Measurements of leakage current and offset voltage on some OptoFETs and relays

[Echo88]

Few months ago i started a measurement project which had the aim to evaluate the usability of optocouplers with MOSFET-outputs (known as Solid State Relay/OptoFET/ PhotoMOS/OptoMOS...) in voltnut-applications.
The chosen measurement parameters were output leakage current, offset voltage/TEMF and charge injection with about 30 OptoFETs to be tested.
Those parameters are usually specced very conservative (leakage current), seldom (offset voltage) or not at all (charge injection), while they are of interest when designing high precision applications.
This project is ongoing, but will take quite some time due to the lenghty measurement time of each component and other projects.
Therefore i present the so far gathered data here for the interested reader, who might decide to use such OptoFETs.
Keep in mind that not enough samples were tested to guarantee the measured specs/they might differ from batch to batch.

Used measurement equipment:

The used equipment for measuring leakage current/offset voltage are a Keithley 617 and a Keithley 2182A, together with a fitting electrometer-box/thermally insulated box.
The K617 was found to be measuring <=20% low compared to K220/K2500/100GR-resistor, so a measured value of 10pA would actually be 12pA.
This will still give us good data for component-comparison.
All leakage data is taken from the K617 and the data in the spreadsheet is not corrected for the <=20%-error.

The 2182A hasnt yet been tested against a suitable nV-source, since it has been repaired by me, but shows good matching to my µV-calibrator.
So take that data with a grain of salt, till i have tested the 2182A against a nV-source.

The offsetvoltages of the H11FX-samples were measured with a HPAK 34465A, since that was convenient and sufficient due to the high offset voltage.

The measurement data was gathered with a PC connected to the K617/2182A and saved to spreadsheets/graph-pictures, so the process is somewhat automated.
The leakage-samples were properly cleaned in an ultrasonic-bath, rinsed, heated in a small oven and then cooled before testing to mitigate errors due to fingerprints/residue on the OptoFET-case.

Tested OptoFETs:

I wanted to test about 30 OptoFETs, from different manufacturers with different maximum voltage/maximum switching current, to represent the whole spectrum of available components.
So far the data consists of ~28 OptoFETs:
-most have been measured for the leakage current with at least one test sample each at different voltages, up to the maximum specced device breakdown voltage or K617-limit of 100V. Sample is tested bipolar and then worse leakage is taken for the spreadsheet, usually the leakage is very symmetrical as should be the case.
-offset voltages of some PVA-family-OptoFETs were measured and some relays for comparison so far
-exact small charge injection measurements prove to be very difficult and need different test conditions and equipment; im working on that

TLDR:

Attached data:

The data is available per attached spreadsheet and i also attached the offsetvoltage-measurements of some relays and a PVA-family-OptoFET.
I also attached the original document of the H11F1/H11F2/H11F3-OptoFET "A BILATERAL ANALOG FET OPTOCOUPLER.pdf", before it was released to the market. The document seems to be not available anymore from other internet-sources.
It details the offset-voltage (Figure 8.), which is omitted by every other datasheet of the component and fits nicely with the measured data of the H11FX. 

Regarding the attached Offset voltage-measurements:

"AGQ200A4H Offsetvoltage Diagram_03.11.19.jpg" and "AGQ210A4H Offsetvoltage Diagram_03.11.19.jpg" show why latching relays in low TEMF-switching-applications are a must. While the latching-relay takes some 3-4 minutes for the TEMF to settle, after that its offset-voltage is indistinguishable from the copper-short, which was used at the beginning of the measurement.

"PVA3054NS Offsetvoltage Diagram_03.11.19.jpg" shows the component measured at different LED-drive-currents, with a copper-short used at the start of the measurement.
Nicely visible is the TEMF produced by the LED with its slow settling. This shows that its necessary to use a low led-current, when aiming for low offset voltage. Keep in mind that this measurement was done with the DUT hanging in air/connected by thin copper-wire inside the thermally insulated box, therefore the LED slowly heated the component, which led to fairly high offset-voltages.

"Multicomponent Offsetvoltage Diagram_05.01.2020.jpg" shows different components and configurations which are connected to two big copper-tapes on a pcb, inside the thermally insulated box. The copper-tape acts as a thermal conductor to get an isothermal environment for every component.
The components are activated/deactivated sequentially after disconnecting the copper-short:

AGQ210A4H latching relay in differential configuration: low TEMF
AGQ210A4H latching relay in single ended configuration: low TEMF with visible settling due to self heating and maybe a bit heat from the copper tape
NAIS DS2E-S-05 non-latching relay in differential configuration: high TEMF ~10µV
AGQ210A4H latching relay in single ended configuration again: low TEMF at visible higher settling presumed due to the residual heat from the NAIS-relay/copper tape
AGQ210A4H latching relay in differential configuration: low TEMF
PVAxxxx OptoFET at 1mA LED Drivecurrent: low TEMF ~25nV
PVAxxxx OptoFET at 2mA LED Drivecurrent: low TEMF ~50nV
Copper short inserted again
PVAxxxx OptoFET at 2mA LED Drivecurrent: low TEMF ~50nV

One can see that its useless to differentially connect a non-latching relay while expecting lower TEMF.
The PVAxxxx (cant see the partnumber atm) which are connected to the copper tape show a way lower TEMF compared to "PVA3054NS Offsetvoltage Diagram_03.11.19.jpg", seemingly due to better thermal balancing. 

