For price, delivery and to place orders: Hittite Microwave Corporation, 2 Elizabeth Drive, Chelmsford, MA 01824
Phone: 978-250-3343 Fax: 978-250-3373 Order On-line at www.hittite.com
Application Support: Phone: 978-250-3343 or apps@hittite.com
DATA CONVERTERS - SMT
10
HMC660LC4B
v04.0514
0.02 - 4.5 GHz WIDEBAND
3 GS/s TRACK-AND-HOLD AMPLIFIER
DISCONTINUED PRODUCT
Not Recommended for New Designs
Linearity Measurement and Calculation
When characterizing the linearity of a T/H, the transfer function linearity of the held samples (referred to as T/H-mode
linearity) is usually the quantity of most interest to the user. These samples contain the signal information that is
ultimately digitized by the downstream A/D converter. Since the T/H-mode linearity is often different than the track-
mode linearity, this presents a unique measurement issue in that the linearity of only the hold-portion of the analog
output waveform must be selectively measured.
This issue is aggravated for high speed T/Hs because there are few wide-band time domain instruments (oscilloscopes
or A/D converters) with sufficient linearity to characterize a high linearity T/H. Therefore a frequency domain instrument
(spectrum analyzer) and measurement technique are used which allow selective characterization of the hold-mode
portion of the waveform
A common approach to this requirement has been to cascade two T/Hs in a dual rank conguration such that the
second T/H (T/H 2) re-samples the output of the rst T/H (T/H 1). The two T/Hs are usually clocked 180 degrees out-
of-phase in master-slave operation to eliminate the track-mode portion of the output waveform from the rst T/H. This
arrangement produces an output waveform that consists of two time segments. The rst segment is the T/H 1 hold-
mode output as observed through the T/H 2 track-mode transfer function. The second time segment is the T/H 1 hold-
mode output re-sampled and held by the T/H 2 device. This measurement approach is not a perfect representation
of the linearity of a single T/H due to the impact of the second T/H on the overall linearity. However, it does eliminate
the track-mode portion of the T/H 1 output and permits a spectrum analyzer linearity measurement of the cascaded
devices. Since T/H 2 only has to sample the held waveform from T/H 1, the linearity impact of T/H 2 is primarily
associated with its DC linearity. An often used approximation is that the DC linearity of T/H 2 is much higher than the
slew-rate dependent, high frequency linearity of T/H 1 so that the total non-linearity of the cascade is dominated by
the high frequency linearity of T/H 1. In this case, the dual rank conguration has a net linearity that closely resembles
the linearity of a single T/H, particularly at high frequencies. However, this approximation is not always valid. If not, the
dual rank conguration fails to represent the linearity of a single T/H. The HMC660LC4B represents such a case; the
3rd order nonlinearity of this device varies relatively slowly with frequency and is high enough over the T/H bandwidth
that the DC linearity of the 2nd T/H signicantly impacts the overall dual rank conguration.
Another linearity measurement issue unique to the T/H device is the need for output frequency response correction.
In the case of a dual rank T/H, the output waveform resembles a square wave with duration equal to the clock period.
Mathematically, the output can be viewed as the convolution of an ideal delta-function sample train with a single
square pulse of duration equal to one clock period. This weights the output spectral content with a SIN(πf/fs )/(πf/fs)
(Sinc) function frequency response envelope which has nulls at harmonics of the clock frequency fs and substantial
response reduction beyond half the clock frequency. The spectral content of the held samples without the envelope
weighting is required for proper measurement of the sample’s linearity. Either the impact of the response envelope
must be corrected in the data or a measurement method must be used that heterodynes the relevant nonlinear
harmonic products to low frequencies to avoid signicant envelope response weighting. This latter method is referred
to as the beat-frequency technique.
The beat-frequency technique is commonly used for high-speed T/H linearity measurements, although the
measurement does impose restrictions on the specic input signal and clock frequencies that can be used. For
example, with a clock frequency of 512.5 MHz, a single tone input at 995 MHz beats with the 2nd harmonic of the
sampling frequency (through the sampling process) to produce a 1st order beat product at 30 MHz. Likewise, the 2nd
and 3rd harmonics of the input signal (generated via distortion in the T/H) beat with the 4th and 6th harmonics of the
sampling frequency respectively to produce 2nd and 3rd order beat products at 60 MHz and 90 MHz. In this manner,
the T/H nonlinearity in the vicinity of 1 GHz can be measured even though the 995 MHz fundamental and the 1.99
GHz and 2.985 GHz nonlinear harmonics are well beyond the 206 MHz bandwidth of the Sinc response envelope.
The possible input frequency choices are overly limited when the low frequency beat-product technique is used at high
clock rates. A related high frequency beat-product measurement utilizing correction for the Sinc envelope weighting
must be employed to measure linearity over a wide range of input frequencies. Hittite uses both low frequency and
high frequency beat product methods to measure linearity for a wide range of clock and signal frequencies. Our high
frequency beat-product measurement avoids excessive envelope correction error by maintaining all beat products
within the 4 dB bandwidth of the Sinc function, where the envelope response is well behaved and easily modeled.
Hittite has also developed a method for accurately measuring the held-sample linearity of a single T/H using a beat-
frequency technique that avoids errors due to nonlinear products associated with the track-mode portion of the