Hcc15-3-ag power supply odd behavior

[deepblackjoe@gmail.com]

So I'm putting together my first busbar system. I got a new power one, tapped out the bars, and wired it up after much reading, and everything looked good,. Hooked up a few modules, cycle power a few times, add a few more. Not getting hot, add a few more. I got up to about 3 racks, and then flip the switch and the LED's are just very dim, so I turn it off, all the LED's flash once a second later. Turn it back on, fine again. So I started trying different combinations and trying to find one that was a problem, and up to about 15 or 20 modules, the power comes on about one in three times, without changing anything at all, just on/off. I checked the voltages on the rails when it latches up, and get -.07 on the + rail, and -15 on the neg, like it should be. What could cause this to happen? The modules are mostly blacet and diy cgs, all of which worked fine before this on my patchwork of busboards setup. I'm using an enclosed iec inlet, mains earth next to it goes to 0v bar, at the point of distribution, and to the frame of the brick. Sometimes it will turn on fine 10 times in a row until I add enough modules, but it should be able to do this. I guess what I really don't get is why it would work like every other time when nothing is changing? Would love to hear some insight. I'll try to post some photos.

[deepblackjoe@gmail.com]

[deepblackjoe@gmail.com]

I thought of something I need to add. At one point I had all these but one plugged in, and it turned on fine. Once or twice. I thought it would maybe help to triple up the wires from the posts to the busbars, and connect them to the middle of the bars instead of the end, but get the same behavior.

[EATyourGUITAR]

maybe you should install current sense resistors so you can meter your current consumption by probing with a multi meter set to volts.

https://www.ebay.com/itm/Shunt-Resistor ... 1709456409

using an oscilloscope you can watch your current spikes. then using the oscilloscope again probing the output voltage with reference to ground you can see the rise time after the circuit is energized. these two tests will tell you if you have too much in-rush current, too much load, or too much capacitance. when you get 0.7v it is doing one of 3 things most likely. the voltage can overshoot before it overcompensates and latches to a mode that shunts maximum current or it just can't supply the total current that the load draws while maintaining +15v. or it oscillates and gets latched.

power cycling with strange results %75 of the time to me would indicate either maybe thermal problems or perhaps you are getting lucky as the caps are discharging you are turning it on with the right amount of charge in the caps to null the in-rush current just enough to not trigger the over current protection that is built in to the HCC-15-3-AG

more testing is required. you can also setup relays to stage every row turning on after a 3 second delay.

[deepblackjoe@gmail.com]

One thing is the inlet is at the bottom and the supply at the top. I used 18 awg because I had it, not thinking it mattered. It's almost three feet long. Should I fatten this up to 14? I also didn't use those star washers under the ring to cut thru the oxide layer on the bar.. is this essential? I did use antiox paste tho and lock washers. Maybe shortening the wires as much as possible and making them all 18 awg would also help minimize the effect of the inrush? Thanks for your quick reply... I can look into these tests you mention. I have another older hcc15-3a that powered these modules easily. There's actually two of these cases each with own busbars, to use side by side. I have yet to load the other one up. What is the difference between 3a and 3ag anyway?

[Graham Hinton]

One thing is the inlet is at the bottom and the supply at the top. I used 18 awg because I had it, not thinking it mattered. It's almost three feet long. Should I fatten this up to 14?

It doesn't matter. A tiny voltage drop on mains at mA is not going to effect anything.

I also didn't use those star washers under the ring to cut thru the oxide layer on the bar.. is this essential?

No. Shakeproof washers are to prevent the screw coming loose. Put them in contact with the screw head above the terminal. They might cut through the oxide initially, but they won't prevent re-oxidisation. Use jointing compound intended for the purpose.

Maybe shortening the wires as much as possible and making them all 18 awg would also help minimize the effect of the inrush?

No. It won't do anything for that, you can't reduce current by reducing resistance anyway. The purpose of heavier gauge wire is to reduce voltage drops in places where there is common impedance coupling. That is between PSU and distribution (if you are not using sensing) and in the 0V wires to each module.

There's actually two of these cases each with own busbars, to use side by side.

Then you need a good connection between each 0V busbar. Several large bolts through the sides will do.

You are simply overloading the PSU you have and then grabbing at straws that are nothing to do with your problem. Read the manual with particular reference to the word "derate" in the small print. Those PSUs are intended to be mounted on larger metalwork and/or used with fans at higher outputs.
Have you any idea what currents your modules need?

[deepblackjoe@gmail.com]

Well I couldn't find the derate detail in the power one lit, not yet. I tried to figure out the current draw for the five 19" racks hooked up in the one case so far and it's about 1.4a, knowing about 2/3 of them, and guestimating the rest. There's one more I was hoping to add, case being six high, in theory it should be able to do it. It seems the inrush current is more of a thing than I imagined. Could my former chain of busboards have created a pillowtype buffer of resistance to this before? I looked at NTC thermistor info and that seems to be what they would be doing, not sure how to calculate what type I would want, or if it would even be helpful. It will be interesting to see if the other case with the same busbar setup will behave different with the hcc15-3-a in that one. If it does maybe I could put more heavy draw modules on that side (?). They have yet to be linked together. I will connect the commons well, and earth in one place. Also I ended up changing the wires between the switch and the supply to 14awg, and now it turns on about every other time. When I leave it on for awhile, the still exposed regulators don't really get hot, which makes me think the inrush is the problem. Thanks for your help, always enlightening

[Graham Hinton]

Could my former chain of busboards have created a pillowtype buffer of resistance to this before?

