EEVblog #814 - Keysight N8762A 600V 5100W PSU Teardown

[Phoenix]

Another reason might be some safety issue if the component get loosen.

There is a standard (can't remember which) that says if a component is close to the chassis and can be pushed to make contact with the chassis using 10N it's unsafe - the capacitor on the input board would need glue to pass this... but the chassis appears to have a layer of plastic on the inside.

Are the extra wires in the CT's some sort of self test?

[Arjan Emm]

This power supply looks very similar to the genesys 2U series from TDK Lambda.
http://uk.tdk-lambda.com/products/product-details.aspx?cid=3&scid=273

Looks the same!
Schematics for that one I wonder?

On first glance the usermanual doesn't seem to mention an option to use only 2 phases. It can produce an ac fail error. I can't find the specifics of that error. But that error could come up if one of the phases is missing. If, after replacement of the fuses and mov this error comes up, this could be an indication that unit is working fine but needs it's 3rd phase.
The manual does state that there are 2 versions, 240 en 400v. This one is clearly for the american market.

[VK3DRB]

Dave, a good tear down.

Australia is NOT a 240V country. It is a 230V country. Vic and NSW still run 240V as nominal, but Australia as a country is officially classed as a 230V nominal voltage country. In any case, 247 V is nowhere near the "extreme limit of what is allow here". The limit is 264V in Vic and NSW. I get 255 volts here at the fuse box here but it is well within the maximum limit here in Victoria.

The reason for this power supply blowing up reminded me of a colleague at IBM who was electronically as thick as a brick. He plugged our mains into a 110V machine imported from the USA, blowing the crap out of it. He had checked the voltage on the machine but thought we had 110V in our mains and the Americans has 240V. It turns out he was a lemon because he had no idea whatsoever about basic electrical safety or anything he was hired to do. I believe he faked his qualifications and his experience. Some months after he joined the company, he was fired and literally marched to the door. A big lesson learnt.

I bought one of these Agilent power supplies before. 60V max, and very high current. It was badge engineered from TDK-Lambda, as I suspect this power supply you reviewed is. In fact you could buy pretty much the same power supply under different brand name. A terrific power supply, but mine only had the huge screw terminals at the back. For lower current applications, they could have furnished a few banana sockets, cutting of the output at these sockets at say 10A via a relay. It was programmable, so it was used to stress test designs and for general R & D use.

It is annoying not having intelligent labels on connectors etc, but they don't want anyone in the general public to mess around with the insides. I noticed a few small devices have their top over lay white ink cut off by vias. Time pressure on the PCB layout bloke maybe, but usually it is just plain laziness. Their component libraries are not consistent either. The fly vomit pin 1 markers on IC's for same package pin spacing have different sizes. (Triangles are a much better pin 1 markers than fly vomit). Plus some components have their designators inside a box, others do not. I can only guess this is because they have no component library standards. This is not what I would call PCB artwork. It is clearly not Hewlett Packard quality from the days when it was a test equipment company.

[bigsky]

We'd all love to see this beast repaired. If you are very lucky, it may just be the fuses and MOV that blew. Next would be the input bridge rectifiers.

I found the manual here:

http://www.matsolutions.com/Portals/0/Product%20documents/Agilent%20Technologies/N8736A/N8736A%20User's%20Guide.pdf

I can't see any mention of missing phase detection so it may be OK running it from a single phase supply across two of the input terminals.

The manual says (page 95) that the input range is 170-265 Vac for this version, so it should be OK off your somewhat high mains supply.

It wouldn't be too difficult to get a proper 120/208v three phase supply if needed. There are many ways of doing this.

If you can get a 3 phase 240/415v supply (which I'm sure your building will have), this could be stepped down. A three phase transformer might be a bit hard to find, but you could do it with three identical 240-120v single phase transformers. For testing purposes they could be small ones. You could also use 6 transformers and go via an intermediate voltage (eg 240-12-120v).

