Better Linearity For Frequency-To-Voltage Converters

June 14, 1999
In many applications, such as frequency-locked-loop circuits or tachometers, a dc voltage proportional to an input frequency is required. Some special ICs are specifically designed for a highly...

In many applications, such as frequency-locked-loop circuits or tachometers, a dc voltage proportional to an input frequency is required. Some special ICs are specifically designed for a highly linear frequency-to-voltage conversion (e.g., the AD650 from Analog Devices). However, these devices aren’t commonly available, compared to simple CMOS 4000 series or 74HC series ICs, and their price is typically much higher. On the other hand, if an inexpensive one-shot such as a 74HC423 or a CD4528 is used for frequency-to-voltage conversion, the linearity is generally unsatisfactory.

Adding a very simple RC network improves the linearity between dc output voltage UDC and input frequency f by least one decade. The figure illustrates a standard frequency-to-voltage conversion arrangement with a CD4528 one-shot, except that it’s enhanced by three additional components —two resistors (R1 and R2) and one capacitor (C1).

R1 and C1 form a first-order lowpass filter, similar to R0 and C0, to get rid of the frequency-dependent ripple. The filter’s time constant, t = R1C, is the same as for R0 and C0, and must be large compared to the longest input period: τ >> TMAX, = 1/fMIN. The resistor R2 slightly changes the threshold level at pin 2 as a function of the dc voltage across C1 (i.e., as a function of the input frequency). This provides compensation for first-order linearity errors.

As an example, the table shows the linearity error obtained with the circuit, with and without the compensation network. The maximum linearity error is reduced by a factor of 30 from 1.8% to 0.06%. The values in the table were achieved for R1 = 30k, R2 = 680k, and C1 = 0.1 µF. The time-constant determining network was chosen as Rτ = 10k, Cτ = 220 pF. The supply voltage was VDD = +5 V.

The exact value of R2 must be determined by measurement because it depends on the frequency range and the individual IC selected for the circuit. It’s usually in the range of several hundred kilohms. Usually, R1 is not larger than a few tens of kilohms. Therefore, the value of R2 primarily determines the compensation current into the threshold pin. If, for some ICs, the dependency should be of opposite polarity, as shown in the figure, R1 merely needs to be connected to the non-inverted output Q instead of —Q.

As an example of a practical application, the network was inserted in the frequency-locked-loop circuit of an amateur radio SSR transceiver. Because each amateur radio shortwave band is subdivided in 100-kHz ranges, the display of a 10-turn helix potentiometer could be used directly for the frequency display. Without the compensation network, an error of up to 5 kHz occurred. With the compensation network added, the error was well below 1 kHz and depended mainly on the linearity error of the potentiometer.

References:

  1. Jirmann, J., “A Spectrum Analyzer for the Radio Amateur,” VHF Communications, 4/1987, p. 232-242.
  2. Oppelt, R.,“A Continuously Tunable VCO for the 2m SSB Band (part 2),” VHF Communications, 1/1989, p. 46-56; “Stabilizing the VCO-Frequency by Means of Monostables (part 1),” VHF Communications, 4/1988, p. 238-245.

Sponsored Recommendations

Comments

To join the conversation, and become an exclusive member of Electronic Design, create an account today!