General remarks and future experiments:

Yellow and Green fillings in the spreadsheet indicate worst/best spec from the column.
White package OptoFETs can be influenced by external lighting, so do the measurements in a dark box.
H11FX-offset voltage can be adjusted to near zero µV by controlling the LED drive current, but component-aging will certainly make the offset voltage drift over time.
Surprisingly both the Normally Open and Normally Closed-counterpart from IXYS: CPC1017N and CPC1117N deliver the best and worst leakage-performance compared to all tested OptoFETs.
Only some of the Panasonic-family-datasheets and the AQV212S datasheet contain a TEMF-spec.
PVA3054 and PVA3055 show very slow settling leakage drift of unknown cause.
Generally AQW210S delivers low leakage at low cost and contains two switches in one package, while the Panasonic-family-OptopFETs are more expensive with similar low leakage.

I also had good leakage-results with 2N4117A controlled by a small photovoltaic cell build from BPW34, although the draincurrent through 2N4117A is 75µA max.
By help of member lukier i bought a few of those expensive LS627 OptoFETs with exposed window, which still need to be characterised.
AoE3 Page 260/302 mentions SD210 as a possibility to build extremely low leakage switches, might build and test those also.

I crushed some PVAs with care in a vise, to expose the photocell-assembly within the OptoFET (visible on photo in sixth post, lowest white package on the bottom side).
This enables measurements with external lighting, avoiding TEMF due to led-heating and will maybe get interesting data when tested for charge injection later.

Hope this will help someone.

Regards.

[Cerebus]

Without circuit diagrams of the test setups it's very difficult to know what you mean by 'offset voltage' and 'leakage current'. The same applies to the actual test conditions used to derive these results.

[Echo88]

The offset voltage is the voltage measured at the two output-pins of the OptoFETs by the 2182A, while the OptoFET-internal LED is driven by stated drive-currents coming from a K220.
The offset of the used cables is nulled out by using a coppershort in the thermally insulated chamber after sufficient thermal equilibrium. When the coppershort shows stable readings, then the coppershort gets cut and the measured offset voltage is substracted from the actual offset-reading of the relays/optoFETs.

The leakage current was measured by using a K617 in the configuration shown on page 37: https://www.univie.ac.at/photovoltaik/umwelt/ws2015/K617_original.pdf
Only that i didnt get the reading in Ohms, but instead control the internal voltage source and take the measured current reading.
The slight K617-internal offset current of a few fA is measured at applied 0V to the DUT and is then substracted from the actual measured leakage current at the used voltages.

[Cerebus]

The offset voltage is the voltage measured at the two output-pins of the OptoFETs by the 2182A, while the OptoFET-internal LED is driven by stated drive-currents coming from a K220.
The offset of the used cables is nulled out by using a coppershort in the thermally insulated chamber after sufficient thermal equilibrium. When the coppershort shows stable readings, then the coppershort gets cut and the measured offset voltage is substracted from the actual offset-reading of the relays/optoFETs.

So what you're calling "offset voltage" is the photovoltaic effect on the JFET V_DS of various values of I_F through the LED?

Do you have any evidence that this "offset voltage" isn't affected by V_DS or I_DS, or to put it differently, will this figure change when the JFET is being used in an actual circuit with a voltage imposed across it/current flowing through it?

The leakage current was measured by using a K617 in the configuration shown on page 37: https://www.univie.ac.at/photovoltaik/umwelt/ws2015/K617_original.pdf
Only that i didnt get the reading in Ohms, but instead control the internal voltage source and take the measured current reading.
The slight K617-internal offset current of a few fA is measured at applied 0V to the DUT and is then substracted from the actual measured leakage current at the used voltages.

So are you saying that you are measuring the JFET I_DS at various values of V_DS, presumably with the LED having an I_F of 0, and calling that the leakage current? Or to put it differently, you're effectively measuring the R_DSoff of the JFET at various V_DS values.

[TiN]

Do you have a photos and more details of experimental setup? I could try to replicate that using multi-channel parametric analyzer with both DC and pulsed mode (to avoid self-heating of DUTs). I don't have fancy optoFETs but could steal some optos from various junk boards.

