1. An E-band 40dB dynamic range multi-tanh power detector in 0
An E-band 40dB dynamic range multi-tanh power detector in 0.13μm SiGe technology
R.Levinger1,2, B.Sheinman1, O.Katz1, R.Carmon1 R.Ben-Yishay1, N. Mazor1, S.Pivnik1, D.Elad1 and E.Socher2
1IBM Haifa labs, Haifa, Israel
2Tel-Aviv University, Tel-Aviv, Israel
Hello, my name is Run Levinger from IBM Haifa research labs, and today I will present our work on an Eband 40dB dynamic range multi-tanh power detector in 0.13um SiGe technology.
EuMW – Rome, October 5-10, 2014

2. Outline
Introduction
Conventional Bipolar Translinear Power Detector analysis
Multi Tangents Hyperbolic Power Detector analysis
Technology
Measurements
Comparison to state of the art mmW power detectors
Summary
I will start with a short introduction, then briefly review the conventional bipolar PD analysis. I will then present our MtanH PD analysis of both large and small signal. I will review the technology that was used in this work, and then present some of our measurements including a comparative measurement to the classical PD. I will conclude this talk with a comparison to state of the art mmW power detectors and a short summary.

3. E-band Transceiver Motivation & Applications
Introduction
E-band Transceiver Motivation & Applications
71-76/81-86GHz range for fast deployment, low cost “Last Mile” Point-to-Point solution in high-bandwidth 4G mobile networks.
Light licensing.
Low oxygen/rain attenuation compared to 60GHz unlicensed band.
Small chip size, high integration to enable future smart phased array systems
E-band frequency range starting from 71-76GHz for the lower band and 81-86GHzfor the upper band, offers high capacity channels, used for point-to-point wireless applications, mainly in the newly deployed 4G backhul. The E-band frequency range shows lower oxygen absorption then the 60GHz unlicensed band, and therefore can be used for larger transmission distances. Finally, smaller and less expensive cheaps with high integration levels will enable a wider deployment of E-band transmission systems.

4. Introduction E-band Super-Heterodyne Sliding IF Transmitter
Integrated Power Detector requirements:
Wide dynamic range
High power capability
Composite
Quadrature
Modulator
and IF
Up-mixer Q I X4 ÷2
PFD CP
LPF
DIV
IF VGA
RF OUT
POWER
DET
DRIVER
REF CLOCK
55.5 MHz
fRF
fIF
f4xLO
fLO PA
ATT.
Ext. LO
Power detectors are widely used in wireless communication systems, which often use automatic gain control circuits to improve efficiency, linearity or to ensure the transmit power is within specification. This E-band super heterodyne sliding IF transmitter is an example of such a system. We would like to have a PD with a wide dynamic range and high power capabilities to comply with todays SiGe designs.
Upper Band
Lower Band
fRF(GHz)
18 – 19.1
15.7 – 16.9
fLO(GHz)
72 – 76.4
62.8 – 67.6
f4xLO(GHz)
fIF(GHz)
O. Katz et al. “High-power high-linearity SiGe based E-BAND Transceiver chipset for broadband communication”, RFIC 2012.

5. Conventional bipolar power detector analysis
A pair of common-emitter HBT transistors (Q1, Q2) convert the differential voltage swing into a current that is proportional to the incident power.
The transistors are biased in a class-B regime where the device will be significantly nonlinear.
Current is then multiplied using a PMOS current mirror (M1, M2) and filtered to receive DC current.
Operating current is subtracted (Q3) and the remaining current is converted to voltage using a resistor
R. Ben-Yishay et al. “High Power SiGe E-Band Transmitter for Broadband Communication”, EuMIC 2013.
The principle behind this type of power detector is rather simple. We use a transconductance stage, in this case an HBT differential pair biased in B class regime to convert input power to current. When applying small signal approximation and using some basic trigonometry we can notice that the current has a DC component and a second harmonic component both proportional to the input power. We continue and multiply that current using a current mirror, filter the second harmonic using a LPF then subtract the quicent current and convert the remaining current to voltage using a resistor to receive the expression here. So as long as we are not in saturation mode we have an output DC voltage that is proportional to the input power, the problem is that in the conventional PD we get saturated with a low input voltage, about 1.5 kT/Q.

