EEVblog Electronics Community Forum
Products => Test Equipment => Topic started by: Gary350z on January 21, 2017, 01:28:25 pm
-
I recently purchased an Instek GPD-3303S Power Supply.
It works good with the following exception.
The channel 3 output, which is switchable between 5, 3.3 and 2.5 volts, has a very large overshoot when the output is switched on as shown in the attached image.
The peak value of the overshoot is about 9.2V with the output loaded or unloaded. The overshoot is the same regardless if the output is set to 5, 3.3 or 2.5 volts.
The pulse is narrow, about 40us, so it probably has a low energy content.
A 470uf cap across the output will eliminate the overshoot.
My question: Is this overshoot acceptable or bad? If bad, how bad is it? Will it blow out circuits attached to it?
Thanks for any advice.
-
I'll just add some data from my unit, for comparison.
My unit's channel 3 overshoots to approximately 7 V, regardless of which output voltage is selected. For me, the overshoot is ~30µs long (time with Vout > 5) with the channel set to 5 V.
-
I've confirmed the same effect on my GPD-4303S which is the 4-channel version of that supply. On this model channels 3 and 4 are fully adjustable, but I get almost the same transient at start-up on channel 3 at any voltage setting. Interestingly my channel 4 shows no overshoot at any setting.
As far as damaging circuits, I think it will depend on the circuit. I know I'll keep this in mind before I connect anything sensitive to channel 3 from now on - maybe I need to print up a warning sticker.
-
This is a significant overshoot. It really should not happen. If a 470 µF or similar capacitor can fix it, it would be a good idea to add one (maybe a little smaller so the overshoot is limited to 100-400 mV).
-
That startup waveform looks really bad. Not only does it severely overshoot, but it doesn't even rise smoothly afterward. Interestingly, Fgrir's captures of channels 1, 2 and 4 show proper and smooth startups. Something's definitely wrong with channel 3 on these supplies.
Hopefully, a not-too-big capacitor will smooth things out. Don't make it too big, though, because it does store energy, energy which will be suddenly released (i.e., with no current limiting) if your circuit ever has a short in it during operation. So, the capacitor should be just big enough, but not larger.
Of course, the ideal solution is to fix the regulation problem.
-
Apparently Siglent SPD3303() power supplies also have / used to have this problem, and some fixes are discussed here:
https://www.eevblog.com/forum/testgear/siglent-spd3303d-review/25/ (https://www.eevblog.com/forum/testgear/siglent-spd3303d-review/25/)
Depending on how close the designs of these supplies are, it could be useful to read that thread.
-
Thanks for digging that up, djnz. And here's the post with the details of Siglent's fix:
https://www.eevblog.com/forum/testgear/siglent-spd3303d-review/msg299246/#msg299246 (https://www.eevblog.com/forum/testgear/siglent-spd3303d-review/msg299246/#msg299246)
-
OK, warranties are for losers so I cracked my brand-new GPD4303S open to see why this spike is happening.
First theory was that the service manual schematic has R727 going to +5VCC instead of +5VC as it should. This would have eliminated the intended 7mV offset in the voltage feedback signal and would prevent them from setting the voltage setpoint down to zero, or even better below zero to ensure that the integrator stage would saturate low instead of high when Q706 is engaged to disable the output. No joy here though, probing the board it appears that R727 is actually connected to +5VC as it should be and this is either a typo or an old schematic that has been fixed already. :(
So now I am pretty sure it is just a bad design of their disable circuit. With the unit powered on but output disabled, on my channel 3 I am seeing both V and I integrator outputs (U705) saturated high. This means that when the Q706 clamp is released at turn-on the main output FET will be turned hard on until either of the integrators can slew down into regulation. Looking at the inputs to the voltage integrator I see the feedback signal at ~5mV and the control DAC output at ~6mV.
On channel 4 the added offset to the feedback voltage is larger (9mV) due to the higher feedback gain. Looking at my channel 4 while still disabled the DAC output for the voltage control actually sits at about 2.5mV which is below the feedback signal of 9mV so it is able to keep the voltage feedback integrator saturated low when the output is disabled. I think this is why my channel 4 doesn't glitch at startup - the voltage control loop is starting from a disabled state rather than the enabled state that channel 3 was starting from.
The datasheet for the DAC (AD5643R) shows a 2mV typical and 10mV max for the zero-scale output, so it could be a part-to-part variation between my channels, but I think it is more likely a firmware thing that they don't drive channel 3 to a hard zero when off but they do for channel 4. It makes me wonder if a firmware update could fix the problem on the 4303S. If not that then I think a slight increase in the voltage feedback offset voltage might cure the glitch at the expense of a small setpoint error. Changing R727 from 1M to something about 700K would add 3mV to the feedback offset which should be enough to keep the integrator low when disabled. I think I might be willing to trade 3mV offset error for getting rid of this glitch.
However, for the 3303S there is no DAC for channel 3, just a set of resistor dividers to choose the setpoint and a saturated 2N3904 to clamp the setpoint while the output is disabled. I don't expect that transistor will be able to hold the voltage anywhere near the 7mV offset on the feedback signal, so for this model I expect that the integrator will always be saturated high when coming out of the disabled state. I can't really see a simple fix for this model.
Now to decide if I want to solder on my brand-new power supply or just put the case back on and live with the glitch... :-//
-
Interesting investigation, Fgrir. Thanks for the sharing. As a first step, you could temporarily put a resistor in parallel with R727 to confirm the fix. Then, decide if you want to make it permanent. :-/O
-
There is 2nd thread on the GPD3303 supply rather similar. https://www.eevblog.com/forum/testgear/thoughts-on-gw-instek-3303s-supply/msg1118794/#msg1118794 (https://www.eevblog.com/forum/testgear/thoughts-on-gw-instek-3303s-supply/msg1118794/#msg1118794)
I there suggested a possible "fix", by making the control slightly faster. This way the existing 200 µF cap at the output would be big enough to absorb the than 5-10 times smaller peak. It is not a true fix for the root problem but could have other advantages too.
