Test Challenges in MMICs, RFICs and High Speed Analog Circuits Dror Regev
Dror Regev
Agenda
• Presto-Engineering
• Challenges in Wide-Band and mm-Wave Test
• MMIC/RFIC Probing at high mm-Wave Frequencies
• Standard Automatic Tester limitations
• mm-Wave DUT docking & interfacing challenges
• Solution Approaches & RFIC Test Economics
• Frequency conversion technical consideration example
Dror Regev
Challenges in Wide-Band and mm-Wave Test
Ever growing demand for Data Rate, drive extended BW and Freq.
Wide Band multi-carrier modulation schemes requiring high EVM.
Reduced Dynamic Range, impairments worsen.
5 GHz 60 GHz E-Band
802.11ac Wi-Fi
802.11ad Wi-Gig
PTP 4G/LTE 1.8 GHz 1
60
M
Hz 500
MHz
𝑻𝒉𝒆𝒓𝒎𝒂𝒍 𝑭𝒍𝒐𝒐𝒓 ↑, 𝑵𝑭 ↑, 𝑻𝑿 𝑷𝒐𝒘𝒆𝒓 ↓, Linearity ↓, Phase Noise ↑, Ripple & Reflections ↑
Dror Regev
MMIC/RFIC Probing @ mm-Wave Frequencies
• GSG probe is dispersive and radiates energy
• For good S-par calibration, 30dB isolation
between probes is required. At 150um pitch, calibration
substrate BW is not guaranteed beyond 65-70GHz but perform
reasonably up to E-band frequencies.
• λ/20 GSG Probe pitch, should be maximum pitch for accurate testing. Hence 150um pitch for εr=1 can perform to ~ 100GHz. However, DUT εr will affect test accuracy as :
• 43um GSG probe is needed for best performance at ~ 110GHz.
Dror Regev
MMIC PA Probing @ E-band (71-86GHz)
• CW Gain and Saturation test @ Wafer Level
• 50um Thick GaAs wafer on insulator carrier pose Thermal Stress -> Pulse?
• Recent market drive towards full S-par test.
Dror Regev
RFIC Transceiver Probing @ E-band (71-86GHz)
• 500MHz BW wafer level RX & TX EVM tests • Employing WB Frequency Converters to/from E-band
Dror Regev
Standard Automatic Tester limitations
• Max I/O frequency limited to 6
or 12GHz!. Needs “Frequency
Extension” to drive/test higher
frequencies.
• Instantaneous BW limited
~ 200MHz.
Dror Regev
mm-Wave DUT docking & interfacing challenges
Sockets are limited in frequency (<30GHz) and pose high parasitic and no perfectly consistent mm-Wave contact resulting test errors.
Many 60 GHz DUTs are integrated with antennas.
Path Loss over the air (68dB @ 1meter/60GHz)
Crosstalk and Multipath especially for multisite
VSWR of socket and TL degrade Test Flatness. De-embedding?. 𝟏𝒏𝑯 𝑺𝒆𝒓𝒊𝒂𝒍 𝑷𝒂𝒓𝒂𝒔𝒊𝒕𝒄
@𝟑𝟎𝑮𝑯𝒛 = j𝛚𝑳 ≈ j200 ohm
𝟓𝟎𝒑𝑯 𝑺𝒆𝒓𝒊𝒂𝒍 𝑴𝒊𝒔𝒔𝒎𝒂𝒕𝒄𝒉 @𝟑𝟎𝑮𝑯𝒛
≈ j10 ohm
SiBEAM
WHD module
(Ali M. Niknejad)
Antennas
Dror Regev
More Challenges in WB and mm-Wave Test
Link Budget Example 1GHz BW @ 60GHz:
1 dB out comp. 60GHz converter (no PA): -20dBm
TX Back Off: 8dB
Maximum TX power: -28dBm
TX + DUT Antenna Gains: 20dB
1 Meter path loss: 68dB
Total over the air (OTA) loss: 48dB
Conclusion: Such stand alone converter can not support WB EVM testing OTA.
𝑅𝑋 𝐷𝑈𝑇 𝑁𝑜𝑖𝑠𝑒 𝐹𝑙𝑜𝑜𝑟: 𝑷𝑵𝒐𝒊𝒔𝒆 @ 𝟏𝑮𝑯𝒛 𝑩𝑾 = -174+10 𝒍𝒐𝒈𝟏𝟎𝟗+NF(𝟔𝒅𝑩 )= -78dBm
𝑅𝑋 𝐷𝑈𝑇 𝑆𝑖𝑔𝑛𝑎𝑙 𝐿𝑒𝑣𝑒𝑙: 𝑷𝑹𝒆𝒄𝒊𝒆𝒗𝒆𝒅 𝑺𝒊𝒈𝒏𝒂𝒍 = -28 dBm - 48= -76dBm
Dror Regev
Solution Approaches & RFIC Test Economics
“Frequency extension” NB or WB:
• Convert (or multiply) I/O Frequency of automatic tester, signal generator and test instruments to/from mm-Wave frequency.
Testing Modules integrated with Antennas :
• Employ wafer probing when relevant
Economic OTA testing or “Contact less”
probing.
