AD8363: Part-to-Part variation

Question What is the expected spread from chip to chip for the slope and offset parameters of this power detector? We see a spread of more than 5 dB from chipto chip when measuring power and applying a slope and Offset from datasheet or when we calibrate it on our own. Answer In the following I have attached a figure, which is a filtered version of one of the plots in the datasheet. This shows the part-to-part variation inoutput voltage at around 900 MHz. So what this is saying is that if you apply an input power of -10 dBm, output voltage will vary from part to partfrom about 2.35 V to about 2.6 V. In dB terms this is a variation of around 4 dB - this is approximately what your are seeing. From this plot it is also visible, that intercept varies a lot more than slope on part-to-part basis.

Operating the AD8364 dual RMS Detector at frequencies below 450 MHz

Q. The AD8364 datasheet provides specification data down to 450 MHz but says that operation down to "Low Frequencies" is possible. Do you have any advice on operating this device below 450 MHz. -------------------------------------------------------------------------------------------------------------------------- A. Normally, we recommend driving the RF inputs differentially using a balun. Providing a balanced signal to the RF input tends to maximize dynamic range, particularly at the top end. However, at lower frequencies (e.g. lower than around 450 MHz), this becomes less critical and you can consider driving the device with a single ended signal (the unused input is then ac-coupled to ground). This is described in more detail on page 36 of the Rev 0 data sheet. Of course, you can still drive the device differentially using a balun. In choosing a balun/transformer, you should look for good amplitude and phase balance (i.e. outputs are 180 degs out of phase) at the frequency of interest. When operating at relatively low frequencies, you also need to be conscious of an internal offset nulling loop. If your input frequency is low enough, this loop will mistake it for an offset voltage and will try to servo it away. Fortunately, you can get around this by adjusting the size of the CHP capacitor (pins 1 and 24). This is also described on page 36 of the rev 0 data sheet.

Reducing the Response Time of an RMS Detector

Q. As I increase the value of the AD8362 filtering capacitor, the device's fall time after an RF burst has decayed becomes very long. How can I get around this? -------------------------------------------------------------------------------------------------------- A. As the CLPF capacitor is increased, the residual ripple on the output decreases at the cost of a slower response time. However, as the value of CLPF increases, the fall time increases at a faster rate than the rise time. The shape of the falling voltage becomes more linear than exponential (see the attached oscilloscope plot) In applications where the fall time is critical, it is better to reduce residual ripple by placing an RC filter at the rms output of the AD8362 rather than further increasing CLPF. In the second oscilloscope plot below, such a scheme is implemented. A filtering capacitor of 390 pF is placed on CLPF and an RC filter (75 ohms, 0.15 uF) is placed at the rms output pin. With this configuration, the fall time reverts to a faster exponential decay while residual ripple is a little less than the previous case (150 mVpp vs 200 mVpp). The attached file shows resoponse times for other filter component values and also includes response time plots for the AD8363 and AD8364 rms detectors. The ADL5902 acts much like the other RMS detectors when a filter is put on its output. However the filter also causes some peaking in the detector response. Placing a resistor in parallel with the CLPF capacitor, has the same affect as an output filter but without the peaking. More information is included in the ADL5902 datasheet. Attachments: RMSDetectorPulseResponse.pdf

ADL5902 RMS Detector Output Voltage Clamping for High Input Power Levels

Q. At high input power levels (above 0 dBm), the output voltage of the ADL5902 rms detectors (and some other rms detectors also) loses linearity and clamps to the rail. What causes this and is there any way to prevent it? --------------------------------------------------------------------------------------------------------- A. The plot below illustrates the effect you are describing. In this case, as the input power approaches +3 dBm, the output voltage starts increasing rapidly and clamps at a level of around 4.7 V (this clamping level varies with temperature). This effect is caused by the internal architecture of the ADL5902. In its normal operating mode, Vout connects to Vset. This closes an internal AGC loop which uses a VGA to present a constant input level to the internal square law detector (one of the X2 blocks). As the input level increases, the gain control voltage to the detector (which is also the final output voltage), keeps increasing which causes the gain of the VGA to decrease. When the gain cannot be decreased any further, the AGC loop's integrator voltage (Vout) goes open loop and clamps to the rail (approx 4.7V). If you do not want the device to behave in this manner, the most practical solution is to increase the slope of the transfer function by connecting a resistor divider between Vout, Vset and ground (as described in the datasheet). This will give higher Vout voltages for a particular input power. So you could set the slope so that the output voltage is equal to 4.6 V for an input level of +3 dBm. If you increase the input level above +3dBm the loop will still open up but the output voltage jump will be much smaller. In this example, the voltage will jump from 4.6 V to approx 4.7V (clamp level varies with temperature).

