @squonk, strange you should mention this. I already have the magnetics,the FETs on order to make one this past friday. I am using the TS1001 as a comparator instead of the LT comparator. Will post some scope captures once I get it to work!.

    This would be the Jim Williams Memorial circuit. What a genius :-). Respects.

    1. Fantastic, and the first one on the list describes a hack using an olive jar to measure temperature (an131f.pdf). This could see me reading for days…

Leave a comment

Your email address will not be published. Required fields are marked *

Notify me of followup comments via e-mail. You can also subscribe without commenting.