UWB on load board E-Band converter 60GHz IQ converter 60GHz Transceiver
Dror Regev
Frequency conversion technical consideration IQ Conversion 101:
IF @ fLO
fLO
fLO
BBI
00/900
BBQ
IF1
IF2
BBI
00/900
BBQ
𝑰 + 𝑸
𝟐
𝑸 − 𝑰
𝟐
IF @ fLO
fLO
fLO
Ideal Up-Conversion
Ideal Down-Conversion
𝑩𝑩𝑰 = 𝑺𝒊𝒏(𝝎𝑳𝑶𝒕) ∙
𝑰+𝑸
𝟐𝑪𝒐𝒔 (𝝎𝑳𝑶 − 𝝎𝑩𝑩)𝒕
−𝑰−𝑸
𝟐𝑪𝒐𝒔 (𝝎𝑳𝑶 + 𝝎𝑩𝑩)𝒕
=
= 𝑰
𝟐𝑺𝒊𝒏 (𝝎𝑩𝑩)𝒕 +……
𝑩𝑩𝑸 = 𝑪𝒐𝒔(𝝎𝑳𝑶𝒕) ∙
𝑰+𝑸
𝟐𝑪𝒐𝒔 (𝝎𝑳𝑶 − 𝝎𝑩𝑩)𝒕 −
𝑰−𝑸
𝟐𝑪𝒐𝒔 (𝝎𝑳𝑶 + 𝝎𝑩𝑩)𝒕
=
= 𝑸
𝟐𝑪𝒐𝒔 (𝝎𝑩𝑩)𝒕 +…….
Dror Regev
Frequency conversion technical consideration Transceiver Frequency Plan using a Single Synthesizer:
Popular case : M=2, N=1.
32
21
RFLOLO
FFF
LO1, LO2
Generation:
= LO1
= LO2
GHzFF
F RFLOLO 20
32
21
GHzFF
F RFLOLO 20
32
21
Dror Regev
Frequency conversion technical consideration Ideal performance of 1/3, 2/3 frequency plan:
fRF @ 3fLO1
WB Signal IF @ fRF/3 RF=1LO2+1IF
fLO2 =
2*fLO1
fLO1 =
fRF/3
BBI
00/900
BBQ
IF1
IF2
fIF=fRF/3
𝑻𝑿𝟏𝟏 =desired output = 𝑪𝒐𝒔(𝟐𝝎𝑳𝑶𝒕) ∙𝑰+𝑸
𝟐𝑪𝒐𝒔 (𝝎𝑳𝑶 − 𝝎𝑩𝑩)𝒕 −
𝑰−𝑸
𝟐𝑪𝒐𝒔 (𝝎𝑳𝑶 + 𝝎𝑩𝑩)𝒕
= 𝑰+𝑸
𝟒𝑪𝒐𝒔 (𝟐𝝎𝑳𝑶 − 𝝎𝑳𝑶 + 𝝎𝑩𝑩)𝒕 + 𝑪𝒐𝒔 (𝟐𝝎𝑳𝑶 + 𝝎𝑳𝑶 − 𝝎𝑩𝑩)𝒕
- 𝑰−𝑸
𝟒𝑪𝒐𝒔 (𝟐𝝎𝑳𝑶 − 𝝎𝑳𝑶 − 𝝎𝑩𝑩)𝒕 + 𝑪𝒐𝒔 (𝟐𝝎𝑳𝑶 + 𝝎𝑳𝑶 + 𝝎𝑩𝑩)𝒕
= 𝑰+𝑸
𝟒𝑪𝒐𝒔 (𝟑𝝎𝑳𝑶 − 𝝎𝑩𝑩)𝒕 + ⋯ . -
𝑰−𝑸
𝟒𝑪𝒐𝒔 (𝟑𝝎𝑳𝑶 + 𝝎𝑩𝑩)𝒕 + ⋯
Outputs @ ω = 3 𝝎𝑳𝑶𝟏
Dror Regev
Frequency conversion technical consideration Inter-Modulations in Mixers:
𝑻𝑿𝟐𝟏 = Iinter-modulation IMD21 = In Band Impairment=
𝒂 ∙ 𝑪𝒐𝒔(𝟒𝝎𝑳𝑶𝒕) ∙𝑰+𝑸
𝟐𝑪𝒐𝒔 (𝝎𝑳𝑶 − 𝝎𝑩𝑩)𝒕 −
𝑰−𝑸
𝟐𝑪𝒐𝒔 (𝝎𝑳𝑶 + 𝝎𝑩𝑩)𝒕 =
a𝑰+𝑸
𝟒𝑪𝒐𝒔 (𝟑𝝎𝑳𝑶 + 𝝎𝑩𝑩)𝒕 + 𝑪𝒐𝒔 (𝟓𝝎𝑳𝑶 − 𝝎𝑩𝑩)𝒕 -
𝑰−𝑸
𝟒𝑪𝒐𝒔 (𝟑𝝎𝑳𝑶 − 𝝎𝑩𝑩)𝒕 + 𝑪𝒐𝒔 (𝟓𝝎𝑳𝑶 + 𝝎𝑩𝑩)𝒕
Outputs @ ω = 3 𝝎𝑳𝑶
𝑰𝑴𝑫𝒏𝒎 = ±𝒏𝒇𝑳𝑶 ± 𝒎𝒇𝑹𝑭 Where n, m integers 0, 1,2,3
Special interest should be paid for IMD products with m=1 which
inherently have higher power levels
Dror Regev
Frequency conversion technical consideration M21 Impairment in Heterodyne 1/3, 2/3f0:
Up converter (as well as Down converter) mixer present:
significant 2LO-1RF inter-modulation product around the same center frequency.
IF @ fRF/3 RF=1LO2+1IF
IM21=2LO2-1IF fRF
WB Signal
Intermod. a
fLO2 =
2*fLO1
fLO1 =
fRF/3
BBI
00/900
BBQ
IF1
IF2
fRF/3
a- Mixer IM21 suppression can be improved to a certain level by
improving Balun Balance
Dror Regev
Summary
• High mm-Wave wafer probing relevant when:
– Chip ASP and performance is critical as with PAs.
– Packaging or final module expensive as with MCMs.
• Socket testing is limited today to ~ 30GHz
• OTA testing is good for characterization and pose challenges in HVM environment.
• “Contact less” probing has potential advantages.