Lowpass filter, cutoff 10kHz, fast response time, very low noise addition

Probably the lowest noise kind of low pass filter you can make is an LC (series L, shunt capacitor). But thermocouples are often prone to common mode noise (voltage or current that is common to both leads of the couple) because the couple is often grounded, and the voltage at that point may be very noisy, compared to signal ground at the receiving end. I have often had very good results suppressing this kind of noise bu adding a common mode filter inductor in series with both leads of the couple, at the receiving end. This sort of thing is made to be used as a power line filter in switch mode supplies and is very available. It adds only an ohm or two to the couple, so open couple detection methods work right through it. Here is an example of what I am talking about:

The beauty of this type of filtering is that it can have a very low frequency attenuation for common mode noise, but pass differential mode voltages (the couple voltage) with essentially no filtering.

--
John Popelish

I don't kow what type of thermocouple you are using but most thermocouples are low pass filters with a 3 db frequency less than 100 cycles per second. They won't measure rapid changes in temperature. This means with your 125 kHz sampling frequcny you could use a simple one pole low pass RC filter with a 100 cps 3 db point. Then if you have any unusual noise problems like 60 cycle power line noise or high frequency noise from an induction heater you could get rid of this noise with a digital filter. As a side note it seems unusual you are not using an A/D comverter better suited for thermocouple signals then this 125 kHz device you are using.

Hi,

The answers are all generally yes. But to start with, a lowpass filter by definition doesn't have a "very fast time response". You must accept the time response that goes with the filter you select.

It can certainly be passive and low noise - just add a series resistor R that doesn't upset the DC accuracy or ADC noise too much, then shunt the output to ground with a cap so R.C = T (the filter's time constant

- about 16us for 10kHz).

But you probably want more than 20dB/decade rolloff, and the best simple circuit that does that is a variation of the above single pole circuit. It uses a low noise bipolar transistor wired as a simple common emitter amplifier, with resistors from supply to collector, and from collector to base. Then instead of the single cap to deck that was on the single pole filter, you wire in two caps - one to the transistor's base, the other to its collector. Eg, try making the series resistor 10k, the other two resistors 33k, and the two caps

1n0. It's only 2 poles, but produces no DC error and very low noise, so maybe it's enough? To be sure you need to know what it is you're trying to reject (what frequencies, how much rejection).

Failing that, you could make an L-C filter; at 10kHz, you could get away with regular cheap axial chokes, but you'd need to allow for the very poor Q (not easy, and not the sort of thing you can tweak to get it right).

Or, if you only want one, use "proper inductors" and hang the expense.

And I have to ask - who makes thermocouples that resp>Hi,

Tony (remove the "_" to reply by email)