EEVblog 1691 - Uni-T UDP6731 360W Bench PSU REVIEW

[EEVblog]

The Uni-T UDP6700 series UDP6731 360W slimline Bench Lab Power Supply REVIEWED

00:00 - Uni-T UDP6731 PSU
01:30 - Beware of thermal stacking
02:55 - Teardown
07:35 - Sneaky fan
11:24 - Remember the Riden RD6006, also 350W
13:10 - Power up
16:04 - The display and UI
18:14 - Instantaneous knob
19:16 - Only 600 memory settings, meh.
20:10 - No ability to disable voltage sense, and some oddities.
23:22 - Accuracy spec
26:13 - What fan noise?
27:45 - Have I KILLED it?
28:46 - WTF is going on here?! Rigol or Uni-T problem?
32:05 - Did my Rigol Electronic Load survive?
33:03 - The Uni-T survivied a brutal oscillation test
33:47 - Ripple & Noise
35:36 - 2ohm load waveform
36:41 - Switch on and other waveform tests
40:31 - Remote software
44:00 - Lister mode tested
47:39 - Woah, this is a problem!
49:46 - Conclusion

[samipetri]

Rigol has bug at FW. Input terminals are shorted at bootup and toggle CV CC short is gone if remember corretly. I found this bug and reported to Rigol at 2020. We used DL3021 at automated tester and burned DUT's because of this weird bug.

This was my original message to seller and after that i talked to Rigol support and received new FW witch corrected problem.

Problem is when i switch power to Rigol load and after bootup, input connectors are 0.2ohm.
When i change mode to CV and back to CC load inputs goes normal "open ohms" over 200kohms.
Input goes short also when remove voltage from input connectors before i switch load ON ->OFF.
When load is OFF input connectors are still shorted.

@Dave Maybe some Rigol debug video before update FW?

[Kleinstein]

Having 2 separate heat sinks is kind of a must for saftey: the switching transistors are at mains potential, the diodes are at the output.

The 200 banks of memory looks like the memory left. A few (like 10) can makes some sense, more are rarely useful.

The turn off when it detects a large difference between drive and sense is nice. It still is a bit slow to really save a circuit.
Some high resistance linke from sense to drive also makes sense - so just a loose sense wire would not cause the voltage to go way to high and blow the circuit.

The programmable part is a thing one would really want to do via the PC and not the fiddely interface at the PS.

[EEVblog]

Having 2 separate heat sinks is kind of a must for saftey: the switching transistors are at mains potential, the diodes are at the output.

Yes, you are right.

[Nominal Animal]

I really appreciate the ramp/overshoot tests.  Those, the oscillation, and the noise (both peak-to-peak and RMS), are the only things that really matter to me –– besides the user interface, of course.  The listing mode rising edge "bump" is very concerning to me.  I've had components burn due to the power-on voltake "spike" on a cheapie 30V/5A switching supply.

To me, it looks like a software design problem: see how the bottom of the bump is the centerpoint between start and end of the ramp?  Both in time and voltage?  I suspect that instead of just going to the target voltage, they add a midpoint to avoid overshoot, but at low voltages the rise time is faster than they calculated, causing the bump.  The fix is to adjust the midpoint voltage and/or time calculation, taking into account the rise/fall time differences.


If I were to create a new programmable PSU, my UI design/simulation phase would be exactly what Dave did! 

I've talked about this before, but I believe all UIs (software and hardware) really should be mocked up and tested on unwitting victims users at the design phase.
Grab a display with the same pixel pitch, 3D print a mock front bezel, and use an USB microcontroller interface board for the individual buttons and encoders and LEDs, connected to a suitable computer.  Throw it to an UX design team with a junior Python/Javascript RAD developer.  For the test scheme, the application simulates the supply and the load.  Observing how a few unwitting vi- users interact with the device, "simulating" a couple of typical scenarios perhaps –– just like Dave did with the PSU here! –– is worth more than what an UX design specialist thinks it should be like, because it is a tool and not an art piece.
The design end result documents exactly how the UI should work, making the electronic design much easier –– fewer iterations, you know exactly what kind of data the UI needs to work correctly ––, plus later on, allows comparisons between real world behaviour and the simulator/specification, which can help when fixing firmware bugs.  If done before ee design has started, it can help estimate the manufacturing costs and thus whether the product is likely to be profitable.

It always puzzles me why others don't see the value in such testing on tentative users, observing how they interact with the device.  Everyone just copies what others do, instead of making the UIs actually better.  It's not like it is expensive to do; just not part of the application/tool/gadget design culture, methinks.

[EEVblog]

The listing mode rising edge "bump" is very concerning to me. 

