R&S HMP4040 voltage spike at PSU turn-off

[electr_peter]

Rohde & Schwarz HMP4040 PSU has outputs enabled/disabled by relays. Output relays should eliminate negative/positive spikes (or DC leakage in off position). That is what you would think until you discover that all outputs at PSU turn-off become charged to 2.3V!

Still investigating WTH is happening.

[nctnico]

In what time frame? Scope image showing the effect?

[electr_peter]

In what time frame? Scope image showing the effect?

~500ms rise time, see attached pic. Didn't capture a tail, it dropped off quickly after that. Voltage was visible on DMMs with 10M and 1M input impedances for few seconds.

I will upload more detailed shots in next few days, need to check across all 4 channels at once and check potential power.

[electr_peter]

There are some measurement results of switching PSU on/off via mains switch:

• Measurement setup - each channel of HMP4040 is connected to a Micsig STO1104E scope (ground on output-, probe on output+), 10x input.
• All 4 channels behave almost the same each time - there is small spike on PSU turn-on and bigger spike (charge & decay) on PSU turn-off.
• Turn-on spike is ~100ms and ~0.5V
• At turn-off, voltage ramps up to ~2.3V in 220-300ms, stays stable for 5.4s and then decays in 5.8s (with no load, only 10M scope probe).
• With 10kOhm, 1kOhm and 100Ohm loads, turn-off voltage spikes are lower in magnitude, but have the same duration and shape.
• Estimated current with different loads - 10M scope probe 0.23uA; 10kOhm load 190uA; 1kOhm load 650uA; 100Ohm load 1.2mA.

HMP4040 is capable to produce sustained (5-10s) voltage spike (up to 2.3V and up to ~1.2mA) on it's outputs when outputs are in turn-off state and PSU is switched off from mains. Likely same happens on R&S HMP2020/HMP2030 and HMP4030 as they are very similar inside (same channel module).

Potential power to damage circuit under test is limited, but not zero. Should there be a concern with 2.3V/1.2mA turn-off spike?

I do not have schematic, but observations point to PSU sense circuitry. Sense terminals are always active, there is no option to turn them ON/OFF. In off state there is 10kOhm resistance on output terminals for each channel, that should be sense circuitry. Output is disconnected via relay, while sense terminals are probably disconnected via some transistor. My guess is that during PSU turn-off transistor may be opened earlier than voltage decays inside.

[nctnico]

That doesn't look too good and a proper PSU should not behave like this. It also looks as if there is quite a bit of energy available. There are plenty of chips out there that run from 1.8V and can't take much more. It is surprising nobody else has noticed this though so you might have a defective unit.

[electr_peter]

Take in mind that sample size is one until other HMP users confirms this. This issue is sneaky and hard to notice unless looked at very carefully.

In Gough Lui road test review of HMP4040 something related was noticed, but not looked in too deeply.

The only potentially major issue was the behaviour of the rails when power is removed from the power supply unexpectedly. In the test case, an unloaded 1V rail rose to 10V as the hardware power switch was actuated – so it’s best to turn off the rails first using the output button before turning off the power to the supply to ensure the DUT is not damaged.

It was also of interest to investigate what the behaviour of the HMP4040.04 is when power is abruptly removed during operation. Some other power supplies have specific warnings not to do this as regulation can be lost on such a power-down. Looking at the scope traces above, an abrupt power-down using the red hardware power button results in a less-well-defined rail power down timing, along with significant over-voltage on the first rail configured as 1V which shoots up to 10V. As a result of this behaviour, in case of emergency, it is better to use the master output switch rather than the hardware power button to ensure no risk of damage to the DUT.

Links:
R&S 4-Output Bench Power Supply, Prog (HMP4040.04) - Review
R&S HMP4040.04 PSU RoadTest in Depth – Ch5: Instrument Performance Testing

[mawyatt]

About 40 years ago we lost some very expensive ($50K) custom chips to a power supply that produced an over voltage glitch at turn-off like OP has shown. Expensive lesson and this was a quality, expensive lab PS, think it was Power Designs.

Anyway, we made sure it wouldn't do this again after taking a hammer to it, then sending it to in-house cal lab!!

This is something not to be taken lightly, it can and will destroy sensitive ICs as mentioned by nctnic0

Best,

[2N3055]

About 40 years ago we lost some very expensive ($50K) custom chips to a power supply that produced an over voltage glitch at turn-off like OP has shown. Expensive lesson and this was a quality, expensive lab PS, think it was Power Designs.

Anyway, we made sure it wouldn't do this again after taking a hammer to it, then sending it to in-house cal lab!!

This is something not to be taken lightly, it can and will destroy sensitive ICs as mentioned by nctnic0

Best,

 

[nctnico]

After checking the review: having a 1V rail shoot up to 10V is a no-go. A long time ago an intern at work tested a batch of boards (20 or so) and he used a PSU that also had the output shoot up when switching it off (way worse compared to the R&S PSU thouch) and he accidentally blew up every board right after putting a 'test passed' sticker on it.  CMOS is not forgiving where it comes to overvoltage.