I very much support the analysis of Fgrir: the clamping transistors can not bring the voltage down to below about 50 mV (CE saturation) and thus the integrator goes to the positive limit. One possible way around would be to replace Q707 and Q706 with N MOSFETs, that would allow a lower clamping and thus likely negative saturation (if the OPs offset is not too large). Does not work with channel 4 though, as there is no second transistor.
Another option might be a (PNP) transistor and maybe a diode on the collector side in parallel to c717 with base control from behind the diodes D704/D705. This would limit overshoot to about 1.4 V. It is not perfect anti windup, but would also work with CC to CV transitions.
-
Yes, I hate that this got split into two threads but here seemed more specific to the overshoot problem.
If I was going to make my mod I was leaning towards adding a resistor from the cathode of ZD703 to pin 3 of U704. This would have a similar effect to changing the value of R727 while in the disabled state, but with a much smaller error when enabled. However I am a bit less enthusiastic about my idea for ensuring the integrator stays low while disabled since I realized that in this state the op-amp output is basically short-circuited. On my channel 4 (which as-built enters this state) I am measuring -1.6V at U805A's output. The TL072 can theoretically handle this, but with a 15V rail and the typical short-circuit current of 40mA it is dissipating over 500mW which would raise its temperature by 67C over ambient using the 125C/W published in it's datasheet. For the worst-case 60mA short-circuit current we are looking at a 100C rise above ambient. I don't have any thermometers handy, but touching U805 confirms it is working very hard in the disabled state.
I think all of Kleinstein's suggestions make sense as well, but for me these solutions are becoming more complicated than I want to try to hack in to a $500 power supply. I am leaning towards putting the case back on my unmodified unit and waiting to see if channel 4 dies an early death before I consider any changes to fix the glitch on channel 3. A warning sticker on channel 3 should be enough to keep me from connecting it to anything that can't handle an 8V spike.
Mostly I am going to add this to my pile of reasons^H^H^H^H^H^H^Hexcuses to pay more for top-name gear. If I had been doing paying work instead of troubleshooting this thing I could have easily spent twice as much ::)
-
Mostly I am going to add this to my pile of reasons^H^H^H^H^H^H^Hexcuses to pay more for top-name gear.
Nah, not an excuse -- just prudent shopping. Of course, not everyone lives in an area with good access to top-name gear at good prices, so that's a factor too. In my case, all my power supplies are HPAK or Power Designs, except one -- a Mastech rebrand that behaves badly on shutdown. But that's a topic for another thread.
-
Having the OP to go into negative saturation could in deed be a problem. I agree that adding true anti windup (or changing to MOSFETs would be a bigger modification.
My first suggested "fix" by making the control faster should is relatively straight forward and simple.
As in the plan the control of channels 3 and 4 is rather slow and is missing the usual phase boost one needs in a LDO type regulator.
According to my simulation, the unmodified circuit should show quite some ringing when there is a significant low ESR capacitor (e.g. 470 µF) at the output and a step to low load current. It depends on the ESR of the existing 200 µF capacitor at the output how bad things get. So just adding a capacitor to the output is not such a good idea.
Adding an RC combination parallel to R712 (for channel 3) gives phase boost and is the usual way to dampen that kind of ringing. Channels 1 and 2 also have it, though in a slightly different configuration. As a side effect the speed of regulation gets higher and by this way the overshoot is reduced. In the simulation something like 1,5 K and 50 nF worked well, but the values chosen are not that critical. Already 3.3 K and 10 n gives a significant reduction in the peak and dampen ringing to an acceptable level. So one can chose on how much faster it should get. As this is only AC, it should not have an influence on the set voltage and voltage reading.
-
Even with no external capacitor the transient control on channel 3 is not very good. Channel 3 recovery from overcurrent is a disaster, and while it is better it's not even very good for channel one. I am pretty much set at this point give up on improvements and to consider my GPD-4303S as a low-quality dual output power supply and only use channels 3 and 4 in cases of dire need. It is nice having the 1mA set and readback, but as a power supply this thing leaves a bit to be desired.
A few overcurrent recovery test results:
Load is Aim-TTI LD400P set for transient between 0A and 1A, f=0.5Hz, Duty = 99% at 0A, Slew = 4A/ms
In all cases the power supply is set to 5V 0.9A output.
For comparison (and my own comfort) I also tested my trusty old Topward 6303A which has been powering my prototypes for the last 15 years. IMHO this is what an overcurrent recovery should look like, and this wasn't a particularly expensive supply. :-+
-
Even with no external capacitor the transient control on channel 3 is not very good. Channel 3 recovery from overcurrent is a disaster, and while it is better it's not even very good for channel one. I am pretty much set at this point give up on improvements and to consider my GPD-4303S as a low-quality dual output power supply and only use channels 3 and 4 in cases of dire need.
Thanks for the additional info Fgrir. Really appreciate it. I had been looking at getting one of the GPD-3303S units myself as a first benchtop PSU since the ITT liquidation has made them cheap and widely available. Just pulled the trigger yesterday and bought one NIB for $200. Now I'm not sure I made the right call. Can you help a newbie out and put your findings into a beginner's context for me?
Is this still acceptable for a first benchtop PSU? I had been looking at the cheaper Chinese made units and figured an Instek would be better quality for roughly the same $$. Should your findings scare someone off like me who's just starting out and ready to get their first dedicated PSU or are these findings more for the hardcore "pixel peepers" or whatever the EE equivalent is? If I leave Channel 3 set to 5V and never move it while I've got a circuit hooked up I should be OK with overvoltage issues right? What about overcurrent recovery? Is there a similar technique I can use to mitigate the effect you've found or is the overcurrent a dealbreaker for a newbie? Thank you again.