Not enough to stop this.

I looked at NTC thermistor info and that seems to be what they would be doing, not sure how to calculate what type I would want, or if it would even be helpful.

The correct place to limit a power up surge is on the mains side of the power supplies. Any resistance you add to the outputs will defeat all the benefits of having busbars.

I will connect the commons well, and earth in one place.

Earth each case separately so that it is usable standalone, then connect the 0Vs together. You are making a triangle of resistances, your two mains leads are about 150 milliohms each so you need the link to be about 1 milliohm or less. I have used M10 bolts through case sides before. Also busbars bolted directly between busbars (not possible with your layout) and 25mm2 cables.

[deepblackjoe@gmail.com]

If I am using the cases together, which I imagine will be most always, would I then earth only in one side or the other? It would be easy enough to do both if preferable. I've fitted a 4 pin XLR jack next to each inlet , with all 4 pins connected to the common bar with 18 awg just a couple inches long, with a cable also of 4 18's to link them. Maybe not lowest of the low mohms, but not a banana lead. It was the best I could come up with for now, with a mind to upgrade in the future if necessary. The current limiting I was just reading about, the NTC thermistors, would go between+15 and -15 posts on the supply and the busbars, slowing down the rush, then heating up and becoming non resistive, not changing the current draw but effecting the rate. The also mentioned a fixed resistor to limit this rate, but will allow the full current demand to flow unimpeded after dampening this first surge. If there was none in the common path and only the rails, could this help me? Not sure how to calculate what size. Is it possible for one side of the supply to clamp immediately like this because one side is drawing much more or faster, like a safety mechanism of some kind? The fact that it's not hot when it does power up makes me wonder

[EATyourGUITAR]

It is possible to get latchup when you are not overloading your power supply. That power supply has over current protection I know for a fact. Neither you nor grahm have actually measured the current draw of the modules. This is fine for an expert like Graham who is pretty good at guessing the problem. I continue to disagree with him because I was trained to take measurements and follow procedures. I would never teach someone electronics as the science of speculation. Get the current sense resistors. That way you will always know you current draw as you add modules to the system. With an oscilloscope you can see exactly how much in rush current trips the over current protection. Real numbers. Real answers. Less back and forth. All these power supply threads become safety ground threads which doesn't fix the problem of capacitance from the modules accidentally tripping over current protection. It is just a side bar. While it is good information, it doesn't really take the problem head on.

[Rex Coil 7]

Well I couldn't find the derate detail in the power one lit, not yet.

Lacking credible specifications I tend to go with about a 60% duty cycle when spec'ing out a linear power supply. Not that ~linear~ is somehow less reputable when it comes to published specs (no more incorrect than specs on switching supplies) but rather without credible duty cycle charts I go with a "soft 60%" so to speak. Never hurts to have "too much" capacity, and always hurts to have too little capacity. 40% headroom is something I'm comfortable with.

Note the emphasis on the word "credible". Someone a lot smarter than I once said "there are lies, there are damned lies, and then there are manufacturers' specifications". Wise words.

[Rex Coil 7]

... This is fine for an expert like Graham who is pretty good at guessing the problem. I continue to disagree with him because I was trained to take measurements and follow procedures. I would never teach someone electronics as the science of speculation. Get the current sense resistors. That way you will always know you current draw as you add modules to the system. With an oscilloscope you can see exactly how much in rush current trips the over current protection. Real numbers. Real answers. Less back and forth. All these power supply threads become safety ground threads which doesn't fix the problem of capacitance from the modules accidentally tripping over current protection. It is just a side bar. While it is good information, it doesn't really take the problem head on.

In motorcycle racing what you just did there is called "avoidable contact". Riders are penalized for doing that. For good reason.

I see no reason to take a swing at Hinton ... you'll retain a lot more credibility if you simply address the issue rather than trolling Hinton.

[EATyourGUITAR]

He takes shots at me all the time. I toned it down. I don't see the problem. There was no sarcasm there. Maybe you read it as sarcasm. It is just giving context. An expert can guess the problem. A student needs ohms law and a multi meter. I see an expert giving expert advise. I am talking about it in the most respectful way giving context and explaining why I respectfully disagree with Mr Hinton. I say this all the time in my posts. Hinton is an expert. Hinton is right. I have a lot of respect for Mr Hinton honestly. There is no hate in my post at all. I have recommended his power supplies to people. I repeat his words when people ask me about power supplies. I already apologized for the fights we had in the past and I still stand by my statement that I don't want to fight anymore.

[Graham Hinton]

If I am using the cases together, which I imagine will be most always, would I then earth only in one side or the other? It would be easy enough to do both if preferable.

