This is my first stab at the input amplifier for the Archaeology Resistance Meter Project.
It's tagged as armp
The schematic has (almost) no component values but I shall talk about them in the text.
All two terminal components are in 0603 packages unless marked otherwise.
A better print of the schematic is uploaded as a .pdf in Open Source Hardware
The original plan was that the maximum possible (working) input voltage would be +/- 15V but comments
from davemartin have persuaded me that this may not be enough.
Accordingly the design now has a 20dB front end attenuator which will allow it to cope with +/- 150V signals.
For some of the possible applications for this design the attenuator offers no benefits at all - the pcb could
be built without fitting RL3, R7, R8, VR5 and VR6.
The input is differential, and is buffered by U1 and U2, the differential; signal is then converted to single ended
by U3, low pass filtered by U5 and made differential again by the ADC driver U4.
This may seem a little odd, to go from diff to single ended to diff again but there is a reason.
The signal is centred around the instrument GND net for the input differential and single ended stages but the ADC
differential signal is centred around one half of the ADC reference voltage.
No doubt a fully differential input and output level shifting ADC driver is possible but I don't know of one to buy.
Both inputs follow identical paths until they reach the outputs of U1 and U2 so I shall describe only the INP1 path.
The input is defended against RF by the input PI filter. C1, C3 and L1. The capacitors must be rated for 200V.
C1 is in an 0805 package in case we decide to use a transient suppressor type device after tests or second thoughts.
R1 and R3 protect the input from large continuous AC or DC voltages, a very unlikely form of abuse, but capable of
mass chip destruction. (When the attenuator is not switched in and the DC path is select by RL1, or the frequency
is high enough for C5 to be a low impedance, then the protection diodes D1 and D2 attempt to clamp the voltage
to 0.6V outside the +/- 15V supply rails. If the abuse voltage comes from a low impedance the supply rails will rise
unless R1 and R3 limit the current to a low enough level.
In order to keep the noise as low as possible we want R1 and R3 to be as low as we can get away with.
To survive continuous application of 250V AC R1 + R3 must be at least 31k (for 1W per resistor) - this won't matter
much when the attenuation is selected but will add 22nV/rtHz to the input noise (4x as much as the amplifier !) when
the direct path is selected.
(If you want to stand 250V AC then C1 and C3 must be rated for 500V.)
Relay RL3 switches the attenuation in or out of circuit, it is much larger than the other relays so that it can be capable
of sustaining high voltages between contacts. (Farnell 4219960 will do.)
C5 is the AC coupling capacitor, 0.1uF @ 400V (Farnell 1890167). This will have an impedance of 16k at 100Hz which
will increase the system noise compared with direct coupling.
R23 sets the input impedance when the attenuator is out of circuit and is part of the attenuator (with R7 and VR5
when it is used. I'm suggesting R23 = 1M and R7 + VR5 = 9M (it works OK for scopes !).
VR5 is necessary so that the gain of the two inputs can be exactly balanced when the attenuation is used. It isn't
important in this application that the attenuation be exactly 10x, just that the two inputs are the same to within
0.1% or better if possible.
The AC adjustment found on scope probes is not needed here because the operating frequencies are so low. If the
shunt capacitance across R8 and R23 was 5pF each things would start going wrong at about 10kHz - our max
operating frequency is 200Hz, so there is a good margin.
U1 and U2 are unity gain buffers, low noise, high impedance rail to rail op amps operating from +/- 15V supply rails
are needed to maximise common mode range. TI's OPA192 offers a good combination of features with adequate
bandwidth and very low DC offsets as well.
U3 is a differential amplifier. According to the state of RL2 it can have one of two gain settings. The actual gains are set by
the resistor ratios. To get the best possible common mode rejection the ratios in the upper and lower resistor chains
must be as close as possible. Even if 0.1% resistors are used you can get a 0.2% ratio mismatch which would degrade
the common mode rejection to only 54dB, I'd like to get to 80dB which needs adjustments.
The rather odd arrangement of R13 and VR1 (and the other pots and resistors) is so that the adjustment of one gain
will not interact at all with the other. (VR1 and VR3 work on one setting and VR2 and VR4 on the other). It may not be
necessary to have VR3 and VR4 but I like symmetry.
The values of some of the resistors are quite low so the post are shunted by resistors so that the actual resistance of the
pot is unimportant.
It's easier to understand when there are some actual values.
The input buffers and attenuator can have a gain of 1 or 0.1.
With 15V of input we would like the ADC (PCM4201) to be fully driven, which needs the diff amp to have a gain of1/3 (the low pass filter
and U4 have again of 1).
But for special purposes (like measuring low value resistors which is the alternate use for this circuit), we would like more gain
so in the high gain mode we would like U3 to have a gain of 10/3.
(We then end up with three input ranges, +/-150, 15 and 1.5V full scale.)
For low noise we need the resistors to be as low a value as possible but we don't want to have to use huge currents to drive
If we want the maximum output current for U3 to be nor more than 3mA that sets a minimum value for R19 of about 5k.
So lets make R19 = 5k1.
For a gain of 1/3 we need R11 + R15 to be 15k3 (5k1/(1/3)) (we're ignoring R13 and R17 because they are small by comparison.
For a gain of 10/3 we need R15 + R19 = R11 * 10/3
With a bit of juggling we can work out that R11= 4k7077, R15 = 10k5923 and R19 = 5k1.
We'll use R11 = 4k7, R15 = 10k5 and R19 = 5k1 (all 0.1%, 20ppm/C or better), with R13 and R17 = 51 R and VR1 and VR2 = 1k
If we use 0.1% resistors R13 and its symmetrical allies need to be about 0.5% of 10k5 to be sure that we have enough adjustment
which makes them 50R each. (actually 51R // 1k pot which is close enough to 50R)
The high gain will be (5100 + 10500 + 75)/(4700 + 25) = 3.317 and the
low gain = (5100 + 25)/(10500 + 4700 + 75) = 0.3355, so the gain difference is 19.9dB which is close enough !
The diff amp is powered by +/-15V to keep the common mode range as high as possible and it's centred on the instrument ground.
A Sallen and Key type low pass filter, using another OPA192 amp, restricts what can get into the ADC drive chip ( an LTC6362).
The ADC will sample at between 8 and 10 kHz (we may have reason to move it about within this range as will become clear later.
It can cope with unwanted signals between 4kHz and 2MHz but we can help it and the system noise along by using the
low pass filter to get rid of anything much above 500Hz.
The LTC6362 takes our GND referenced single ended signal and converts it to a differential signal centered on VOCM which
in this design is one half of VP5Q (VoltagePositive5Quiet). VP5Q is used as the analogue supply an reference for the ADC.
For simplicity I've powered U5 and U4 from VP5Q as well - this will cost us a bit of the ADC's range since it's full scale input range
is 0V to Vref (VP5Q again). Yet more power supplies (-0.5V and +5.5V) seems a step too far for the 2dB loss of dynamic range.