-
Can you help a newbie out and put your findings into a beginner's context for me?
I would summarize it as follows. Avoid using channel 3 on anything that can't withstand 8-9 volt spikes. If unsure, don't connect it to channel 3. When using channel 3, it's safer to enable the output first, then attach the device or circuit. Although this doesn't mitigate every issue, it should cover most bases. When in doubt, leave it out (of channel 3).
Is this still acceptable for a first benchtop PSU?
It'll work, but it's unfortunate that channel 3 has many caveats. It's extra cognitive load for a beginner. Perhaps stick a warning note on channel 3 as Fgrir mentioned in one of his posts.
I had been looking at the cheaper Chinese made units and figured an Instek would be better quality for roughly the same $$.
That tends to be the case, but as shown here regardless of the name on the badge, there's no substitute for testing/verifying the actual operation and quality of a device.
-
I'll second most of what bitseeker advised - I think it can be a decent supply if you understand it's limitations and for $200 new I doubt you'll find anything much better. I think channels 1 and 2 are not too bad and I wouldn't be too worried about using them in most situations. Channel 3 is pretty nasty and I think I would avoid using it except in cases where the load can tolerate the expected overvoltage, for example a board that has it's own voltage regulator or a load that is tolerant of short spikes. I wouldn't use channel three to directly power a 3.3V microcontroller or logic circuit since I think that might not survive many turn-on cycles.
-
Adding the RC combination in parallel to R712 should also improve the CC-CV mode transitions: In the simulation it does it, if the voltage goes significantly down for some time (e.g. ms range) before coming back from the CC mode. There is a little overshoot when the voltage only drops slightly, but usually not that much, as the CC mode control is also rather slow.
This same limitation (poor reaction after a short phase of CC mode with the voltage not going down significant), may apply to other supplies too. Often it is a kind for RC combination in the feedback path (like the supposed addition) that makes a relatively slow recovery from a lower voltage and still a fast recovery from normal load changes.
To get a significant better CC to CV mode transition, a kind of true anti windup might be needed. Many commercial supplies get away without this and thus might have similar limitations for a short time current limit.
I have an idea how to add true anti windup (described in an earlier post), but this only works really well if the OPs output is always higher than the input (=ADC output here). It depends on the on the power MOSFET gate threshold whether this condition is met. So this type of anti-windup would get difficult or not very accurate (need the extra series diode).
At least in the simulation, an extra RC parallel to R712 to a large part fixes the problems (turn on, CC-CV transition and ringing with capacitive load) and I don't see to much possible downsides.
The simulation showed a possible problem of the circuit not directly related:
If the OPs go significantly negative there can be a significant current flow through the small PNPs (e.g. Q701) base that drive the MOSFETs gate. This might even violate there specs (base current) and thus could be a potential reliability issue. It could also explain why the OP might get quite warm to touch - though under normal conditions this should only happen for a short time during transients. It could be longer time if an external voltage source is applied though. So it might be a good idea to add an extra series resistor to Q701,Q801,Q102,Q402.
-
I made a quick attempt at adding the RC combination in parallel to R712 as proposed by Kleinstein.
However, this turned out to be a bit too much for my soldering skills.
R712 turns out to have a 0603-sized package, and the only 1.5k resistors I have are 0805. I also have no 50nF on hand, so I tried to combine the 0805 resistor with two 100nF 0603 capacitors, in a kind of lopsided Pi-shape on top of R712. This was a bit too difficult for me to pull off at the moment. Maybe there's an easier way to do this. :-DD
-
A resistor of only 0603 size is making soldering a little difficult in deed. My idea was still old style through hole. :palm: Now I understand Fgrir hesitating.
However I doubt there will be an easier solution than adding just 2 small parts with 2 connections to the board.
The free standing version with 3 parts directly on top of a 0603 could be tricky. Even just 2 parts is not easy to solder.
It does not have to be directly at the resistor: one side would also be at the negative output and other bigger parts (e.g. D703). So it could be a fine wire to the right point.
The values are not that critical - 100 nF or 2.2 K should be OK too.
-
You could probably use a small piece of protoboard for the RC filter and then wire it to R712. If it works well, mount the filter board with Kapton tape, hot glue, etc.
-
I gave it another shot and managed to get the tiny components stacked over R712. 100nF, 100nF and 1.5k?. I measured the connections after soldering to verify.
The spike is reduced but still present (expected), but the reduction appears to be less than in your simulation. The spike now reaches only 3.24 volts on this unit.
I don't know why I don't get a reduction as significant as in the simulation, but for me this might be good enough (I mainly need 3.3V and 5V).
-
That doesn't sound too bad, henken. Congrats on getting the spike under control. Post a pic of your handiwork?
-
There are a few uncertainties and simplifications in the simulation. For the peak on enable, I did not include the current limiting and I don't think it will have an influence. Also the ESR of the output capacitor can have an important influence: to little ESR might cause instability (also for the unmodified circuit) and higher ESR allows a higher peak. Similar the resistance of the shunt and the connecting wire and it's inductance has an influence. The extra resistance from the wire can help a little. So simulations one power supply circuit have it's limitations.
In my first simulations I had used a slightly small FET and this way got a smaller peak.
p.s.:
To make sure there is no problem with stability one should do a few tests with the modified supply: Something like load transients, maybe just with a switch and resistor.
-
A resistor of only 0603 size is making soldering a little difficult in deed. My idea was still old style through hole. :palm: Now I understand Fgrir hesitating.
