Finding a Low Voltage JFET

[aep9690]

I am building a sensor and I need some help finding a part.

The sensor measures the phase difference between two signals.  I then intended to convert that phase difference to an analog voltage by low pass filtering it.  HOWEVER, because the phase difference is measured with an XOR gate (due to power requirements) the upper and lower voltage limits are not well defined.  The solution I have come up with is to use either a MOSFET or JFET to act as a pass transistor between the low pass filter and a 2.5V reference.  This is the hard part.

Because I need to maintain the accuracy of the sensor at greater than 1%, the voltage error going into the ADC must be less than 1mV.  This means the noise and voltage drop across the FET must be as low as possible.  This seemed like a good time to use a JFET.

My requirements are that the JFET is being driven with a 42KHz signal that varies between 0 and 3.3V.  The reference voltage is 2.5V, I will probably end up buffering this with an op-amp.  I found this part: http://www.newark.com/vishay-siliconix/sst201-t1-e3/n-channel-jfet-40v-to-236/dp/06J8328 and this: http://www.farnell.com/datasheets/910015.pdf but I don't think these will due the job due to the variance in the gate source turn off voltage.

Any help is appreciated, ranging from finding a part for me to pointing me to a website where they deal with this kind of thing.  Thanks.

[MK]

All fets have at least a 2:1 figure for IDSS and other parameters, you have to consider that in any ciruit that uses them, they have a minimum temperature coefficient at around somewhere near 80% of IDSS, but you may need to select parts, it depends on the quantities you want to make.

[dannyf]

Most jfets are of low voltage - 20 - 30v max typically.

[rs20]

You'll probably get a lot more help with a diagram -- maybe I'm just simple, but I don't have a clue what "a pass transistor between the low pass filter and a 2.5V reference" means.

[David Hess]

I do not quite understand what you are trying to do from just your description.  A schematic would help.

I found this part: http://www.newark.com/vishay-siliconix/sst201-t1-e3/n-channel-jfet-40v-to-236/dp/06J8328 and this: http://www.farnell.com/datasheets/910015.pdf but I don't think these will due the job due to the variance in the gate source turn off voltage.

Variance in threshold voltage is going to be a problem with any FET and especially a problem in low voltage circuits.  The usual solution is to grade the parts yourself for a low or known threshold voltage.

[aep9690]

This is the basic principle:

  Except in my case the 12V supply is replaced with a buffered 2.5V reference, and I am using a JFET instead of a BJT.  I have found a bunch but after some more consideration I don't think the JFET would cut it.  When the source voltage becomes low I wouldn't be able to turn the JFET off.  However, to help other people looking for low voltage JFETs here are some that I found.

1. NXP BF862: http://www.nxp.com/documents/data_sheet/BF862.pdf
2. Vishay SST201: http://www.farnell.com/datasheets/910015.pdf
3. Toshiba 2SK209: http://www.toshiba.com/taec/components2/Datasheet_Sync/200711/DST_2SK209-TDE_EN_6915.pdf
4. On Semiconductor 2SK3557: http://www.onsemi.com/pub_link/Collateral/2SK3557-D.PDF

At this point I am pivoting from JFETs to MOSFETs, BJTs or out because the collector-emmitter voltage drop is going to be too large.  The tricky part is that I need to make sure the leakage current through the MOSFET is sufficiently low, in the nA range.  Not sure if this is possible but I am going to try.

[David Hess]

I assume you mean that circuit but with a common emitter PNP transistor so the Vbe is not a problem.

Since you were considering a bipolar transistor and the collector-emitter saturation voltage was the only problem, you could use the same transistor but with the collector and emitter swapped.  Vce will only be like 5 volts then and the beta will be low, possibly below 1, but the saturation voltage will be a lot lower also.

[aep9690]

You misunderstand, I did not want to use a BJT, I wanted to use a JFET.

[David Hess]

You misunderstand, I did not want to use a BJT, I wanted to use a JFET.

Then it is not clear from your schematics and descriptions what you are trying to do which makes it difficult to make good suggestions.

[aep9690]

I explained that the schematic is a representation, and that there were differences.  Due ITAR restrictions I can't post a schematic.

