LOG114 Input Resistance

Hi Paul,

I have some questions and comments:

• Is the internal circuitry that generates the -0.1 V bias derived from the LOG114, or some other part of the circuit?
• The log amplifier input does not have any added series resistance at the input.
• Beyond 1 mA of input current the log conformity of the LOG114 becomes less linear. This is evident from the bottom 4 graphs on page 7, in the data-sheet. I wonder if that might be a factor producing the equivalent of the series input resistance effect you are seeing?
• Are you able to provide your schematic?

Regards, Thomas

PA - Linear Applications Engineering

Thomas,

Thanks. Here is my schematic:7824.piezo_logamp.pdf. The bias voltage is generated using A5 and applied to the Vcm_in terminal. When A2 is in its operating regime there should be no voltage between Vcm_in  and i2 (virtual ground) yet I get ~25mV when a 100 ohm resistor is connected between the input and ground on the 100mv setting (i.e. there is 75mV across the resistor. The Output voltage is still within spec. The effect is linear. When I choose the 10mv setting I get 7.5mv out. Now the current is 10x less so it is not a saturation effect, but a linear input resistance (Req = 25mv/750uA = 33 ohm).

Paul

Hi Paul,

I am thinking about the result you are observing. Disecting your latest response:

 The bias voltage is generated using A5 and applied to the Vcm_in terminal. - Okay, you are depending on the virtual ground I₂ input to track the adjustable V_CM level and develop the current through R_DUT . That seems like a good method.

 When A2 is in its operating regime there should be no voltage between Vcm_in  and i2 (virtual ground) - Other than the offset voltage that is what I would expect too.

yet I get ~25mV when a 100 ohm resistor is connected between the input and ground on the 100mv setting (i.e. there is 75mV across the resistor.The Output voltage is still within spec  - As you described earlier it is if there is a resistor in series leading from R_DUT to the I₂ input. Since this is a linear effect are you seeing that the output voltage is within the correct range at all current settings?

The effect is linear. When I choose the 10mv setting I get 7.5mv out. Now the current is 10x less so it is not a saturation effect, but a linear input resistance (Req = 25mv/750uA = 33 ohm).  - So in the case of the 10 mV setting you would 7.5 mV across the 100 Ohm resistor and 2.5 mV across the log amp inputs. The I₂ input current would be 75 uA and the proposed series input resistor would remain at 33 Ohms. Is that correct?

How are you making the actual measurement? Are you connecting a DVM between V_CM and the I₂ input and making a direct measurement, or is the voltage being extrapolated based on the level observed at Vo4? The LOG114 is a wide-band log amplifier fabricated on a high-speed process and one has to be careful when making measurements not to cause high-frequency oscillation. DC errors can occur when oscillation is present.

Regards, Thomas

PA - Linear Applications Engineering

Thomas,

Thanks for taking the time on this. I measured the voltages directly on the inputs with a DVM. I did not check on the scope while I was doing this measurement but I did check previously. The amplifier did pick up some 60Hz at high input resistors (>10Mohm) but not at the low values we are talking about. The amplifier was remarkably stable. I don't think AC oscillations are the problem, although I will check more thoroughly. The effects I am seeing are linear and predictable, fully consistent with a simple series resistor. This resistor would not affect the Vout vs Iin transfer function of the amp so maybe for this reason it is not seen on normal testing. I would appreciate it if you are able to find any info on this.

Thanks,

Paul

Hi Paul,

I have been going over the internal schematic for the LOG114 die looking for something that might explain the series input resistor phenomena. Basically, the amplifier input pins go directly to a JFET differential pair and the input ESD protection consists of normally-off current steering diodes to the supply rails. Therefore, I am not finding anything that simply acts as a series resistance. It appears to be something much less obvious then that.

The LOG114 is most often driven by a current source such as a photodiode. The impedance asspciated with that kind of current source is very high. Your application is quite different in that it involves a series voltage source with a low resistance; nevertheless a source of current, but with much lower impedance than something like a photodiode. This operating condition is more similar to what is described in the data-sheet Voltage Inputs section, on page 18 - except in this case the resistances are very low.

