Tektronix Technical Forums are maintained by community involvement. Feel free to post questions or respond to questions by other members. Should you require a time-sensitive answer, please contact your local Tektronix support center here.

SMU 2657A overheating?

Model 2657A
Post Reply
JSchoeck
Posts: 3
Joined: October 27th, 2016, 7:36 am
Country: Germany

SMU 2657A overheating?

Post by JSchoeck » October 27th, 2016, 7:52 am

Hello,

we are using a 2657A SMU for testing of high power devices. Until recently we were using the web interface for breakdown measurements on devices that were manually contacted with a needle. A ramp from 0 to 3000 V with 20 mA current compliance was used.
To increase throughput and statistics we adapted our wafer prober to integrate the 2657A with the prober's installed 4200-SCS SMU via GPIB. Using KITE we can measure the same ramp successfully.

Two problems occur:
1. The ramp measurement, using identical parameters, is much slower. With a voltage step of 5 V, no added delays and nPLC 3 it takes 36 s in the standalone case and about 2 min with the 4200/prober setup. Why? How can I change this?

2. When scanning a whole wafer the 2657A outputs an error message after about 20-30 devices measured that it is "too hot" and slows down the measurement to about 15 min per device. I haven't seen this during standalone use, but of course there were more breaks in-between measurements - and they ramp was faster.

I'd be glad for any input. Thanks in advance!

Best regards,
Johannes

brian.d.smith
Keithley Applications
Keithley Applications
Posts: 402
Joined: June 10th, 2013, 6:00 am
Country: United States

Re: SMU 2657A overheating?

Post by brian.d.smith » October 29th, 2016, 2:54 am

We need some more information.

1. How have you integrated the 2657A with your 4200? Did you write a UTM or are you using ACS Basic? Or some other method?

2. What is your ramp rate when sweeping from 0 to 3000V?

3. What current do you consider to be breakdown?

20mA at 3000V sounds like more reverse current that I have seen in any breakdown test. This sounds like a destructive test. I have seen a breakdown test where the voltage reached 250V at 1 mA and the devices are always destroyed. For this device, 1uA was destructive and the test sweeps three decades in current farther than destruction and six decades farther than the knee of the device's breakdown curve.

A non-destructive breakdown test will choose a current that shows the knee of the breakdown curve such as 1nA. The exact current will depend on your device.

You force this current with a voltage compliance higher than needed. When the 2657A output is enabled, the instrument output automatically rises to the voltage where 1nA is achieved and the breakdown voltage can be measured. No sweep is required. No device destruction occurs.

Be aware that the 2657A is slew limited in the range of individual volts per microsecond. So, a ramp from 0 to 3000V will require some milliseconds. If you try to sweep from 0 to 3000V faster than a few milliseconds, the instrument will not comply. The slew limitation is an analog phenomenon and cannot be changed by the user.

JSchoeck
Posts: 3
Joined: October 27th, 2016, 7:36 am
Country: Germany

Re: SMU 2657A overheating?

Post by JSchoeck » October 31st, 2016, 1:03 am

1. We are using KITE on the 4200 and are transmitting the commands to the 2657A in a UTM.

2. When using the 2657A by itself, the ramp rate was quite fast (maybe 50-200 V/s), but via the 4200 the ramp rate slows down significantly (using the same parameters!). 0 - 3000 V with 5 V steps takes about 2 min, not miliseconds.

3. The measured current is < 1 mA, usually around 50 µA at 3 kV. I chose max. compliance to see as much as possible of the reverse behaviour. We are working on SiC power devices, so destructive breakdown should neither happen at 3 kV, nor at 20 mA. So the 2657A does not have the voltage to actually trigger breakdown on functional devices, but the purpose is to find the yield of devices stable at 3 kV.

Andrea C
Keithley Applications
Keithley Applications
Posts: 1262
Joined: October 15th, 2010, 10:35 am
Country: United States
Contact:

Re: SMU 2657A overheating?

Post by Andrea C » October 31st, 2016, 6:43 am

If the UTM is slower than the TSP-Express based exection, it would be due to some difference in implementation.
Possible areas of difference: NPLC setting, filter, filter factor, limited auto low range for current measurements.
Also, how is the sweep being implemented? In the UTM, are you using one of the KISweep functions or is each individual source voltage being written from a for/next loop?

As for the too hot situation, are you getting the error 5046 from the SMU?
If yes, this concerns me.
Are the cooling vents obstructed?
Check pg 31 in the 2657A Reference Guide about the cooling vents.

JSchoeck
Posts: 3
Joined: October 27th, 2016, 7:36 am
Country: Germany

Re: SMU 2657A overheating?

Post by JSchoeck » November 4th, 2016, 2:53 am

Thank you for the feedback about timing issues. We didn't yet implement the filter functionalities, but I'd expect that the default value for the TSP-Express and UTM would be identical, so it shouldn't make a difference. We will test this. Is the "samplingCount" setting in the configuration save file (.prj from TSP-Express) the filter setting? Or is this the equivalent to smua.measure.count for the fast ADC?
And what does "stepPoints" exactly do - is it only used when programming a sweep internally? Sadly I cannot find any information about these settings in the reference manual.

Auto range was used with the TSP-Express routine and it was quite fast. We will try how different settings are affecting the measurement time.
As for the implementation, we are sending all commands individually to the SMU, so every source voltage is set by itself and the measurement value retrieved. Would a different implementation significantly increase the speed?

Regarding the overheating: yes, it is the error 5046. When it occurs the SMU waits for several minutes (up to 10 min have been observed) before it continues the measurement. The wafer-map time increases from under one hour to over five hours this way.
All cooling vents are free, the SMU is rack mounted with enough space for circulation and there is basically nothing hot below it. The reference guide doesn't help further than stating that the error is serious and that you shouldn't obstruct vents, so we are thankful for all hints.

Is the 2657A actually capable of long-term testing at high voltages? What is the expected power dissipation that will not lead to overheating?

brian.d.smith
Keithley Applications
Keithley Applications
Posts: 402
Joined: June 10th, 2013, 6:00 am
Country: United States

Re: SMU 2657A overheating?

Post by brian.d.smith » November 4th, 2016, 2:11 pm

If you are never exceeding 50uA at 3000V and below, you should never have a heat problem.

My guess is that the 2657A is oscillating and that the oscillation is what is heating up the instrument. Can you get a high voltage probe and record the sweeps with an oscilloscope?

Is your device highly capacitive?

Post Reply

Return to “2657A High Voltage SourceMeter”

Who is online

Users browsing this forum: No registered users and 1 guest