[David Hess]

The 2.5 volt reference has a low impedance either by itself or because it is buffered so base/gate current/leakage is irrelevant.  Further, the available bias voltage is low but you have not said what the load impedance is.  That is a perfect application for a bipolar transistor with reversed collector and emitter which can switch at low voltages and provide a low saturation voltage.

BJTs or out because the collector-emmitter voltage drop is going to be too large.

I have just described how to use a BJT with significantly reduced collector-emitter saturation voltage which is very useful at low supply voltages.

An alternative if you do not want to grade FETs for low threshold voltage is a low voltage monolithic analog switch like an ADG701, ADG801, MAX4624, or MAX4625.  I might also consider a photo FET like an H11F1 but I do not know if any have a low enough on resistance.  You could make a photo FET switch but a discrete implementation would tend to be slow.

[ejeffrey]

Your premise confuses me:

The sensor measures the phase difference between two signals.  I then intended to convert that phase difference to an analog voltage by low pass filtering it.  HOWEVER, because the phase difference is measured with an XOR gate (due to power requirements) the upper and lower voltage limits are not well defined.

CMOS logic should have pretty well defined voltage rails.  A CMOS output is just a MOSFET driven to hard conduction to either one rail or the other, and it is pretty much as good as the rail voltage.  That means you just need to power the phase detector from your analog reference voltage or use a dedicated LDO as both the supply and ADC reference voltage.  Since an ADC gives you a ratio metric measurement, the exact supply voltage won't matter.

[aep9690]

CMOS logic should have pretty well defined voltage rails.

CMOS voltage rails can vary by about 100mV over temperature, this is significant error.

The 2.5 volt reference has a low impedance either by itself or because it is buffered so base/gate current/leakage is irrelevant.  Further, the available bias voltage is low but you have not said what the load impedance is.  That is a perfect application for a bipolar transistor with reversed collector and emitter which can switch at low voltages and provide a low saturation voltage.

I'm not concerned with base/emitter leakage, I am concerned with collector/emitter leakage current.  If the input to my base is zero I expect the input to my ADC to be zero.  The load impedance is upwards of 20k so a leakage current of anything over about 10nA would be troublesome.

The saturation voltage would also be a problem but not because it would be large.  The issue is that it is not well defined, it can vary by as much as an entire volt over temperature and from unit to unit.  Sensor accuracy isn't so much dependent on small error, rather some known error that is stable over time and temperature.

[David Hess]

CMOS logic should have pretty well defined voltage rails.  A CMOS output is just a MOSFET driven to hard conduction to either one rail or the other, and it is pretty much as good as the rail voltage.  That means you just need to power the phase detector from your analog reference voltage or use a dedicated LDO as both the supply and ADC reference voltage.  Since an ADC gives you a ratio metric measurement, the exact supply voltage won't matter.

That makes another way to do it.  Power a CMOS tri-state driver with the 2.5 volt reference and toggle the output using the enable input.  I doubt it would be better than an analog switch but it would work; digital logic will not be specified and tested for low leakage like an analog switch.

[David Hess]

I'm not concerned with base/emitter leakage, I am concerned with collector/emitter leakage current.  If the input to my base is zero I expect the input to my ADC to be zero.  The load impedance is upwards of 20k so a leakage current of anything over about 10nA would be troublesome.

The saturation voltage would also be a problem but not because it would be large.  The issue is that it is not well defined, it can vary by as much as an entire volt over temperature and from unit to unit.  Sensor accuracy isn't so much dependent on small error, rather some known error that is stable over time and temperature.

A bipolar transistor operating in reverse will have a saturation voltage of millivolts or less.  In the past, they were commonly used to reset integrators to that level.

Reverse leakage, Icbo and Iebo in this case, is not typically specified below 50 nanoamps because it costs too much to test below that level.  Typically it is below 10 picoamps.  I have graded transistors down to the 100s of femtoamps for use as low leakage diodes and the only reason I did not test lower was because of the added difficulty.

The analog switches I mentioned have on resistances measured in ohms and leakages in the 10 picoamp range.  They would be ideal.