Design Article
Optimizing a switch system for mixed signal testing
Dale Cigoy, Keithley Instruments Inc.
5/25/2011 8:21 AM EDT
Channel-to-channel Isolation
Channel-to-Channel Isolation: This specification is related to leakage and crosstalk between adjacent signal paths through the switch assembly. Troubleshooting problems caused by leakage and crosstalk can be a difficult task. It’s much easier to start system development with the proper switch design and specifications than to spend precious time troubleshooting an elusive problem, which applies to other potential error sources.
Most switch assemblies are printed circuit board (PCB) cards that are inserted into a switching type of measurement instrument, or into a switching mainframe used with separate instruments. Therefore, the electrical isolation between any two adjacent switches may be expressed in different ways depending on the intended use of the switch card. Normally, the switch channels on the PCB are aligned in order to achieve proper voltage isolation and accommodate the physical dimensions of the switches and other components, such as connectors. This spacing and the PCB material provide a certain level of isolation between channels. The higher the isolation, the lower the chance of crosstalk or leakage. Typical values of channel-to-channel isolation are up to 10GW with capacitance of less than 100pF see figure 2.
In high-frequency applications, leakage capacitance becomes an important consideration. For these applications, isolation is usually stated in dB. For instance, 60dB would be an isolation of 1000 to 1 from channel to channel, meaning that a 1V signal on one channel could bleed over and become a 1mV signal on an adjacent channel. Keep in mind that open contact isolation resistance must also be taken into consideration when developing the switching portion of a test system. The higher the isolation between open contacts and between adjacent channels, the better the integrity of signals passing through the system.
Offset current: This current can occur on switch cards even when no test signal is present. The largest magnitudes are due to finite coil-to-contact impedance in electromechanical relays. It is also generated by triboelectric, piezoelectric, and electrochemical phenomena on the switch card, regardless of relay type.
Offset current is important, for example, in low level and high impedance measurements taken during semiconductor parametric testing on wafers and individual devices. Low offset current becomes a critical specification when conducting leakage current measurements on semiconductor devices and materials. It is also important when doing semiconductor C-V characterization.
Depending on card design and intended usage, the offset current specification could range from less than 1pA up to 1nA. Manufacturers of cards designed for relatively high level DC current and voltage switching may not provide an offset current spec, because it is normally unimportant in those applications.
Relay Switching Speed: A relay’s operating speed has a direct affect on the throughput of a switching type of test system. The system developer has to pay attention to relay speed specifications to also ensure that measurements are accurate. A typical test scenario is to apply a stimulus to the device under test (DUT), wait a short amount of time for the test system and DUT to react and settle to a final value, and then measure the DUT’s response. If measurements are taken before the system has settled sufficiently, results can be inaccurate.
Relay operating speed is a measure of the rate at which its contacts can be cycled and still obtain reliable operation. This rate is limited by the relay’s actuation and release times. Actuation time is measured from when power is applied to the coil until the contacts have settled. Thus, actuation time includes contact bounce time. Release time is the opposite of actuation time. It is measured from the time power is removed from the coil until the contacts have settled to their open position and includes bounce time.
A significant amount of system settling time is associated with relay bounce time, which must have settled out before a solid reliable connection is established in the signal path. Settling time varies from relay to relay with typical times in the millisecond range. Sometimes relay switching cards have a built-in delay to avoid problems associated with contact bounce. In addition, some switching equipment may even have a user programmable delay time.
Use of Solid-State Switches: Standard electromechanical relays can switch from one state to another in as little as a few milliseconds, which is fast enough for some applications. However, in production applications where test time carries a significant dollar value, this switching time may be too long. Solid-state relays (e.g. transistors, FETs) have a much faster switching time, generally below one millisecond. Going from a few milliseconds to a few hundred microseconds could shave off substantial test time and increase test throughput.
Another advantage of solid-state relays is their reliability. Solid-state relays have a switching life of almost 100 times that of electromechanical relays. This would be on the order of about 10 billion switch cycles instead of a good electromechanical relay’s life of about 10 million cycles.
One disadvantage is the “on” resistance of solid-state relays, which is on the order of tens of ohms. Such a high resistance could lead to measurement inaccuracies in a twowire resistance measurement. Trying to measure a few milliohms with upwards of 10W of resistance in the circuit from the “on” resistance would effectively bury the low-resistance measurement.
One way around this is to use a so-called golden or standard channel. This is a channel with a short on the device side. The channel is closed, the resistance measurement is made, and the measurement is subtracted from all other channels. Therefore, the “on” resistance is essentially zeroed out. The problem is that this holds for only the golden channel and would be slightly different on each channel. Using this method would depend on the magnitude of the resistance to be measured and the accuracy required.