Uni-T are unable to reproduce the Lister step waveform issue that I saw. There could possibly be an issue with my unit, they are sending a replacement.
I can't imagine what the fault would be that doesn't show up in other modes.

[EEVblog]

If I were to create a new programmable PSU, my UI design/simulation phase would be exactly what Dave did! 

I've talked about this before, but I believe all UIs (software and hardware) really should be mocked up and tested on unwitting victims users at the design phase.
Grab a display with the same pixel pitch, 3D print a mock front bezel, and use an USB microcontroller interface board for the individual buttons and encoders and LEDs, connected to a suitable computer.  Throw it to an UX design team with a junior Python/Javascript RAD developer.  For the test scheme, the application simulates the supply and the load.  Observing how a few unwitting vi- users interact with the device, "simulating" a couple of typical scenarios perhaps –– just like Dave did with the PSU here! –– is worth more than what an UX design specialist thinks it should be like, because it is a tool and not an art piece.
The design end result documents exactly how the UI should work, making the electronic design much easier –– fewer iterations, you know exactly what kind of data the UI needs to work correctly ––, plus later on, allows comparisons between real world behaviour and the simulator/specification, which can help when fixing firmware bugs.  If done before ee design has started, it can help estimate the manufacturing costs and thus whether the product is likely to be profitable.

It always puzzles me why others don't see the value in such testing on tentative users, observing how they interact with the device.  Everyone just copies what others do, instead of making the UIs actually better.  It's not like it is expensive to do; just not part of the application/tool/gadget design culture, methinks.

This is why I like to "fiddle" with it straight out of the box, as much as that annoys a lot of people.
I think it's an important test of the "obviousness" of the UI if I can just play with it out of the box without reading the manual.

This test with David2 was funny, this is a horrible PSU UI

[Nominal Animal]

This is why I like to "fiddle" with it straight out of the box, as much as that annoys a lot of people.
I think it's an important test of the "obviousness" of the UI if I can just play with it out of the box without reading the manual.

I definitely agree.  And I do read the manual after wards, before using the tool for realsies.

Intuitiveness is important, because it dictates how easily you reach for that particular tool.
It is okay if some features take an "oh, so that is how they thought the workflow would go" moment, but you do want most common use cases to be as intuitive as possible.

This test with David2 was funny, this is a horrible PSU UI

I remember the time in the oughties when I did some web UIs.  At that time, it was common-ish that if you pressed the Back button in your browser, the server side lost its marbles (i.e., the data you were submitting got lost, and you had to re-login).  (Mine didn't, of course.)  I don't know who or what idiot thought that that would be acceptable in a tool UI, because I know for a fact it caused hundreds of hours of lost man-hours of work at at least the Uni I worked for.

If anyone does the kind of unwitting user test, even if you tell them beforehand that pressing the browser Back button will cause the tool to fail, the first time they notice they forgot something on the previous page they'll press the browser Back button anyway, not your custom "please go to previous page" button-thingy on the page itself.  We humans just work that way.

I don't expect to idiot-proof tools, though; "mostly intuitive" suffices.

With appliances that have both touch screen and physical buttons, we see how stupid most UX design still is: they repeat functionality, instead of complementing each other.  If you do the design the other way, looking at how people use such tools in anger (especially mock-up interfaces and unwitting test users), you end up with more powerful but simpler UIs.

(And I don't mean the UI should be what most people expect it to be; efficiency is more important.  You measure the time it takes for each victim to achieve a task, and add penalties for errors they do, plus extra points/penalties based on user's mood/attitude/satisfaction/frustration.  This is what the UX specialist is for: looking at how people try to use the thing, and then devise a layout/structure/interface that lets them do those things efficiently.  Then they collaborate with an industrial graphics artist, and make it pretty.)

[HeMan]

Great review, thanks for that. But I note you missed any comparisons with similar supplies, so the question on everyone's mind (or maybe just mine) is, if I were buying a few hundred dollar PSU (in my case I probably don't need 80v, so would go with a lower model from the range), would I buy this or is there still a better alternative for the same kinda money?

The Siglent SPD3303X-E seems a well regarded option; more flexible, but much lower power output, or the Rigol DP811; USB\LAN, but far more expensive.

Thanks.

[Brumby]

This is why I like to "fiddle" with it straight out of the box, as much as that annoys a lot of people.
I think it's an important test of the "obviousness" of the UI if I can just play with it out of the box without reading the manual.

The introduction of the Tektronix 465 is testament to that.


This comes under the heading of "I just want to drive my car to do stuff, not spend 2 weeks of intensive training just to get it out of the garage."

