Design Requirements Spec

INTRO

If the Use Case provides the WHY, then the Requirements define the WHAT for the instrument to be created.

A Design Requirements Specification describes the product in terms of its
- Functions (goals, objectives)
- Features (tools, attributes to accomplish goals)
- Performance (measureable qualities)

Where are the requirements derived from? These specs can

1. fall naturally out of the Use-Case
2. reflect industry standards
3. require creating a strategy of making these decisions

Back to Design Series
 

FUNCTIONS

• Measure voltages (listed in Use Case: 1.5V to 20V).
• Display voltages.

FEATURES

• Two voltage ranges: +4V and +20V Range.
• Pushbutton selects voltage range
• Laptop window displays voltage.
• Positive input only (unipolar)
• No Calibration

PERFORMANCE

• Accuracy
  • 4V Range: Verror = ±15mV ±1.5% of Reading
  • 20V Range: Verror = ±75mV ±1.5% of Reading
• Measurement overange: 20%
  • 4V Range: 4V x 1.2 = 4.8V
  • 20V Range: 20V x 1.2 = 24V
• General
  • Input R: 1M both ranges
  • Input V max: +/--50V (must safely handle overvoltages)
  • Display update: 0.25 to 0.5s (industry standard)
  • Settle to within accuracy < 1 sec
  • Operating Temp: 25 ° C (temp controlled home.)

VOLTAGE RANGE STRATEGY

Why two voltage ranges, why not a single 20V range?

• Many ADCs have an input voltage range of 0V to 4V (or 0V to 5V).
• A 4V input range is an easy choice to directly digitize the voltage below 4V.
• However a 20V range requires an 5:1 attenuator to scale down the 20V input to a 4V input level of the ADC. Unfortunately, scaling down comes with a price. Any error of the ADC, 10mV for example, will scale up by 1:5 for a 50mV equivalent error at the battery, degrading the accuracy at lower voltages! 
• So, we'll keep a separate lower range for better accuracy for inputs below 4V.

ACCURACY STRATEGY

Where did the accuracy requirement come from? The answer came from digging into the Use Case and making some reasonable assumptions . 

• The Use Case states that customers typically replace a 1.5V AA at 1.1V and a 9V at 7.0V.
• An Industry Rule-of-Thumb:  the instrument should be 10x more accuracte than characteristic to be measured.

1.5V AA Accuracy Required

• What is the expected voltage change from nominal to depleted battery?
    ∆V = 1.5V - 1.1V = 0.4V
• For a measurement with 10x more accuracy, what is the max error allowed:
   Verr = 0.4V / 10 = 40mV

9V Accuracy Required

• What is the expected voltage change from nominal to depleted battery?
   ∆V = 9.0V - 7.0V = 2.0V
• For a measurement with 10x more accuracy, what is the max error allowed?
    Verr = 2.0V / 10 = 200mV

OFFSET & GAIN ERROR STRATEGY

Most instruments exhibit two significant non-ideal behaviors: an Offset Error (Voff) and a Gain Error (Kerr) giving a total error of

  Verr = Voff  + (Kerr x Reading)

The goal is to select these errors in order to meet the total accuracy of 40mV (for a 1.1V Reading on the 4V Range) and 200mV (for a 7.1V Reading on the 20V Range). This may take some some trial and error as well as past experience.

Example: A total error Verr = 40mV at 1.1V Reading
One approach is to initially split the total error budget of 40mV by allocating 20mV to the Offset Error and 20mV to the Gain Error (∼ 2% of the 1.1V Reading). To include some design margin, we trim both back a bit for a total error of
Verr = 15mV + 1.5% of Reading. Later, we can adjust the offset / gain values as a practical design develops.

Let's check if the initial errors chosen will pass spec.

• 4V Range: 15mV + 1.5% of Reading
  • Check Verror = 15mV + 1.5%/100 x 1.1V = 32mV (< 40mV PASS)
• 20V Range: 75mV + 1.5% of Reading
  • Check Verror = 75mV + 1.5%/100 x 7.1V = 182mV (< 200mV PASS)

BANDWIDTH STRATEGY

Ideally, the voltmeter should respond quickly to a voltage change - implying a high bandwidth.

However, the long battery test leads may pickup unwanted signals like 60Hz mains or a multitude of radio frequencies. Unfortunately, these signals can cause fluctuating readings or harrnful transients. What's the fix? Add a Low Pass filter before the ADC to reduce these interfering signals. The filter bandwidth should be

• LOW enough to attenuate 60Hz mains pickup.
• HIGH enough to settle to specified accuracy without a significant perceptable delay to the user.

HANDS-ON

Play in the Excel file - modify values, see what happens!

• Excel File:  DVM1_Offset_Gain_Errors1.xlsx
• Right Click on the filename, select "Save link as...".

Offset and Gain Error

• Suppose the system had a larger offset error. Increase Voff from 15mV to 25mV. What is the Verr at 1.5V?
• Can you reduce the Gain Error (Kerr) sufficiently to still meet the spec of Verr = 40mV at 1.5V?

ADJUST THE SPECS?

Are these specifications set in stone? During development, specs can be refined and adjusted for a number of reasons.

• A competitor has an improved performace.
• The product achieved better than expected performace.
• The accuracy cannot be met given the budget or technology.
• The assumed needs were over or understated.

 

NEXT UP

The next step is to draft the Implementation Strategy to answer HOW we can achieve the Design Requirements.

Back to Design Series