If getting access to three phases is awkward, then the VSD approach suggested by gardner might work. Or a small motor/alternator set. Or some sort of static converter. It's a common problem faced by people who need to run three phase motors from single phase supplies, and there are plenty of solutions.

[bigsky]

L1 Fuse blown
L2 MOV blown
L3 Fuse Blown
Seems to be some sort of pattern here. 

Both fuses blew instantly, the remaining Line was then shorted to earth by the MOV.
Never seen a ceramic fuse blow hard enough for the magic white dust to escape from inside.

I think the MOVs will be delta connected, across the phases. I wonder if the MOV that blew took out the fuses on the phases feeding it.

I'd advise caution with using it over single phase.
As 5,100W over 240V is 22Amps, and the start up current could be higher, after all it does have 32 Amp fuses on all 3 phases (well did have).

I don't think you realise that the AC current draw will depend on the load connected to the DC side. With no load I can't imagine it would take more than 0.5A per phase. Obviously its full capacity will not be possible from a single phase supply (assuming it will run from one) but that's not relevant to a repair.

[EEVblog]

Australia is NOT a 240V country. It is a 230V country.

I know. I still call it 240V.

[EEVblog]

I'd advise caution with using it over single phase.
As 5,100W over 240V is 22Amps, and the start up current could be higher, after all it does have 32 Amp fuses on all 3 phases (well did have).

Not unless you were silly enough to fully load it during repair 

[Monkeh]

I don't think you realise that the AC current draw will depend on the load connected to the DC side. With no load I can't imagine it would take more than 0.5A per phase. Obviously its full capacity will not be possible from a single phase supply (assuming it will run from one) but that's not relevant to a repair.

But there may be little or no inrush current limiting.

[LINACboy]

Spot the difference!
This is was taken from a TDK-Lambda GEN300-17:
Same series, but 300V/17A and with proper 3~/400V input.
That Agilent unit is a complete rebadger.

[Arjan Emm]

Just turn the voltage up using a variac. No inrush problems..

[Spamlobster]

Could be, but the exact same leaded parts don't have it elsewhere, so if so it wouldn't be (or shouldn't) because of an actual leaded component modal reason. So could be a PCB modal component involved. But I suspect the reason for the stuff is not nearly this complicated.

I think it's just additional mechanical support to prevent twisting them when you're operating heavy machinery (ie fingers) around them. I keep noticing something that requires handling near those glued caps, such as the fuses, pin headers/connectors, screwed on cables, the output board you need to semi-blindly mount upside down...

[TheAmmoniacal]

Of course the Agilent is a rebadger, but that's not a problem as long as they rebadge from the best.

I also found a technical manual (no schematics unfortunately) for the Genesys U2 5kW from TDK-Lamda which looks identical: https://dl.dropboxusercontent.com/u/15526481/docs/83515000.pdf

[rs20]

Spot the difference!

No ethernet interface on that one?

[TheAmmoniacal]

What are the financial incentives to rebadging this stuff? Agilent buys 1000 units for 30% off to make 30% profits?

[PChi]

I have had large electrolytic capacitors make a disconcerting click when they break down inside and blow the fuses leaving no external evidence. They may work OK with some performance loss after the event but can't be trusted in the long term.

[Yansi]

There are many VFDs available that are designed to run three-phase industrial equipment (pumps, saws, mills etc) from single/split phase supplies.  I believe you will find something from China that is designed for 120V that can generate three-phase 208V at 0-300Hz for $150 or so.

Pretty stupid and dangerous suggestion. Don't fall for that!

VFD's must be NEVER EVER used as a single to three phase voltage converters. Unless you know hell of a lot what you are doing.  In the VFD there is absolutely NO output voltage ripple filtering, the output is pure square pulse output, as the current is being smoothed by the motor winidng inductance. Depending on the modulation scheme and control algorithms, the VFD can (or most times cannot) be modified with external LC filters to function as a one-to-3-phase converter to supply other equipment, than plain 3ph motors.