[Kleinstein]

Great review, thanks for that. But I note you missed any comparisons with similar supplies, so the question on everyone's mind (or maybe just mine) is, if I were buying a few hundred dollar PSU (in my case I probably don't need 80v, so would go with a lower model from the range), would I buy this or is there still a better alternative for the same kinda money?

The Siglent SPD3303X-E seems a well regarded option; more flexible, but much lower power output, or the Rigol DP811; USB\LAN, but far more expensive.

Thanks.

The mentioned Rigol and Sigilent supplies are linear lab supplies with a classic mains frequency transformer.
The UDP6731 is a pure switch mode supply. This makes it way more noisy, but also relative cheap and low weight for the power level. So it is a thing if one needs quite some power, but not a really clean voltage.

The other, DIY set together supply Dave showed is similar. The PeakTech P 6227 or OWON SPE6103 are likely similar, though not sure if they have an extra linear stage after the SMPS part, as some higher end ones have. Given the low price - likely not.

[EEVblog]

Great review, thanks for that. But I note you missed any comparisons with similar supplies, so the question on everyone's mind (or maybe just mine) is, if I were buying a few hundred dollar PSU (in my case I probably don't need 80v, so would go with a lower model from the range), would I buy this or is there still a better alternative for the same kinda money?

Unless I did a complete shootout with half a dozen different units (which I don't have), I couldn't tell you.

[Furna]

Dave,

I guess you need to update the fw on SDS824X HD ... micro symbol is now µ while your still has u ?

One day I could even buy one of those switching thing but currently I do not feel the need.

[nctnico]

Great review, thanks for that. But I note you missed any comparisons with similar supplies, so the question on everyone's mind (or maybe just mine) is, if I were buying a few hundred dollar PSU (in my case I probably don't need 80v, so would go with a lower model from the range), would I buy this or is there still a better alternative for the same kinda money?

The Siglent SPD3303X-E seems a well regarded option; more flexible, but much lower power output, or the Rigol DP811; USB\LAN, but far more expensive.

You can't compare those PSUs with the UDP6731 because the latter is a switching power supply. You have to divide power supplies in two types:

1) linear ones (with a big mains transformer) which have low output noise at the expensive of poor efficiency

2) switching ones which are lighter, can output more power but also output a lot more noise. Basically these are current limited, adjustable AC/DC power supplies.

For doing measurements, a switching PSU is a big no unless you can't avoid it due to power levels.

If you want to buy a cheap, linear power supply then look at Korad. These seem to be well regarded on this forum.

[Macao2Blue]

at 11min you are comparing the size and what you get from a quality point of view.

However, the RD6006 uses synchronous topology correct ?
And I assume with those large diodes with heat sinks the Uni-t uses asynchronous ?

I have recently looked inside a 27amp DC-DC converter with max. ripple less then 100mVolts, and the size of it is less then a credit card, full length heat sink, but it has NO fan, gets warm at most. Could not believe it!

[Nominal Animal]

I'm happy with my Gophert NPS-1601 0-30V 0-5A switchmode supply that VoltLog reviewed in #255.  It's my only "real" lab PSU.
Ripple is just 8mV peak-to-peak, and it does not have any voltage overshoot when powering on or enabling the output.  This is extremely important to me, because a lot of my stuff cannot tolerate too-high supply voltages at all.

(In constant current mode it reacts to sudden decrease in load resistance rather slowly, and can seriously exceed the current limit momentarily due to bulk capacitance on its output, but this does not affect my use cases.)

It is quite simple single output switchmode supply, easy-ish UI, surprisingly trustworthy in my use cases, and obviously cheap: I paid under 54€ including shipping and VAT 24% five years ago from Banggood.

The Gophert NPS-1600 (0-16V 0-10A) would fit my needs even better –– my SBCs and Linux appliances that take 5V to 12V in can draw 6A-8A (short repeated peaks) when loaded with spinning rust external drives and 4G/5G/LTE modems and other wireless radios and such ––, but I haven't seen any reviews NPS-1600; not the kind Dave did for the Uni-T here, and VoltLog did for the NPS-1601 five years ago, checking the power-on ramping and noise characteristics on a scope.

This just shows that this kind of review does affect people, at least us hobbyists who have burned their fingies with poorly behaving supplies before.

[ballsystemlord]

@Dave , You seem to have forgotten to test the PSU at it's 360W like you wanted to.
You got distracted by the Rigol instrument.

[EEVblog]

@Dave , You seem to have forgotten to test the PSU at it's 360W like you wanted to.
You got distracted by the Rigol instrument.