Another way to correct for this resistance is the four-wire (Kelvin) measurement technique, which involves using two channels instead of one. One channel is used to source the current and one to sense the voltage. This is a standard method to measure low resistance. Using an electromechanical or reed relay would only have a contact resistance of tens of milliohms, which would be more advantageous in low-resistance measurements using the two wire method.
Other Settling Time Issues: In addition to mechanical issues, there are electrical issues associated with the opening and closing of switches. When a mechanical relay opens or closes its contacts, there is a charge transfer on the order of picocoulombs that causes a current pulse in the test circuit. This charge transfer is due to the mechanical release or closure of the contacts, the contact-to-contact capacitance, and the stray capacitance between signal and relay drive lines. This phenomenon can affect both the signal settling time and signal integrity.
The nature of the signals must also be considered. Some signals originating from a DUT take longer to settle than others. As a general rule, the rise time of a DUT output signal is defined as the time for it to rise from 10% to 90% of its final value when the stimulus rises instantaneously from zero to some fixed value. If the signal originates from an extremely high impedance (producing a very low current), then it may require several seconds or even minutes to settle. The settle time is directly related to the small current charging the cable or stray capacitance in the circuit. The higher the impedance the lower the current and the more time it takes to settle out.
Making sure a test system has settled sufficiently is the key to good measurements. Specifications listing relay actuation time are only the starting point in determining the total test time for a measurement sequence. The mainframe or switching instrument holding the switch card also contributes some overhead time, which is a function of its command to connect times in a test sequence. This varies according to the test sequence design, but some switching instruments and mainframes have displays that depict when a relay has closed, providing some indication of how fast a sequence is progressing. However, keep in mind test system design always involves tradeoffs between throughput and accuracy issues.
A much more detailed version of this paper which additionally covers topics of "Switch System Architecture and Topology", and "Switching Test System Design Tradeoffs" is available at Optimizing a switch system for mixed signal testing.
About the author:
Channel-to-Channel Isolation: This specification is related to leakage and crosstalk between adjacent signal paths through the switch assembly. Troubleshooting problems caused by leakage and crosstalk can be a difficult task. It’s much easier to start system development with the proper switch design and specifications than to spend precious time troubleshooting an elusive problem, which applies to other potential error sources.
Most switch assemblies are printed circuit board (PCB) cards that are inserted into a switching type of measurement instrument, or into a switching mainframe used with separate instruments. Therefore, the electrical isolation between any two adjacent switches may be expressed in different ways depending on the intended use of the switch card. Normally, the switch channels on the PCB are aligned in order to achieve proper voltage isolation and accommodate the physical dimensions of the switches and other components, such as connectors. This spacing and the PCB material provide a certain level of isolation between channels. The higher the isolation, the lower the chance of crosstalk or leakage. Typical values of channel-to-channel isolation are up to 10GW with capacitance of less than 100pF see figure 2.
Fig 2: Representation of channel-to-channel isolation with shunt capacitance and resistance.
In high-frequency applications, leakage capacitance becomes an important consideration. For these applications, isolation is usually stated in dB. For instance, 60dB would be an isolation of 1000 to 1 from channel to channel, meaning that a 1V signal on one channel could bleed over and become a 1mV signal on an adjacent channel. Keep in mind that open contact isolation resistance must also be taken into consideration when developing the switching portion of a test system. The higher the isolation between open contacts and between adjacent channels, the better the integrity of signals passing through the system.
Offset current: This current can occur on switch cards even when no test signal is present. The largest magnitudes are due to finite coil-to-contact impedance in electromechanical relays. It is also generated by triboelectric, piezoelectric, and electrochemical phenomena on the switch card, regardless of relay type.
Offset current is important, for example, in low level and high impedance measurements taken during semiconductor parametric testing on wafers and individual devices. Low offset current becomes a critical specification when conducting leakage current measurements on semiconductor devices and materials. It is also important when doing semiconductor C-V characterization.
Depending on card design and intended usage, the offset current specification could range from less than 1pA up to 1nA. Manufacturers of cards designed for relatively high level DC current and voltage switching may not provide an offset current spec, because it is normally unimportant in those applications.
Relay Switching Speed: A relay’s operating speed has a direct affect on the throughput of a switching type of test system. The system developer has to pay attention to relay speed specifications to also ensure that measurements are accurate. A typical test scenario is to apply a stimulus to the device under test (DUT), wait a short amount of time for the test system and DUT to react and settle to a final value, and then measure the DUT’s response. If measurements are taken before the system has settled sufficiently, results can be inaccurate.