[mahi]

Another Rohde & Schwarz HMP4040 user here...

I can confirm that the mains power switch does affect the (turned-off) outputs. On switch-on there's a short 0.3-0.4 V pulse on all outputs, and on switch-off the outputs ramp up to 1 V and stay there for several seconds.

However, I am unable to reproduce output 1 climbing above 1 V, let alone 10 V... I've retried multiple times, with different set voltages (outputs off), but the traces always looked the same. I did not experiment with different loads - only the oscilloscope or a multimeter.

The screenshots below are switching on and off with the mains switch, no load (other than the oscilloscope) and all outputs off.

Switch-on

In about 10 ms all outputs ramp up to 0.3-0.4 V and then fade back to 0 V in 30 ms. A fluke 289 in peak detect (250µs response time) measures the peaks between 0.33 and 0.38 V.



Switch-off

The outputs ramp up to 1 V in less than a second. Output 1 ramps up noticeably faster than the other three outputs. It gets a 600 ms head start, but even without that the slew rate is much higher.

Just under 6 seconds after switching off, the outputs suddenly drop to 0.5 V in under half a second. From there it takes another 4 seconds before the outputs finally reach 0 V again.

A Fluke 289 in peak detect does not read higher than 1.04 V over the entire period.



For reference: My HMP4040 is running the latest firmware (2.72). I don't know the hardware revision, but it is a "pre-facelift" unit (light front and single fan).

The Element14 review by Gough Lui shows a later "facelift" unit. Note that this power supply was seriously damaged in transport.

Which version do you have, electr_peter?

[nctnico]

Another Rohde & Schwarz HMP4040 user here...

I can confirm that the mains power switch does affect the (turned-off) outputs. On switch-on there's a short 0.3-0.4 V pulse on all outputs, and on switch-off the outputs ramp up to 1 V and stay there for several seconds.

However, I am unable to reproduce output 1 climbing above 1 V, let alone 10 V... I've retried multiple times, with different set voltages (outputs off), but the traces always looked the same. I did not experiment with different loads - only the oscilloscope or a multimeter.

From reading the review, the output seems to shoot up with the outputs enabled. This is something that could happen during an unexpected power outage.

[mahi]

From reading the review, the output seems to shoot up with the outputs enabled. This is something that could happen during an unexpected power outage.

Good point. I missed that.

I've retested with all outputs enabled and outputting 1 V. When switching off via the mains switch a similar profile is visible like posted earlier:



However, when zooming in at the point where the voltage first falls (or should fall), things do not look very good... Output 1 climbs over 10 V, output 2 touches 5 V and the remaining two channels peak around 4 V before dropping to 0 V (and then commencing with the ramp up to 1 V):



The output artifacts do not change with the output voltage. Thus output 1 with a voltage of 10-11 V and the other outputs 5 V suppress the effect:

[electr_peter]

@mahi, thanks for the input. Your pictures of turn-off ramp matches 1kOhm load ramp (ramp, drop, ramp to full voltage). Maybe there is voltage dependancy, channels were initialy set to higher voltages (9-12-32V).



My HMP4040 has grey/white front panel (single color), FW is 2.72.

[mahi]

electr_peter: I just retested, but the set voltage does not affect the voltage ramp-up after mains switch-off at all (with the outputs turned off).

The screenshots I posted were all with 10M input impedance (1M oscilloscope + 10x probe). I don't have a 100x probe but I added 9x 10M on a breadboard for a 100M input impedance. Horrible noise but the voltage and signal profile stayed the same (1 V maximum).

Then I tested with different loads. The waveform always had the same profile but the maximum voltage was lower:

10K: Maximum voltage 0.9 V
1K: Maximum voltage 0.5 V
100R: Maximum voltage 0.1 V

Under no circumstances I get over 1 V on the outputs in your scenario.

The issue that Gough Lui observed on the other hand... Nasty.

[electr_peter]

And things get even worse. If sense terminals are shorted to to output terminals (normal operation, +/+ & -/-), during turn-off DMM reads 2.64V open or 2.2mA shorted. This also points to sense circuit as a main suspect.
I do not find voltage dependency w.r.t. to output spike level as well. In addition, turn-on spike increased from 480mV to 630mV with sense terminals connected.



With all 4 channels set to 1V and turned on, abrupt power off results in up to 5.4V voltage spike and followed by the same turn-off ramp-up/down.



In all these shots violet channel (3rd) has sense terminals shorted to output terminals. Sense terminals do not mean that they are sensitive to DUT, quite the opposite in this case.

[electr_peter]

What are output impedances measured with DMM on PSU turned-off?