It was working fine at I think 365W as I briefly showed in the video (in fact many minutes before I hit record again for that segment), but then I turned it up and then all hell broke lose.

[ballsystemlord]

@Dave , You seem to have forgotten to test the PSU at it's 360W like you wanted to.
You got distracted by the Rigol instrument.

It was working fine at I think 365W as I briefly showed in the video (in fact many minutes before I hit record again for that segment), but then I turned it up and then all hell broke lose.

250W 26:45 seconds in. You said that 250W was the maximum capability of the BK precision 8601.
I just fast-forwarded through the video. Before that segment is the partial teardown and your other remarks before testing. After that is the Rigol bug, then testing with the oscilloscope, then the examination of the SW suite.

[HeMan]

Ah, well, Kleinstein you have made the situation far, far worse! 

The problem I have is, going by Dave's review, the UDP6731 (although I would go with the UDP6730) seemed a good buy.  And although it would have been nice to have a "yes, buy it" from Dave, I appreciate he can't compare them all, so was about to stump up the cash anyway.  But then you mentioned the OWON and looking at their range, the SPE3103 seems a far better choice, on paper at least.  The voltage and current range is better (for my needs), the ripple lower*, accuracy similar, readback resolution better etc. And best of all, it's a third of the price of the Uni-T.

I'm only really using it for hobbyist charging of batteries, powering microcontrollers and related modules and electronics, so nothing particularly sensitive, but the USB and better computer functionality of the OWON, and the better resolution of current readback (1mA) is tempting for helping calculate battery life for any given project**, is enough to cause a real personal turmoil; probably more than it should... 

* But worrying they get this figure with an external capacitor and have a spelling mistake in "eht" [sic] disclaimer in the specs!

** But for anyone doing this, the Nordic Power Profile Kit II is brilliant.

[malman]

I wonder if the list issue is a sw bug related to that each step represents two possible changed settings… for each step it first sets a voltage, and then it sets a current…. The current setting might interfere with the ramping to the new voltage… 

[EEVblog]

I wonder if the list issue is a sw bug related to that each step represents two possible changed settings… for each step it first sets a voltage, and then it sets a current…. The current setting might interfere with the ramping to the new voltage…

Welcome to the forum.
Could very well be. But Uni-T say they are not seeing the issue, so 

[EEVblog]

The problem I have is, going by Dave's review, the UDP6731 (although I would go with the UDP6730) seemed a good buy.  And although it would have been nice to have a "yes, buy it" from Dave, I appreciate he can't compare them all, so was about to stump up the cash anyway.  But then you mentioned the OWON and looking at their range, the SPE3103 seems a far better choice, on paper at least.  The voltage and current range is better (for my needs), the ripple lower*, accuracy similar, readback resolution better etc. And best of all, it's a third of the price of the Uni-T.

Yes, that's the problem with any individual review unfortunately. So you'd have the look for others who have reviewed other similar models to see if there are any showstoppers.
"Shootout" videos between different brand/models are another type of video entirely.

[Kleinstein]

The video with a trear down and test of one such instrument is already quite long. Here the issues with the Rigol load added even more. A note on the types (SMPS , linear) of power supplies would be nice.
From the pictures it looks like the Uni-T is just a SMPS. However the heat sink at the output side is relatively large for just diodes - though a bit on the small side to also have a linear post regulation. There is this 3rd group of supplies with a SMPS followed by a linear stage. These can have a bit faster response and less output ripple, but still have a good efficiency.

For really measuring the current for a battery powered device the dispaly at the power supply is not really the best choice. These type of supplies are made for large currents and thus not good in measuring in the mA range. The higher current capability also comes with some limitations in the accuracy of the current regulation at low currents. It is just the wrong instument for this.

[HeMan]

Yes, that's the problem with any individual review unfortunately. So you'd have the look for others who have reviewed other similar models to see if there are any showstoppers.
"Shootout" videos between different brand/models are another type of video entirely.

OK, so I guess the million dollar question (although, asking as a hobbyist, my budget is a couple of hundred £, so hopefully not a million dollar answer) is would you buy the Uni-T or something else (the OWON?)?

For really measuring the current for a battery powered device the dispaly at the power supply is not really the best choice. These type of supplies are made for large currents and thus not good in measuring in the mA range. The higher current capability also comes with some limitations in the accuracy of the current regulation at low currents. It is just the wrong instument for this.

Yeah, I take your point, which is why I use the Power Profile Kit if I need accuracy, but if I just want a quick idea, or to spot if something is wrong, the 1mA resolution of the OWON seems better then the 10mA of the Uni-T, as does the accuracy (≤0.1%±10mA vs anything up to <0.3%+30mA for the Uni-T).