I have been discussing your inquiry with on of my colleagues and I will let you know if we hit upon anything that provides an explanation. Please note that the support level here in PA-Linear Applications will be much lower during the holidays and response times will lengthen.

Happy Holidays!

Thomas

PA - Linear Applications Engineering

Thomas,

Is the JFET pair part of the input amp of is this part of the ESD protection network?

It should be easy to measure the input resistance, just input a current, e.g. 1mA, and measure the input voltage. You might need to isolate the DVM at the input with a resistor (e.g. 100k) to block noise from the DVM. It would be interesting to see if you get the same value for the input resistance that I do.

Paul

Thomas Kuehl said:

Hi Paul,

I am thinking about the result you are observing. Disecting your latest response:

 The bias voltage is generated using A5 and applied to the Vcm_in terminal. - Okay, you are depending on the virtual ground I₂ input to track the adjustable V_CM level and develop the current through R_DUT . That seems like a good method.

 When A2 is in its operating regime there should be no voltage between Vcm_in  and i2 (virtual ground) - Other than the offset voltage that is what I would expect too.

yet I get ~25mV when a 100 ohm resistor is connected between the input and ground on the 100mv setting (i.e. there is 75mV across the resistor.The Output voltage is still within spec  - As you described earlier it is if there is a resistor in series leading from R_DUT to the I₂ input. Since this is a linear effect are you seeing that the output voltage is within the correct range at all current settings?

The effect is linear. When I choose the 10mv setting I get 7.5mv out. Now the current is 10x less so it is not a saturation effect, but a linear input resistance (Req = 25mv/750uA = 33 ohm).  - So in the case of the 10 mV setting you would 7.5 mV across the 100 Ohm resistor and 2.5 mV across the log amp inputs. The I₂ input current would be 75 uA and the proposed series input resistor would remain at 33 Ohms. Is that correct?

How are you making the actual measurement? Are you connecting a DVM between V_CM and the I₂ input and making a direct measurement, or is the voltage being extrapolated based on the level observed at Vo4? The LOG114 is a wide-band log amplifier fabricated on a high-speed process and one has to be careful when making measurements not to cause high-frequency oscillation. DC errors can occur when oscillation is present.

Regards, Thomas

PA - Linear Applications Engineering

Thomas,

Happy New Year!

I did a calibration on my logamp by setting Vcm=0 and applying input currents from a Keithley current source. At the same time I measured the input voltage (using a 5K resistor for isolation) and monitored the output on the scape. As you see the logamp is very well behaved and there appears to be a resistor or 34.5 ohms in series with the input. It is curious, perhaps just a coincidence, that the departure from the log-linear relationship at high currents can also be attributed to a resistor of about 30 ohms in series with the diode.

Paul

Hello Paul,

I am going to take the LOG114 into the lab and see what we find. I'll report back when I have something for you.

Thomas

PA - Linear Applications Engineering

Thomas,

Great! Waiting for the results. An interesting aside: the effects of the log-diode series resistance and the input series resistance work in opposite directions so under some conditions (Vbias ~0.1v) they compensate each other.

Paul

Hi Paul,

A couple of quick measurements using a LOG114 EVM and input currents of 1 mA (100 mV V_CM and 100 Ohm input resistor) and 100 uA (100 mV V_CM and 1 kOhm resistor) yielded an equivalent phantom input resistor of about 24 Ohms. This is different than the 34 Ohms you measured, but does support your finding that there is a series resistance.

I am going to attempt to get the latest schematic and layout for the LOG114, but that will take some doing. The original schematic I reviewed was an early version and something could have been changed at a later time. Also, the die layout might reveal some unexpected resistance in the input circuit path. Metal resistance and via resistances between layers might account for tens-of-ohms of resistance.

Regards, Thomas

PA - Linear Applications Engineering