With sense terminal unconnected:

• between output terminals - 10kOhm
• between sense terminals - 12.9kOhm
• between output- and sense- terminals - 1.47kOhm
• between output+ and sense+ terminals - 1.47kOhm
• between output+ and sense- terminals - 11.5kOhm
• between output- and sense+ terminals - 11.5kOhm

With sense terminal shorted to output (+/+ & -/-):

• between output terminals - 10kOhm

[mahi]

electr_peter: When shorting the sense lines to the outputs, the ramp-up voltage increases to 1.6 V maximum (instead of 1 V) on my HMP4040. This does indeed suggest that the sense circuitry is to blame for this behavior. Still, the voltages on my unit are much lower than on yours.

There is no change to Gough Lui's scenario with the sense lines shorted to the outputs.

The resistances between the output terminals on my HMP4040 are quite different from yours:

With sense terminal unconnected:

    between output terminals - 1.87 kOhm
    between sense terminals - 4.82 kOhm
    between output- and sense- terminals - 1.47 kOhm (same as yours)
    between output+ and sense+ terminals - 1.47 kOhm (same as yours)
    between output+ and sense- terminals - 3.34 kOhm
    between output- and sense+ terminals - 3.34 kOhm

With sense terminal shorted to output (+/+ & -/-):

    between output terminals - 1.87 kOhm

All outputs show identical values within tens of ohms from each other.

[electr_peter]

@mahi, resistances are quite different. There could be some error due to diodes turning on in circuitry and different DMM effect, but that would not account for such big differences. Lower impedance will dampen voltage ramp effect, so there is explanation why yours HMP is more tame.

This data confirms that HMP4040 has different HW versions for channel boards. R&S probably find out this issue and did a HW revision at some point. Do you recall manufacture/sale date of PSU? Mine is from 2015-2016 or earlier.

[electr_peter]

Based on measurements and general PSU configuration I estimated rough schematic of sense circuitry. There are three main resistance elements and two voltage cut-off points (relay for main output and transistor for sense circuit). During turn-off sense transistor becomes activated and leaks voltage. It is clear that different HW versions have different resistive values.

I will try recreating impedances similar to @mahi PSU and see how it compares in terms of turn-off spike.

[mahi]

electr_peter: I can't find any date markings on the outside, but I purchased my unit in spring 2018. Thus it was probably made in 2017-2018.

By the way, your diagram can't be right. You have a direct connection between the sense circuit and the outputs. In your diagram R+ and R- serve no (useful) purpose.

The diagram below (taken from the user manual) is incomplete, but it shows the link between the outputs and sense circuitry.

[electr_peter]

2017-2018 is 1-2 years newer model. HMP4000 series was made by Hameg from 2012 or earlier, this model is still made today by R&S. Predecessor of HMP4040 is Hameg HM7044.

By the way, your diagram can't be right. You have a direct connection between the sense circuit and the outputs. In your diagram R+ and R- serve no (useful) purpose.

Yes, diagram above is incomplete and does not represent functional sense circuit at all, but it is the simplest circuit representation that matches impedance measurements across output and sense terminals in PSU OFF state. Resistive network is complicated by active circuitry (diodes and ICs), thus impedances look strange.

Actual circuit would be similar (but more complicated) to linked in your post above which in off state reduces to much simpler representation. Exactly what happens in off state is difficult to say without exact schematic. Also, your linked schematic gives a hint why shorting sense and output terminals can give more voltage/current.

[SparkyBruce]

I checked my HMP2030 this afternoon and I also see the 2.3V ish output on all 3 channels after turning the unit off - even with the latest firmware.
I am collecting a second identical unit tomorrow and I will also check that.
This is all unloaded (apart from the 10M of the DVM).

[electr_peter]

@SparkyBruce, can you also check date of manufacture and resistance across output (in OFF state)? I guess there are 10 kOhm on outputs.

[mawyatt]

This is serious if across all these R&S power supplies, and not just a select rogue few.

If true then seems R&S needs to step up and issue a recall. As mentioned many times this can and will destroy modern small feature CMOS, as it produces gate oxide punctures which are non-recoverable and fatal!!!

Best,

[electr_peter]

There is clearer estimated schematic of PSU with locations of sense and load discharge (?) resistors. Sense resistors are 1.5 kOhm and load discharge resistors are either 10 kOhm or 1.87 kOhm

The resistances between the output terminals on my HMP4040 are quite different from yours:
With sense terminal unconnected:
    between output terminals - 1.87 kOhm
...

By the look of PCB, 1.875 kOhm is four 7.5 kOhm in parallel. Source of PCB pic: "EEVblog #1174 - Rohde & Schwarz PSU Teardowns" at time 09:18

I checked my HMP2030 this afternoon and I also see the 2.3V ish output on all 3 channels after turning the unit off - even with the latest firmware.

Same as in posts above, just with HMP2030 instead of HMP4040. I believe this is HW related and FW would not affect it.