Relay operating speed is a measure of the rate at which its contacts can be cycled and still obtain reliable operation. This rate is limited by the relay’s actuation and release times. Actuation time is measured from when power is applied to the coil until the contacts have settled. Thus, actuation time includes contact bounce time. Release time is the opposite of actuation time. It is measured from the time power is removed from the coil until the contacts have settled to their open position and includes bounce time.
A significant amount of system settling time is associated with relay bounce time, which must have settled out before a solid reliable connection is established in the signal path. Settling time varies from relay to relay with typical times in the millisecond range. Sometimes relay switching cards have a built-in delay to avoid problems associated with contact bounce. In addition, some switching equipment may even have a user programmable delay time.
Use of Solid-State Switches: Standard electromechanical relays can switch from one state to another in as little as a few milliseconds, which is fast enough for some applications. However, in production applications where test time carries a significant dollar value, this switching time may be too long. Solid-state relays (e.g. transistors, FETs) have a much faster switching time, generally below one millisecond. Going from a few milliseconds to a few hundred microseconds could shave off substantial test time and increase test throughput.
Another advantage of solid-state relays is their reliability. Solid-state relays have a switching life of almost 100 times that of electromechanical relays. This would be on the order of about 10 billion switch cycles instead of a good electromechanical relay’s life of about 10 million cycles.
One disadvantage is the “on” resistance of solid-state relays, which is on the order of tens of ohms. Such a high resistance could lead to measurement inaccuracies in a twowire resistance measurement. Trying to measure a few milliohms with upwards of 10W of resistance in the circuit from the “on” resistance would effectively bury the low-resistance measurement.
One way around this is to use a so-called golden or standard channel. This is a channel with a short on the device side. The channel is closed, the resistance measurement is made, and the measurement is subtracted from all other channels. Therefore, the “on” resistance is essentially zeroed out. The problem is that this holds for only the golden channel and would be slightly different on each channel. Using this method would depend on the magnitude of the resistance to be measured and the accuracy required.
Another way to correct for this resistance is the four-wire (Kelvin) measurement technique, which involves using two channels instead of one. One channel is used to source the current and one to sense the voltage. This is a standard method to measure low resistance. Using an electromechanical or reed relay would only have a contact resistance of tens of milliohms, which would be more advantageous in low-resistance measurements using the two wire method.
Other Settling Time Issues: In addition to mechanical issues, there are electrical issues associated with the opening and closing of switches. When a mechanical relay opens or closes its contacts, there is a charge transfer on the order of picocoulombs that causes a current pulse in the test circuit. This charge transfer is due to the mechanical release or closure of the contacts, the contact-to-contact capacitance, and the stray capacitance between signal and relay drive lines. This phenomenon can affect both the signal settling time and signal integrity.
The nature of the signals must also be considered. Some signals originating from a DUT take longer to settle than others. As a general rule, the rise time of a DUT output signal is defined as the time for it to rise from 10% to 90% of its final value when the stimulus rises instantaneously from zero to some fixed value. If the signal originates from an extremely high impedance (producing a very low current), then it may require several seconds or even minutes to settle. The settle time is directly related to the small current charging the cable or stray capacitance in the circuit. The higher the impedance the lower the current and the more time it takes to settle out.
Making sure a test system has settled sufficiently is the key to good measurements. Specifications listing relay actuation time are only the starting point in determining the total test time for a measurement sequence. The mainframe or switching instrument holding the switch card also contributes some overhead time, which is a function of its command to connect times in a test sequence. This varies according to the test sequence design, but some switching instruments and mainframes have displays that depict when a relay has closed, providing some indication of how fast a sequence is progressing. However, keep in mind test system design always involves tradeoffs between throughput and accuracy issues.
A much more detailed version of this paper which additionally covers topics of "Switch System Architecture and Topology", and "Switching Test System Design Tradeoffs" is available at Optimizing a switch system for mixed signal testing.
About the author:
Dale Cigoy is a Senior Application Engineer at Keithley Instruments in Cleveland, OH. His major responsibility is helping customers with measurement applications that include Keithley equipment, especially DMMs. Prior to this he wrote technical instruction manuals for Keithley products. Cigoy joined Keithley in 1976 after earning a Bachelor of Science degree in Electronic Technology from Capitol College in Laurel, MD.
|
|
|
|
|
Navigate to related information