6. MtanH power detector analysis
This circuit is very similar to the conventional PD, a was used 3 transistor hybrid MtanH core instead of differential pair and a current source.
Where
The MtanH PD is a bit more complex. The core consists of a current source, two transistors like the conventional PD, Q1 and Q2, and an additional transistor, Q4, which is A times larger then those on the side, Q4 discards it current to the positive supply. Applying an input signal will shift DC current from the larger transistor in the middle to the transistors on the side, that is we convert input power to current, which is what we want to have from the core as was shown in the last slide.

7. MtanH power detector analysis
There is no loading of the current mirror and filter differently then the conventional PD core.
MtanH PD exits the linear domain after 5kT/q (A=4) instead of about 2kT/q in the classical PD.
The expression that governs what I have described is written here and plotted here and is it noticeable that as we enlarge A we are able to stay longer in the linear region and delay saturation. When using this cell we can have the same current mirror and filter as the conventional PD since we do not present a different load.

8. MtanH power detector analysis
It is noticeable see that the small signal approximation is more accurate for larger A values since the MtanH core becomes more linear as we increase A.
The degeneration effect can also be seen by the slope, as A grows the slope becomes more moderate. We observe it in measurement when comparing the Mtanh PD to the regular PD.
In order to observe that here too we get a linear response with respect to power we need to apply small signal approximation. We see that here as well the output voltage is proportional to input power and in fact it is the same as the conventional PD but divided by this factor right here. So the MtanH PD behaves as a degenerated PD with respect to the input power. We can see in the plot here small signal Vs Large signal and see that as A grows the small signal approximations holds for a larger input power, that is we are delaying saturation. Naturally we pay with gain, that is we are able to enlarge the dynamic range but we loose the ability to detect very low power, so we encounter the very familiar linearity - gain trade off.

9. Technology IBM SiGe BiCMOS8HP
Technology features
5 layers of metal
MIM and FET capacitors
fT= 200GHz, fMAX= 280GHz
Comparable to 65nm+ CMOS
Mature mmW technology
Fully characterized actives/passives
CMOS - 130nm
Ic(A)
fT vs. Ic
The designs were fabricated with IBMs SiGe BiCMOS8hp. The HBT’s in this technology have an ft of 200GHz and fmax of 280GHz. The technology also includes 0.13um MOS transistors. The technology is fully modeled up to 110GHz, including passives, S-param measurements that we have conducted show a very good fit up to 170GHz.
E-BAND
~ fT/2.4, ~fMAX/3.5
To preserve HBT peak-fT/ peak-gm/min-NF operation point over temperature and process, bias compensating circuits are used

10. Measurements Chip Micrograph
0.56mm PD
0.01mm^2
DC and output
Digital control
An on chip balun was used for measurement purposes. You can see the die photograph here, chip area is mm^2, the PD core is much smaller however, about mm^2.
BALUN
& Matching
0.9mm RF

11. Measurements setup TX X3
DC and Output X3
2dB loss TX
RF probe
16.7dBm
Maximum
Input
In order to measure output voltage vs input power or PD response, we took a frequency tripler followed by a 34dB amplifier with 19dBm saturation power, we used a low loss Wave Guide probe so we will have the maximum input power entering the DUT. The amplifier had a limited bandwidth, 78 – 86 GHz, below that we measured from the multiplier directly and were able to produce about 0dBm input power to the DUT.
Return loss and balun was measured using an Agilent PNA.
34 dB gain
19dBm Psat
Digital Control
Input power from input stage was recorded as a function of signal source output power prior to the measurement.
The return loss and balun were measured using a PNA.