I probably should have mentioned that - if it was thru-hole I'd have been modding the crap out of this thing :-/O
I might still have a go at it in the future, but not until I finish the job this was bought for. That 1mA/1mV set/readback seems pretty unique at this price point and will make some testing I'm going to do much easier. I'm glad that henken was braver than I so we could see that your suggestion offers a significant improvement to the size of the spike. A 3.24V spike is still not ideal, but I can't imagine very many cases where that would cause damage.
Of course this whole exercise already has me eyeing more expensive supplies. I may need to check into that Test Equipment Anonymous thread...
-
bitseeker: I didn't make any photos, and now the unit is put together. The components are quite hidden by cables, and difficult to frame. There's no pretty handiwork, quite ugly actually, with three components stacked on top of a 0603. It did seem solid enough to stay together though, so it should work out.
I suppose I will cobble together some sort of load transient test before I use that channel for anything. I've been wanting to make some kind of dummy load for other purposes..
Fgrir, I bought mine used for around $120, so there was no warranty to void or anything like that. It seemed pretty safe to give it a try, since there's always the possibility to revert the changes in case things don't turn out well, and the risk of this sort of modification causing permanent damage seemed low. If I had bought one new, I would be much more hesitant to try it out.
You're right that the regulation is still not ideal, but I think it will do just fine for my hobby use cases. Perhaps other component values will fit better with the characteristics of the loop, and reduce it even more. I am hesitant to try others now though, since the modification requires a teardown and tricky soldering (for my skill level at least).
-
I don't see this overshoot (or I'm just measuring it wrong?), so maybe not all units exhibit this.
Measured with 0.5A load and no load (first capture is with the load).
-
I don't see this overshoot (or I'm just measuring it wrong?) ...
To make a educated guess on that you should tell us how you measured.
-
I don't see this overshoot (or I'm just measuring it wrong?) ...
To make a educated guess on that you should tell us how you measured.
I had Channel 3 output connected to a electronic load, and I measured with 10x probe directly from the output...
First with load on at 0.5A, and then with the load at 0A (off). Just normal "single shot" capture when pressing "output" button on the PSU...
-
Looks good, indeed.
So not all of these PSUs are created equal.
-
I don't see this overshoot...
Interesting.
Your unit reaches 5V in 100us.
My unit Ch3 reaches 5V in about 7500us (not including the overshoot).
My unit Ch1 and Ch2 reaches 5V in about the same time.
Yours is quite different.
There are several models of this power supply.
My model number is GPD-3303S, and it is brand new.
What is your exact model number?
How old is your unit?
-
I got this unit from eBay (apparently liquidated stock from ITT Tech schools...), so its probably at least couple years old as fan bearing was making noise in my unit when I got it. I have replaced fan on this unit (https://www.eevblog.com/forum/testgear/gw-instek-gpd-3303s-fan-replacement/ (https://www.eevblog.com/forum/testgear/gw-instek-gpd-3303s-fan-replacement/)), but that should have no effect on how CH3 behaves...
Only model number on the unit seems to be in front panel (just QC and serial number stickers on back panel). Have to try USB, if it would report actual model/revision/firmware over USB (?)
UPDATE:
No luck getting USB connection to work. Unit has FTDI USB to serial chip:
ID 0403:6001 Future Technology Devices International, Ltd FT232 USB-Serial (UART) IC
But USB interface seems to get stuck during power on (boot), since I see following over serial connection (9600bps) when powering on the PSU:
**********************************************
Program utility! V1.06
**********************************************
Do you update programme?
But cant seem get it to respond any SCPI commands at all...
-
sequoia:
Your front panel is different from mine. Notice position of banana jacks and LEDs; see photo.
Otherwise same model number on front and same boot screen.
Maybe they changed something inside too.
-
sequoia:
Your front panel is different from mine. Notice position of banana jacks and LEDs; see photo.
Otherwise same model number on front and same boot screen.
Maybe they changed something inside too.
Seems like they've changed design (but not model number), as front panel is clearly different. Most obvious being third channel labeled "FIXED" vs. "CH3".
Assuming units with "CH3" label are newer patch/design, seems as if they must have cut costs or something since they seem to perform significantly worse...
-
sequoia and Gary350z :
Would you guys be able to post photographs of the internals so that we can possibly figure out what is different between the two variants?
-
sequoia and Gary350z :
Would you guys be able to post photographs of the internals so that we can possibly figure out what is different between the two variants?
That would be good to know, but mine is being returned to seller so I can't do that. :(
Plus you would probably have to remove the main PCB and compare them part by part. Removing the main PCB looks complicated.
-
i have bought ain ITT unit too, the front panel arrangement is the same as sequoia's and as sequoia i didn't measure overshoots or slow rise time on the additional channel (but i used a lighter load).
haven't tried the usb connection yet.
-
That's a very interesting development in the overshoot saga. I hadn't really paid that much attention to the front panel layout. It looks like the panel layout was changed to accommodate the additional channel for the four-channel GPD-4303S in order to share parts. Adding that fourth channel to the PCB design probably resulted in some changes to the existing channels, and introduction of new bugs.
Edit: Actually, Ch3 on the 4303 is adjustable and has two ranges, unlike on the 3303. So, it definitely changed. Ch3 on the 4303 has been reported to overshoot, too.
-
Hi folks - first post just joined. I just bought one of these power supplies off of eBay and it looks to be in good shape. Honestly I'm disheartened to see that they changed the unit and kept the model number the same... mine says CH3 on the fixed supply output so maybe it also has the overshoot problem.... I've yet to put it on a scope but I will. Do we know for sure that the circuit has changed or just the front panel? But of course at lest one person here claims that his doesn't have the overshoot problem.
By all means please post any mods that can solve this problem. I'm not against ripping into the unit and making changes if they actually work.
I downloaded the service manual from Instek but honestly how can I tell if it matches my unit or the other one?...
Thanks!
