High-pass filter implementation with a low cutoff frequency

Hi Mike,

A single op amp active filter stage could accommodate a second-order high-pass filter. The advantage of using a single op amp could be flexibility in terms of supply voltage and selection of an op amp that has very low noise, wide bandwidth etc. A program such as TI's Filter Designer is a useful tool for synthesizing such filters.

Also, there is a universal active filter (UAF) product option that uses a state-variable topology, the UAF42. The UAF topology offers some performance advantages compared to simpler single op amp filter solutions. The UAF42 has precise integrated circuit capacitors and resistors so the difficulty of finding precision capacitors is eliminated.It usually offers lower fc and Q sensitivities compared to a single op amp solution.

I did some work with an old DOS UAF42 program and came up with a 20 kHz high-pass, having a Butterworth response and a gain of 1 V/V in the pass band. You can see the circuit configuration and the response below. Active high-pass filters tend to have a bit of an upward increase in gain before they roll off due to their limited bandwidth. However within most of the passband the gain is nearly flat at 1 V/V (0 dB).

TI offers some switched-capacitor filters, but they only provided as low-pass response. In general, switched-capacitor filter solutions have higher noise than what can be achieved with an op amp analog solution so they may not be a good choice for your application.

Regards, Thomas

PA - Linear Applicaitons Engineering

Q1:  The UAF42 datasheet specifies under "Filter Performance" that the frequency range, 'fn' is typically from 0 to 100 kHz.  Does this mean that if the high-pass filter cutoff frequency is set to 20 kHz, that the pass-band will only extend up to 100 kHz?  For my design, I need the pass-band to extend up to at least 500 kHz.  

Q2:  I have used TI's Filter Pro to design a high-pass, Sallen-Key, Bessel amplifier and filter using OP4322 (see image below).  However, when I built the design, it was extremely noisy.  (As a temporary fix, I replaced this filter/amp with a discrete filter and amp from Minicircuits, and the noise went down by about 20 dB).  I don't know if it is the high gain (+35 dB) that could have been the source of noise or the fact that I was trying to implement an amplifier AND a filter simultaneously.  Any thoughts???  

Q3:  What is the proper way of comparing UAF42's noise to OPA4322 given that UAF42 has noise density of 25 nV/sqrt(Hz) at 10 Hz, while OPA4322 has 8.5 nV/sqrt(Hv) at 1 kHz?

Q4:  Which low-noise op-amp would be a good starting point?

Thanks

Hi Mark,

Q1:  The UAF42 datasheet specifies under "Filter Performance" that the frequency range, 'fn' is typically from 0 to 100 kHz.  Does this mean that if the high-pass filter cutoff frequency is set to 20 kHz, that the pass-band will only extend up to 100 kHz?  For my design, I need the pass-band to extend up to at least 500 kHz.

The UAF42 Applicaitons Information section in the datasheet does state, "Frequency • Q product of the UAF42 has been improved, and the useful natural frequency extended by a factor of four to 100kHz." Therefore, the that fn reference is with regard to the cutoff frequency range. The the device would need to provide a reasonable replication of the expected filter response - at least until the op amps used in the UAF run out of bandwidth.

I re-simulated the UAF42 20 kHz high-pass filter using JFET op amp models that are very simular to that inside the UAF. The results seen below verify that the bandwidth of the filter is maintained until about 4 MHz at which frequency the gain roles off rapidly. The gain-bandwidth of the op amps is around 4 MHz so it is no surprose it rolls off there. There shouldn't be any problem meeting your 500 kHz upper frequency requirment.

Q2:  I have used TI's Filter Pro to design a high-pass, Sallen-Key, Bessel amplifier and filter using OP4322 (see image below).  However, when I built the design, it was extremely noisy.  (As a temporary fix, I replaced this filter/amp with a discrete filter and amp from Minicircuits, and the noise went down by about 20 dB).  I don't know if it is the high gain (+35 dB) that could have been the source of noise or the fact that I was trying to implement an amplifier AND a filter simultaneously.  Any thoughts???  

FilterPro was our baby and I still use it often because of its ease of use. A gain of 35 dB is pretty high, 56 V/V. Noise generated by the op amps and resistors, and externally induced noise will be amplified by that gain. Keep in mind that the OPA4322 has a 20 MHz bandwidth and since you have a high-pass response induced HF noise will be amplified by that gain. I suspect if you placed the board in a shielded metal enclosure, used coax connectors for the input/output ports and carefully bypassed the dc supply lines you would see the noise drop considerably.

Q3:  What is the proper way of comparing UAF42's noise to OPA4322 given that UAF42 has noise density of 25 nV/sqrt(Hz) at 10 Hz, while OPA4322 has 8.5 nV/sqrt(Hv) at 1 kHz?

To compare apples to apples see OPA4322 datasheet Figure 9, INPUT VOLTAGE NOISE SPECTRAL DENSITY vs FREQUENCY graph. The low end of the horizontal frequency scale is 10 Hz. Moving up to the curve we see the OPA4322 input voltage noise spectral density is about 55 nV/rtHz at that frequency. Higher than the UAF42, but still low for 10 Hz.  

Q4:  Which low-noise op-amp would be a good starting point?

Here is a link to the TI low noise op amps. It will allow you to search the amplifiers based on your circuit requirements. If you need further help selecting an amplifier we can provide you with help:

Regards, Thomas

PA - Linear Applications Engineering