Operating the AD8362 Evaluation Board at Low Frequencies

Q. I am having some problems getting the AD8362 evaluation board to work low frequencies (e.g. around 1-10 MHz). Do I need to make some changes to the evaluation board for operation in this frequency range? ----------------------------------------------------------------------------------------------------------------------------------- A. The input balun on the evaluation board (ETC1.6-4-2-3) is specified to work from 500 MHz to 2.5 GHz. So, I would suggest removing this and driving the device single-ended as described in page 19 of the Rev. D datasheet. Make sure that you use two input coupling capacitors that set a low enough input high-pass corner frequency (assume that the input resistance at the device is 200 ohms). Another alternative would be to try and find a low frequency 4-to-1 balun that will fit on the existing footprint on the eval board. One option is MABA-010012-ES4302 (Macom) which will work down to 30 MHz. If you want to operate below approximately 2 MHz, you will also need to add a capacitor to the CHPF pin. This moves the corner frequency of the offset compensation loop down in frequency. Operation at low frequencies and setting CHPF are discussed on pages 20 and 21 of the Rev D data sheet and in AN-691 ( Operation of RF Detector Products at Low Frequency ) . Attachments: AN-691_Low_Freq_Detector_Operation_Final.pdf

Alternatives to AD8362 High Range RF RMS Detector

Q. I have used the AD8362 logarithmic detector in the past. Now I am starting a new design and am wondering if there are any newer alternatives. ----------------------------------------------------------------------------------------------------------------------- A. The AD8362 is a really good part. It operates at frequencies up to 3.8 GHz with excellent rms accuracy and has up to 65 dB of detection range. However, if you are starting a new design, there are some alternatives you should consider. The ADL5902 ( www.analog.com/ADL5902 ) also has up to 65 dB of detection range but operates at frequencies up to 9 GHz. Its temperature drift and transfer function ripple are also superior to the AD8362. Unlike the AD8362, the ADL5902 has a temperature compensation pin. By adjusting the voltage on the TADJ pin (you can do this with an external DAC or you can simply use a resistor divider along with the on-board 2.3 V reference voltage), you can fine tune the temperature drift compensation for a particular frequency. Another device to consider is the AD8363. This device is very similar to ADL5902 but offers a little bit less detection range (approximately 50 dB).

Can the AD8362 Logarithmic RMS Detector accurately WiMax and LTE?

Q. I am looking for an RMS detector which will measure WiMax and LTE signals (OFDM Modulation). Looking at the data sheet of the AD8362, I see no mention of these standards. Can the AD8362 accurately measure the rms power level of such signals? ---------------------------------------------------------------------------------------------------------------------------------- A. The short answer is yes. The AD8362 logarithmic detector will have no problem dealing with these signals. When the AD8362 data sheet was written, these wireless standards did not exist. This is why they are not mentioned. Generally speaking, the accuracy of an rms detector will start to degrade as the peak-to-average ratio of the signal increases beyond a certain level. Beyond this level, there will be some internal clipping inside the circuit which will result in small errors. The AD8362 has been extensively tested with signals with peak-to-average ratios as large as 11 and 12 dB (e.g. single and multi-carrier WCDMA and CDMA2000 signals). These peak-to-average ratios are similar to those of WiMax and LTE signals so the AD8362 will not have a problem accurately measuring their signal level. You should think about using a newer device such as AD8363 or ADL5902. These devices have similar rms accuracy to AD8362 but their temperature stability and transfer function ripple are much better.

S-Paramters for a RMS-responding RF Detectors

Question : Are S-Parameters available rms-responding RF detectors such as AD8362 or ADL5902? Or is there some kind of a behavioral model available for these devices. Answer : We do not publish s-parameters for rf detectors. It is not really meaningful to talk about an RF detector having a full set of s-parameters. While you could measure the input return loss (S11) of the RF input of an rf detector, you can't really talk about the the S21 since the output is not linearly related to the input. Also since the rf detector output looks like dc if the instantaneous input power to the detector is constant, you cannot talk about there being a phase relationship between the input and the output. If you are trying to model the behavior of an rf detector with a linear-in-dB transfer function, the following equation may be helpful. Vout = Slope x (Pin - Intercept) where Vout is the predicted output voltage of the rf detector Pin is the input signal to the rf detector expressed in dBm Slope is the incremental relationship between input power and otuput voltaqge (i.e. deltaVout/deltaPin), expressed in mV/dB or V/dB Intercept is an extrapolation of the rf detector's transfer function. The Intercept is the extrapolated input power level that would result in a detector output voltage of 0 V. This equation is only an approximation. It assumes that the transfer function of the device is a perfect (linear-in-dB) straight line and that the device has an infinite input power range. In reality the RF detector will have a limited range over which the transfer function follows the equation. And the real world transfer function will have some ripple and non-linearity. However, the equation does serve as a useful model for device operation within its linear power range.

AD8363: Using INHI vs. INLO for Wideband Applications

Q. The AD8363 datasheet recommend driving the RF input through INHI for frequencies at or below 2.7 GHz and through INLO for operation above 2.7 GHz. If I have an application where I want one circuit to span a broad frequency range that includes 2.7 GHz and 3.5 GHz, which input should I choose. A. You can use either depending on whether your frequency range is concentrated above or below 2.7 GHz (that is the frequency at which we recommend switching from INLO to INHI. As go go up in frequency driving INHI, you find that the transfer function loses some linearity above input power levels of around -10 dBm. Switching over to driving the signal through INLO improves this a bit. However, if you do drive through INHI at frequencies above 2.7, you will still get performance that is fairly linear. And if you calibrate the transfer function at three points, as suggested in the Rev. A datasheet, that will improve things even further. The same argument applies to using INLO; it will still give good performance below 2.7 GHz, particularly if you 3-point calibrate the circuit.