-
sequoia and Gary350z :
Would you guys be able to post photographs of the internals so that we can possibly figure out what is different between the two variants?
What the heck. I have both of the versions. I bought them from two different vendors on Ebay within a week of each other. I've got some time right now so I will go and take them apart and take the pics. Hopefully we will find something.
-
Just a real quick reaction to what i found. Pics will be coming. The Unit that is like Gary350z's I will call the non-gapped unit as it does not have a gap between the banana posts.
There is quite a difference in the board and component layouts. It has me scratching my head as to why. The main power boards are very similar but some of the parts are in different locations. Like one TO-220 regulator is soldered to the board and attached to the right hand heatsink. On the other unit, it is screwed to the upper part of the left hand heatsink and attached to the board with wires. ?? Almost looks like a afterthought kludge. Only thing is that the original location on the board is still there with no marks on the heatsink or other signs it was ever there.
The front panel pcb's are reversed from one to the other. one has components on the rear side of the PSU and the other has the component side to the front. The power transformers look the same. The wire bundles between the power board and the front panel boards are the same number, color and sizes.
-
Having not seen inside these, but looking at the front panel layout, the changes appear to have been made to accommodate the four-channel model, GPD-4303S, that looks very similar to the redesigned GPD-3303S. The channel 4 binding posts appear to the left of channel 2.
The CV/CC LEDs are in a better location, though. So, that was an improvement. But, apparently, in the process of making space for channel 4, channel 3 was borked in both the 3303S and 4303S.
Looking forward to seeing inside the two versions of the 3303S.
-
i have the pics now. I'm trying to get them from the camera to the forum. Gotta figure out what this forum needs. I can tell you that the PCB's on them are different in some ways but very similar in others. They also have different numbers on the PCB's
Also, there is a display board and then there is a control board behind that that is held on standoffs. The control board has the switches and knobs attached to it.
Non gapped unit.
Power switch pcb PD03SP03A-2
Power PCB PD03SP03A-1
Control Board PD03SP03A
Gapped unit:
Power switch pcb PD01P05C-2
Power PCB PD01P0SC-1
Control Board PD01P02C
Now to get these pics up.
-
I know my camera sucks but this what I have CLICK ON THE THUMBNAILS all the pics have captions.
(http://i.imgur.com/armyclet.jpg) (http://imgur.com/armycle)(http://i.imgur.com/pjj10VYt.jpg) (http://imgur.com/pjj10VY)
NON-GAPPED:
(http://i.imgur.com/AKWLh4Gt.jpg) (http://imgur.com/AKWLh4G)(http://i.imgur.com/ztg9qNet.jpg) (http://imgur.com/ztg9qNe)(http://i.imgur.com/NwhUdjvt.jpg) (http://imgur.com/NwhUdjv)(http://i.imgur.com/Uxc6GJLt.jpg) (http://imgur.com/Uxc6GJL)(http://i.imgur.com/NesONblt.jpg) (http://imgur.com/NesONbl)(http://i.imgur.com/AhL3Wmdt.jpg) (http://imgur.com/AhL3Wmd)(http://i.imgur.com/UAZjqxnt.jpg) (http://imgur.com/UAZjqxn)
GAPPED:
(http://i.imgur.com/ZRFzEZit.jpg?1) (http://imgur.com/ZRFzEZi)(http://i.imgur.com/wEaOihYt.jpg) (http://imgur.com/wEaOihY)(http://i.imgur.com/4MPUWgWt.jpg) (http://imgur.com/4MPUWgW)(http://i.imgur.com/4akUPIxt.jpg) (http://imgur.com/4akUPIx)(http://i.imgur.com/wBxu84Lt.jpg) (http://imgur.com/wBxu84L)(http://i.imgur.com/7V0FHcOt.jpg) (http://imgur.com/7V0FHcO)(http://i.imgur.com/PebPpSit.jpg) (http://imgur.com/PebPpSi)(http://i.imgur.com/Fo92ON3t.jpg) (http://imgur.com/Fo92ON3)(http://i.imgur.com/W0os8RRt.jpg) (http://imgur.com/W0os8RR)
-
Another difference. I was wrong on my first impression that the transformers were the same. The non gapped one has dual orange wires coming out and they are terminated in an unused connector. The other unit does not have them at all. I would bet that those are the power taps for the 4th channel
-
I have found the correct manuals for the two different units. The gapped one is from 2008 to 2011 and is the older unit.
The non gapped one is from 2011 to current. The manuals for this are available on the Instek site.
Gapped early unit user manual and product announcement from 2008:
https://cdn.testequity.com/documents/pdf/GPD-3303-Manual.pdf (https://cdn.testequity.com/documents/pdf/GPD-3303-Manual.pdf)
https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&ved=0ahUKEwiB0qmV_azSAhXDLSYKHc0oDuAQFggcMAA&url=http%3A%2F%2Fwww.testequipmentdepot.com%2Finstek%2Fpdf%2Fgpd3303s.pdf&usg=AFQjCNE1pTkiXEhctUofMJ1p2YHkSWIjKw&sig2=JF4AZFy1pn3Mx6XIkRqPXA&cad=rja (https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&ved=0ahUKEwiB0qmV_azSAhXDLSYKHc0oDuAQFggcMAA&url=http%3A%2F%2Fwww.testequipmentdepot.com%2Finstek%2Fpdf%2Fgpd3303s.pdf&usg=AFQjCNE1pTkiXEhctUofMJ1p2YHkSWIjKw&sig2=JF4AZFy1pn3Mx6XIkRqPXA&cad=rja)
-
The service manual for the later non gapped versions is available online.