[Echo88]

@Cerebus:
Lets clarify a few things. I use the term OptoFET as a term to summarize all solid state relays which ive tested.
Only the H11FX-devices (and LS627) have an internal JFET-like structures, where the measured offset voltage is indeed of photovoltaic origin and not due to TEMF.
I conclude this based on previous measurements where i heated the H11FX input pins with a soldering iron and a resistor, without powering the internal LED. I shared that info a while ago in another thread which i might find again.

All other measured OptoFETs are based on a MOSFET-output-stage.
When i use the term leakage i actually measure RDSoff at various Vds values while Iled=0, as you say.
I havent tested the influence of Uds/Ids on the offset voltage measurement.
Ids influence will be dominated by voltage losses on the RDSon (when choosing an OptoFET with non-significant RDSon), for example Ids 1nA * RDSon 100R =  100nV offset.
I dont really know how i should measure Uds influence on offset voltage, please suggest a suitable measurement scheme.

I can provide more photos of the setup later today, since im at work right now. Attached "IMG_1459" here, since first post attachment size limit is reached. It shows the internals of the thermally insulated box without cover.

[Kleinstein]

It makes no sense to measure the "offset" voltage with different Uds, as the drain source voltage is the measured offset.

The H11Fx type  offsets look really poor. So I would consider them not a good choice any more, as better alternatives are available. About the only application I would consider them is as a variable resistor - here the JFET structure could be more symmetric and better to control than the MOSFET based devices.

The leakage values look really good for most of the devices. So the leakage specs are are very conservative.

With the film capacitors the dielectric absorption cause very slow settling, and makes it difficult in general to distinguish real leakage from the DA effect. Anyway in many uses extra current due to DA is about as bad as real leakage.

[Echo88]

Measuring reed-switch leakage turned out to take really long, apparently due to the DA (or slow surface charge movement?) of the used glass. Havent found anymore info on that topic and maybe someone can point to documents describing it, apart from AoE3?

Also bought chinese H11F1 which have shown excessive offset voltage sometimes over 1mV and two of those six bought samples werent working at all, so beware the low prices.

Attached the picture of my double-box-leakage-setup.

Visible are older DUTs on teflon slabs and the later used alligator clips to make easier measurements, a LS627, split PVA-OptoFET and a reed-switch based contraption which was intended to automate the whole measurement process by measuring the parameters sequentially.
Turned out that the reed-relays produced way too much TEMF due to self-heating, to measure the offset voltage of DUTs.
To avoid the heating i tried latching reed relays KSK-1E85 and a small electromagnet in 15mm distance, which worked.
Havent made any measurements on the resulting offset voltage yet, but should be very good since the reed-contacts arent heated due to the distance to the electromagnet.

[Kleinstein]

One can distinguish real leakage from DA by looking at the current over time curves when turning on the voltage and also turning off the voltage. On turn off one only sees the DA effect, with a current flow even without an external voltage.  However one still has to wait a long time even if just to start the turn off transient from a well settled state.

With a relay the is also the coil as a third conductive part. Chances are the capacitance between the contacts and coil is larger than directly between the contacts. The capacitance directly between the contacts has some part with just gas and not much glass.
So the DA current can also include the coil and it can make a difference how the coil is connected. If left open the coil part can slowly charge to an intermediate potential and this way have another effect similar to DA.

[Echo88]

The long leakage settling time was observed with different reed-contacts. So theres no coil which could affect the settling/leakage in this case.
Reed relays with its coil around the reed-contact might behave different, but i didnt measure them yet.

[TiN]

Also those coaxes look like usual PTFE isolated type. They are quite long on photo, and could act as capacitor as well.
I would try same test with triax wiring, not only guarded box.

[Echo88]

I suspected my measurement setup at first also and did various tests detailed here:https://www.eevblog.com/forum/metrology/measuring-nanoamps-and-below-like-a-ninja/msg2752712/#msg2752712
Turns out that its really the reed-switch and its DA which led to the long settling time, that was my conclusion at least and the settling times ive seen resemble the measurements of other users.

[bsw_m]

NAIS DS2E-S-05 non-latching relay in differential configuration: high TEMF ~10µV

Confirm DS2E-S-DC5V have a very high TEMF.
The relay was connected so that it short the input of the nanovoltmeter both when the coil is energized and when it is de-energized.
Measurements were made with my V2-38.
On the graph at the beginning - voltage is applied to the coil, then removed.

[dietert1]

In February 2020 when i was studying the optical chopper of a Fluke 845AR i also ordered some of those optofets (CRYSTALONICS FF627 = LS627). They were tedious to obtain and cost US$ 40 each.
When they arrived, in a little test they appeared completely useless. They are n-channel JFETs, so in order to turn the channel off you put negative voltage to the gate. When you shine light through the parts window, the gate-channel diode gets charge carriers, i.e. a small photo current from the gate. In order for that photo current to control the JFET, the gate drive needs to be high impedance. Then with enough light the negative gate voltage reduces and the JFET may turn on.
But that photocurrent is leakage. I put the part back into the bag and there it is.

Regards, Dieter