12. Measurements Layered balun characteristics
Introduces 1.7dB IL
Less then 0.6 amplitude imbalance
Less then 1o phase imbalance
Good simulation to measurements correlation
The balun introduces 1.7dB insertion loss with less than 0.6dB amplitude imbalance and better than 1 degree phase imbalance. Simulation and measurements show good correlation.

13. Measurements – Return loss Different bias
Good matching 58 – 92 GHz for all bias points.
This slide shows return loss versus frequency for different biasing. We see that for all bias points we get very good matching from 58 – 92GHz.

14. Measurements – Return loss Measurements Vs Simulation
Good simulation to hardware correlation.
This slide shows Measurements Vs Simulation of the return loss for nominal bias and we see a very good fit indeed.

15. Measurements – PD response Over frequency
* Maximum input power is 0dBm
This slide shows the PD response over some of the frequencies. The PD response is very flat over frequency! This is offcourse a great advantage that can make power control much simpler.
Frequency response is very flat

16. Measurements – PD response Measurement Vs Simulation
This slide shows the PD response at 81GHz, and compares measurement Vs simulation. Once again we see that we get a nice agreement between the measurements and simulation.
Measurements corresponds nicely to simulation

17. Measurements – PD response Sensitivity to temperature
Low sensitivity to temperature variations is important in order to avoid PD calibration over temperature.
The MtanH PD shows very low sensitivity to temperature variation, over 20dB of dynamic range.
PTAT circuitry can be used to compensate the remaining variation.
This slides shows PD response sensitivity to temperature. Low sensitivity to temperature variations is important in order to avoid PD calibration over temperature. The MtanH PD shows very low sensitivity to temperature variation, over 20dB of dynamic range. PTAT circuitry can be used to compensate the remaining variation.

18. Measurements – PD response MtanH PD Vs conventional PD
As was shown in the analysis the MtanH behaves as a degenerated conventional PD.
The MtanH shows flatter frequency response.
This slide shows a comparison between the response of the MtanH PD and that of the conventional one. As was shown in the analysis the MtanH behaves as a degenerated conventional PD where we trade of gain to linearity or dynamic range. The MtanH has a flatter frequency response as can be seen here.

19. Performance Comparison
Technology
Frequency [GHz]
Sensitivity [mV/dB]
Dynamic Range [dB]
This Work
SiGe 130nm
65 – 86
(93 in simulation)
27.5
>40 (44 in simulation)
[1] 65 20
[2]
58 – 65
97.14
17**
[4]
SiGe 180nm
(60 in simulation)
5.65
>40*
[8]
122 17
[9]
SiGe 250nm 40 25
[10]
SiGe 350nm
120
Comparison to state of the art mmW PDs shows that our MtanH PD has the largest dynamic range over the widest measured frequency range.
(*) Includes 4 cascaded stages with attenuators.
(**) Linear dynamic range of 8dB.

20. Summary
An E band highly linear Power Detector was presented and implemented in a low-cost 0.13μm SiGe BiCMOS technology.
By applying mtanh three-transistor hybrid technique, this power detector can achieve over 40dB dynamic range, with a sensitivity of 27.5 mV/dB.
The detector is frequency insensitive, it exhibits less then 1.5dB offset within 70 – 86 GHz frequency range.
This power detector shows outstanding performance when compared to published state of the art mmW silicon based power detectors reported to date.
To conclude, this work presents an Eband PD in IBM’s BiCMOS8hp technology. The PD utilizes a 3 transistor hybrid MtanH cell to achieve a very large dynamic range, over 40dB, with a sensitivity of 27.5 mV/dB and high power abilities that suits the needs of an Eband Tx used for point to point communication.

21. Backup

26. Measurements SBS Vs Layered @ 60GHz
60GHz example
Side by Side
Layered