https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&ved=0ahUKEwjf97DQka3SAhVIKiYKHTHOADgQFggfMAA&url=http%3A%2F%2Fwww.eltest.hu%2Fpdf%2FGPD3303SM.pdf&usg=AFQjCNHYdRXPJymjeCj4AK32pdx1fQgC2A&sig2=D5fjyTfqlNggepUdlhdNdQ&cad=rja (https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&ved=0ahUKEwjf97DQka3SAhVIKiYKHTHOADgQFggfMAA&url=http%3A%2F%2Fwww.eltest.hu%2Fpdf%2FGPD3303SM.pdf&usg=AFQjCNHYdRXPJymjeCj4AK32pdx1fQgC2A&sig2=D5fjyTfqlNggepUdlhdNdQ&cad=rja)
-
Thanks for all the photos! I bought one off of eBay recently and it appears to be the later model, and has the ~7v overshoot on the fixed supply channel 3 as reported by others. Now - I'm more than willing to make a trial mod. The first thing I'd like to know is if the resistor R712 referenced by both @Kleinstein and @henken is indeed the one found in the service manual pointed out here. Those who tried this mod - Can we please get a schematic reference to go with the R712 component number? Or a photo of the location on the board, if you took any when doing it?
Thanks!
-
Thanks for all the photos! I bought one off of eBay recently and it appears to be the later model, and has the ~7v overshoot on the fixed supply channel 3 as reported by others. Now - I'm more than willing to make a trial mod. The first thing I'd like to know is if the resistor R712 referenced by both @Kleinstein and @henken is indeed the one found in the service manual pointed out here. Those who tried this mod - Can we please get a schematic reference to go with the R712 component number? Or a photo of the location on the board, if you took any when doing it?
Thanks!
You are welcome. The schematics that pertain to the later model are in the Service manual that I linked to in the previous post. It also includes complete part numbers and specs in the charts below the schematics.
-
Thanks for all the photos! I bought one off of eBay recently and it appears to be the later model, and has the ~7v overshoot on the fixed supply channel 3 as reported by others. Now - I'm more than willing to make a trial mod. The first thing I'd like to know is if the resistor R712 referenced by both @Kleinstein and @henken is indeed the one found in the service manual pointed out here. Those who tried this mod - Can we please get a schematic reference to go with the R712 component number? Or a photo of the location on the board, if you took any when doing it?
Thanks!
You are welcome. The schematics that pertain to the later model are in the Service manual that I linked to in the previous post. It also includes complete part numbers and specs in the charts below the schematics.
Right and thank you for that! I'm asking for confirmation of the location of the proposed fix/improvement because I don't fully understand why it's chosen to be there, on that specific resistor in the schematic. While I understand the filtering idea I don't understand that location relative to that schematic. That's why I'm asking for a few more details.
-
I understand. I'll try to find it and get a pic.
As I look at the photos I took and compare them to the Siglent unit from the other thread, I'll be danged if they don't come extremely close to the same. they launched their psu within a year of Instek. I think they bought the innards from Instek and added their own intelligent display board and front panel.
All the points on the control board that they mention to add parts to in the Siglent thread are there on the instek. Heck it looks like the control boards are better than 95% the same layout and parts.
I gotta find a schematic of the siglent. That will tell the story for sure.
-
Apparently Siglent SPD3303() power supplies also have / used to have this problem, and some fixes are discussed here:
https://www.eevblog.com/forum/testgear/siglent-spd3303d-review/25/ (https://www.eevblog.com/forum/testgear/siglent-spd3303d-review/25/)
Depending on how close the designs of these supplies are, it could be useful to read that thread.
Yes, thank you for finding that thread!!
I decided that I'm going to pull the control boards from both units and see if I can figure out the differences between the old and new and see just how close they seem to the siglent.
-
Thanks for all the photos. They pretty much confirm that the changes were to support the new four-channel model.
Looking forward to your findings on the comparison with the Siglent version. Hopefully, it really can be fixed without too much trouble.
-
Ok. looking closely at the cad drawings from the Siglent thread, you can see That pin 1 of UC3 goes to RC5 and RC9. RC5 then goes to pin 2 of Ic4. UC3 and UC4 are TLO72 dual op amps. This aligns perfectly with The Instek schematics. Look on page 83 of the Instek service manual and you see U704A and U704B. These are near the exact same position as the siglent board and the rest of the traces shown in the x ray view of the siglent line up perfectly with the schematics in the instek service manual. RC5 on the Siglent thread corresponds to R715 (1.5K) on the Instek.
It seems more and more that Siglent had the schematics and had the board updated and went more SMT. Way too much is the same to be coincidence.
A telling thing is that after installing the transistor and the other components, the OP in the Siglent thread stated that he still had the overshoot. It was resolved only after he changed RC5 to a higher resistance, Changing the voltage going into the other op amp. In other words, they are changing the bias input of the second OP amp. I wonder if this is actually a saturation problem of the second op amp.
Siglent UC3 = Instek U704A
Siglent UC4 - Instek U704B
Well, I finally figured out that my Nikon pocket camera has a macro mode. I have PSU parts all over my table, and I hope to get more comparisons and pics out sometime soon.
I also want to figure out on the schematics what they were trying to do with the other components
-
Due to a request by PM, here are some explanations to proposed modification with an additional RC in parallel to R712 (for channel 3, circuit from linked service manual):
The circuit for channel 3 is a basic low drop regulator with a floating regulator circuit. However it only has basic compensation with a single dominant pole. This known to be problematic, not only for overshoot, but also for ringing in case of a highly capacitive load. I did a simulation of a slightly simplified circuit to reproduce the problem. Adding a RC combination in parallel to R712 is somewhat standard in similar circuits. It is similar to the function of C110 for channel 1, though the circuit is slightly different. One can see this a little like going from an PI regulator loop to an PID loop. It adds response to the rate of change in output voltage and is this way reducing overshoot. It has two purposes:
1) It adds some more phase reserve in the 100 Hz-1 kHz range and this way should reduce possible problems with a capacitive load. This showed up in simulation - no report of this from real life, as highly capacitive load with low ESR is not that common.