[LINACboy]

What are the financial incentives to rebadging this stuff? Agilent buys 1000 units for 30% off to make 30% profits?

I'd assume, the customer will always have to pay extra for the Agilent badge - even if Agilent themselves might get bulk discounts. So the only scenario where it might make sense to buy these devices would be when Agilent won a tendering for some kind of turn-key plant where the contract states that they are responsible to provide maintenance services for all sub-systems - general contractor style.

[gardner]

In the VFD there is absolutely NO output voltage ripple filtering, the output is pure square pulse output

In cheap VFDs that will be the case, yes.  The output waveform is comparable to what you see from a "modified sine wave" inverter.  It sucks.

But you'll have to clarify in what way this is "dangerous" or inappropriate to testing a piece of gear that requires three phase.  It would undoubtedly be inefficient, and would likely confuse the hell out of the PFC.  Where is "danger" or "stupid"?

[Lukas]

Spot the difference!
This is was taken from a TDK-Lambda GEN300-17:
Same series, but 300V/17A and with proper 3~/400V input.
That Agilent unit is a complete rebadger.

Looks like Agilent rolled their own I/O (LAN, etc) board. See also https://www.eevblog.com/forum/testgear/agilent-n5772a-600v-2-6a-1560w-power-supply-teardown/

[LINACboy]

In the VFD there is absolutely NO output voltage ripple filtering, the output is pure square pulse output

In cheap VFDs that will be the case, yes.  The output waveform is comparable to what you see from a "modified sine wave" inverter.  It sucks.

But you'll have to clarify in what way this is "dangerous" or inappropriate to testing a piece of gear that requires three phase.  It would undoubtedly be inefficient, and would likely confuse the hell out of the PFC.  Where is "danger" or "stupid"?

Your're both right. A proper motor VFD produces a PWM sine voltage at its output. It is designed for inductive loads. In fact, it needs the motor inductance to low-pass filter the output current. Because of this, each phase winding is subjected to the modulated square voltage from the VFD. The harmonic content is what increases voltage stress in the motor insulation compared to a mains driven motor and why VFD operated motors require better quality magnet wire for the same life expectancy.

Unlike a motor, the DC-PSU is a resistive load. Simple solution: Add a couple of series inductors between VFD and PSU to make it appear more inductive. However, this will only work for steady state operation of the PSU. Load transients could excite oscillations between the filter inductors and the DC link capacitors. Controlling those requires careful design.

Single phase operation (with a variac for inrush current limitation!) is the way to go for testing, provided that phase-loss detection, active PFC or any other active input circuitry won't protest.

[Fungus]

What are the financial incentives to rebadging this stuff? Agilent buys 1000 units for 30% off to make 30% profits?

It works out to be more than 30% because they have no R&D costs for this unit. Building their own might need them to open more offices, look for more engineers, etc.

Plus: It's a good way to test the water for a new line of products and find out how well they sell. They can build their own units later on if it makes financial sense to do so.

If it means one more "Agilent" badge in a lab then that's good, too.

There's lots of good business reasons if you think about it.

[Len]

What are the financial incentives to rebadging this stuff? Agilent buys 1000 units for 30% off to make 30% profits?

To fill in a gap in their product line, so their customers aren't tempted to look at another company's catalog.

[ConKbot]

I'd actually like to see a davecad™ of the PFC section.  I get how single phase active PFC works, but 3 phase wouldnt play so nice with the simple "plop a boost converter after the rectifier" type used in single phase supplies. I'm seeing a few different typologies online, and I'm curious as to what someone like TDK lambda would use.

[firewalker]

Why 3 bridge rectifiers?

Alexander.

[Monkeh]

Why 3 bridge rectifiers?

Alexander.

Because it's 3-phase?