2) The point more related to the overshoot is that this change speeds up the control loop. The original loop is rather slow and thus needs relatively long to recover from saturation. The OP still goes deep into saturation, but with a faster loop there is less time to charge the output capacitor and thus a lower overshoot.
So the modification is not a cure to the root cause of the overshoot. It is more like reducing the symptoms with few added parts. Reducing the value of R715 as part of the Sigilent mod is also making the loop faster, though without the added advantage to the capacitive load case, and less effective on recovery.
Speeding up the loop could in principle cause instability at higher frequencies - but the loop is still not super fast and the simulation was still OK. Also to one actual change seems to work OK.
Excerpt from circuit:
-
Thanks @Kleinstein! Really appreciate the extra explanations even if I have to look up a handful of terms new to me LOL. Always an opportunity to learn something. No matter what I will probably try this mod and see if it helps in my case.
-
I realize that this topic has been dormant for a while. However, I am wondering whether the fix for the Siglent version (SPD-3303S) is the same as the Instek (GPD-3303S)?
My understanding is that Instek did fix the problem on the current systems but I could not find a more recent schematic for the Instek (GPD-3303S) to compare against my Instek unit. Maybe I am incorrect in this understanding.
Does anyone have any further knowledge to share regarding the fix to the CH3 overshoot?
Thanks
Jim
-
Thanks for bringing this up again. I'll be following with interest.
I realize that this topic has been dormant for a while. However, I am wondering whether the fix for the Siglent version (SPD-3303S) is the same as the Instek (GPD-3303S)?
My understanding is that Instek did fix the problem on the current systems but I could not find a more recent schematic for the Instek (GPD-3303S) to compare against my Instek unit. Maybe I am incorrect in this understanding.
Does anyone have any further knowledge to share regarding the fix to the CH3 overshoot?
Thanks
Jim
-
However, I am wondering whether the fix for the Siglent version (SPD-3303S) is the same as the Instek (GPD-3303S)?
My understanding is that Instek did fix the problem on the current systems but I could not find a more recent schematic for the Instek (GPD-3303S) to compare against my Instek unit.
I don't know anything about the Siglent side of things, but for GW-Instek I think the opposite of your statement is true. I believe that old Instek GPD-3303S units are OK, and newer ones built after the release of the GPD-4303S will have the CH3 overshoot problems. If you read through the second page of this thread there is some discussion on the different versions out there.
I haven't looked into this since all the activity back in February so I suppose it is possible that GW-Instek have cleaned things up in the meantime, but for some reason I highly doubt it.
-
I followed the post: https://www.eevblog.com/forum/testgear/siglent-spd3303d-review/msg299246/#msg299246 (https://www.eevblog.com/forum/testgear/siglent-spd3303d-review/msg299246/#msg299246) Reply 27 where Siglent addressed the issue and contributed a hw patch to those ill siglent power supplies.
Based on the above thread, the designs are identical or near identical between the Instek and Siglent, although I don't know how that came to be. Was this stolen or purchased from one another. In any event I found a copy of the Instek Service manual online and the schematic contains about 90% of the fixes that are recommended in the Siglent post. The Instek service manual has no dates or revision information so I am confused whether it applies to the newer units or not. I need to attempt to compare my service manual schematic against my hardware which I haven't yet done.
I have sent a request to Instek technical support today regarding this issue. So, we will see what they have to say, if anything.
So, to reiterate what I believe to be true(please correct me if incorrect):
There are two designs of the GDP-3303S power supplies
#1 - Early model - only 3 channel - 2 adj (Ch1/Ch2) - 1 (2.5/3.3/5.0)(Fixed) - This version had no issues with overshoot on the Fixed channel.
#2 - Later model - Redesigned support two versions:
4Ch version with Ch3/4 5V adjustable
3Ch version same as #1 above (only with overshoot issues)
rfdes
-
Also on the later three-channel model, all binding posts are evenly spaced (part of the redesign to support the new four-channel model).
-
I heard back from the Instek Service manager in California in regards to any suggested mods, etc. to resolve the overshoot issue. Their recommendation was to piggyback a 100k resistor across R727. This was discussed earlier by several members if I remember.
This did fix the output enable spike but obviously does nothing for the spike incurred following a current clamp release. I then performed a spice simulation to test out the recommended fixes that Siglent had recommended to their customers.
https://www.eevblog.com/forum/testgear/siglent-spd3303d-review/msg299246/#msg299246 (https://www.eevblog.com/forum/testgear/siglent-spd3303d-review/msg299246/#msg299246)
The results were not promising in my simulation model. There was significant instability that I encountered. I believe the prevailing belief is that the Siglent and Instek models are nearly the same. The circuit topolgies appear the same but I cannot confirm the parts values.
Does anyone have a schematic for the Siglent power supply just to see the differences? The fellow that posted the Siglent fix claims it worked for both his output enable and current short release spikes.
Any further help is appreciated.
Jim
-
The additional 100 K parallel to R727 adds some offset to the set voltage and thus makes the added transistor for the output enable effective. It might work without, depending on the OPs offset.
The other suggested changes (a 20 K resistor in series to C717 (in the Instek schematics)) or an RC (e.g. 1.5 K 50 nF) combination in parallel to R712 have a rather similar effect. They kind of speed up the response and thus reduce the peaks, but they don't help against the root problem.
The root problem is integral windup of the voltage loop (U705 A). One could in principle add some windup limiting to this. One scheme uses a PNP transistor with the emitter at the output of U705A, a diode from the inverting input to the collector and a high value resistor (e.g. 100-1M K range) from the base to the output of U705 B or the anode side of D705/D704 (the diode side works with output enable too, the OPs side works better when recovering from current limit). However there is a slight chance for instability, especially if the base resistor is too small. At least in the simulation it works reasonably. A more accurate integrator limiting as difficult to add to this circuit as the input side voltage can be higher than the output side.
-
I heard back from the Instek Service manager in California in regards to any suggested mods, etc. to resolve the overshoot issue. Their recommendation was to piggyback a 100k resistor across R727. This was discussed earlier by several members if I remember.
I've tried adding a 100k resistor parallel to R727 but it didn't remove the overshoot. The only effect I saw was that the 5V output voltage was lowered to 4.9 V.
-
The added resistor in parallel to R727 only helps, with the mod / version that turn off the ref. voltage, when in output off mode. With a BJT used to turn off the set voltage, some offset is needed and the original 5 mV + OPs offset tend to be not enough. By itself it only adds more offset - thus the change in voltage.
-
Yes, the R727 fix will only work if the CE saturation voltage on Q706 and Q707 is lower than the offset voltage. I also tried it and it didn't work for me.
Measuring the voltage at U705 pin 3 (pos inp) I was measuring around 30mV due to the sat voltage.. I changed the Q706/Q707 clamping bipolars to N chan mosfets. The 2N7002 is a drop in pinout. Now, I only see 1-2 mV on U705 pin 3. The output of U705-pin 1 now will clamp at about -800mV. The transient spike is now completely gone. You will need to remove the front panel to do this but it was not bad at all.
- Unplug the one cable at the top of the front panel.
- Remove the ground quick connect attached to the bottom plate.
- Remove the top support bracket.
- Remove the two bottom screws holding the front panel.
- The front pane can now be swung back to gain access to the two transistors.
Give this a try and the Output enable transient will be gone. However, this does nothing for the output short - release transient.
Jim
-
I wanted to present my final fix for the GPD-3303S. The fixes were recommended by several members and I simulated the fixes with Tina-Ti spice, made my modifications and am now quite happy with the results.
In a nutshell:
Replace Q706/Q707 with N-Chan Mosfet 2N7002. These are footprint compatible and will drop right in.
Replace R727 with a 330K resistor.
Replace R717 with a 220 ohm resistor
Remove C717 and 'teepee' a 22K resistor in series,
The photos show the before and after transient responses using these modifications.
The changes were not difficult with the replacement of Q706 being the most challenging.
Good luck to others that decide to make the modifications.
Jim
rfdes
-
Looks really good. I'm going to try it on mine first chance I get.
Thank you so much!
Avi
-
I have done some testing on channel 3 with my 4 channel GPD-4303S power supply. I did nine measurements and wil post a screen of each measurment with a short decsription. Hope this info is usefull. I think the results are good, no overshoot found.
This is my unit :
(http://members.casema.nl/r.kamp/pictures/electronics/gpd-4303s.jpg)
(http://members.casema.nl/r.kamp/pictures/electronics/ch3_s00001.jpg)
GDP-4303S, CH3, Uset 3.3V, no load, switched from off to on
(http://members.casema.nl/r.kamp/pictures/electronics/ch3_s00002.jpg)
GDP-4303S, CH3, Uset 5.0V, no load, switched from off to on
(http://members.casema.nl/r.kamp/pictures/electronics/ch3_s00003.jpg)
GDP-4303S, CH3, Uset 9.0V, no load, switched from off to on
(http://members.casema.nl/r.kamp/pictures/electronics/ch3_s00004.jpg)
GDP-4303S, CH3, Uset 3.3V, 8.2 ohm load, CV, switched from off to on
(http://members.casema.nl/r.kamp/pictures/electronics/ch3_s00005.jpg)
GDP-4303S, CH3, Uset 5.0V, 8.2 ohm load, CV, switched from off to on
(http://members.casema.nl/r.kamp/pictures/electronics/ch3_s00006.jpg)
GDP-4303S, CH3, Uset 9.0V, 8.2 ohm load, CV, switched from off to on
(http://members.casema.nl/r.kamp/pictures/electronics/ch3_s00007.jpg)
GDP-4303S, CH3, Uset 3.3V, Iset 0.3A, 8.2 ohm load, CC, switched from off to on
(http://members.casema.nl/r.kamp/pictures/electronics/ch3_s00008.jpg)
GDP-4303S, CH3, Uset 5.0V, Iset 0.3A, 8.2 ohm load, CC, switched from off to on
(http://members.casema.nl/r.kamp/pictures/electronics/ch3_s00009.jpg)
GDP-4303S, CH3, Uset 9.0V, Iset 0.3A, 8.2 ohm load, CC, switched from off to on
-
I wanted to present my final fix for the GPD-3303S. The fixes were recommended by several members and I simulated the fixes with Tina-Ti spice, made my modifications and am now quite happy with the results.
In a nutshell:
Replace Q706/Q707 with N-Chan Mosfet 2N7002. These are footprint compatible and will drop right in.
Replace R727 with a 330K resistor.
Replace R717 with a 220 ohm resistor
Remove C717 and 'teepee' a 22K resistor in series,
The photos show the before and after transient responses using these modifications.
The changes were not difficult with the replacement of Q706 being the most challenging.
Good luck to others that decide to make the modifications.
Jim
rfdes
The FETs are SOT-23 and all other parts are 0603. They are on the control PCB in the following locations:
(https://i.imgur.com/eapocHf.jpg)
-
Thanks djnz! I'm planning on making this mod in the next few months.
-
I've owned the older version of this GPD-3303S supply that doesn't have the overshoot problem and recently got a pretty good deal on the newer version with the problem. The overshoot on CH3 was up to 5.8V and on the 2.5V setting that could be disastrous. I just finished the modification and it works. CH3 turns on slower, in milliseconds instead of microseconds, and there is no overshoot. Here is what the output looks like on the scope with no load.