Earth both, like I just said. Don't say that will make a ground loop. It won't if you follow what I said.

I've fitted a 4 pin XLR jack next to each inlet , with all 4 pins connected to the common bar with 18 awg just a couple inches long, with a cable also of 4 18's to link them. Maybe not lowest of the low mohms, but not a banana lead. It was the best I could come up with for now, with a mind to upgrade in the future if necessary.

I would use a DINSE-25 socket with 25mm2 cable and then only if the cases needed to be used live.

You would be much better off making the second case mirror image and bolting through the 0V bars. Unless you don't want the cases side by side.

The current limiting I was just reading about, the NTC thermistors, would go between+15 and -15 posts on the supply and the busbars, slowing down the rush, then heating up and becoming non resistive, not changing the current draw but effecting the rate. The also mentioned a fixed resistor to limit this rate, but will allow the full current demand to flow unimpeded after dampening this first surge. If there was none in the common path and only the rails, could this help me?

The type of NTCs used for inrush limiting have a resistance of about 10 to 20 ohms at 25 degC dropping to 0.1 ohms at 100 degC. Putting that in the most sensitive place for common impedance coupling where you really want <5milliohms would be a disaster. Even 0.1 or 0.01 ohm current measuring resistors would compromise your system.

The easiest way to work out the current surge is to get one of those cheap power monitors that plug between an outlet and a mains lead. They have maximum power memories and you can compare that with the steady power. Whatever power your PSU takes from the mains, divide by two for the linear PSU efficiency and then by 15 to get the sum of the currents in both rails. Unless you happen to have an expensive Tektronix current probe laying around...

I'm wondering why you have a power inrush problem at all. Have you got a lot of modules with a ferrite plus electrolytic at the power input? The correct place to limit inrush is there and a ferrite is just a piece of wire at low frequencies so you need some resistance in series.

[deepblackjoe@gmail.com]

Uh oh I do have ferrites and electrolytics on just about every module! Habit I picked up from cgs builds, never thought it would come back to bite me. Maybe I should switch most of those out with 20r resistors, the more accessible ones, and see if that helps, probably don't need most of the ferrites anyway. I definitely wish I had made the cabs mirror image and joined thru the sides with a bar - just didn't think it thru. That upgrade would mean almost starting over at this point.

[Graham Hinton]

Habit I picked up from cgs builds, never thought it would come back to bite me.

I've always wondered why people do that and don't work out the time constant.

I definitely wish I had made the cabs mirror image and joined thru the sides with a bar - just didn't think it thru.

If the cases are going to be permanently sited, get a really large battery cable made and connect it directly between the 0V bars by the shortest route. I mean battery as in articulated truck, not mobile phone. Forget the XLR.

Alternatively, have two shorter such cables, just long enough to get outside the cases and do the longer distance with another busbar.

The way to think about this is: how much resistance is there between the 0Vs of any two modules and could it be lower?

[Synthbuilder]

One thing you may wish to try is adding some Schottky power diodes, such as the 1N5819, across the power rails. Put one across the +15V and 0V bus bars, and one across the -15V and 0V bus bars. Fit them so that they are reversed biased in normal operation. So for the diode connected between the +15V and 0V, the cathode should go to the +15V. For the diode connected between the -15V and 0V, the cathode should go to 0V.

I have found that these Power One supplies really don't like power on surges caused by overly large electrolytics on the power inlets of modules. The imbalance between the surge on the positive rail and the negative rail causes the PSU to latch up. The latch up seems to occur when, say, the negative rail goes briefly positive on power up. The Schottky diodes do not prevent this completely but they do seem to stop it exceeding the voltage that causes the PSU to latch up.

The other option is to make sure each module has no more than a 4.7uF capacitor per rail at the power header. Sometimes I see 100uF in that location which should not be necessary and can cause problems.

Tony

[deepblackjoe@gmail.com]

This is more in the vein of what I suspect is happening. I think all the modules I've built have 10uf there, which is most of them. I'd much rather try this than squeeze another small supply/busboards in there, starts defeating the purpose of these new cases. I have some of those on hand, so will try today. And probably swap out some of the ferrites for small resistors as magic smoke retardant. Thanx for the hot tip. I remember reading somewhere about this imbalance latchup but not the details

[deepblackjoe@gmail.com]

Well the schottkeys don't do it. I had 5817 not 19. I ended up putting an extra small supply in with the 3ag to handle what the 3ag couldn't. I filled up the second case with my older 3a supply and it seems to handle more without latching up, which begs the question...
I'm going to go back and change out some of the ferrites for 10r resistors where it seems ok and put some of the more lopsided current draw modules on the 3a side and see if I can get this down to two supplies, but let it be for now. Thanks everyone. I left the schottkeys in because they're pretty buried at this point, but will remove if it's better for whatever reason, still got alot to learn

[Synthbuilder]

Well the schottkeys don't do it.

Oh well, it was a worth a try. They'll not do any harm so you can leave them in.

The Power One's also have what is called foldback current limiting. This is when it detects an overcurrent state and then reduces the maximum amount of current available to well below the original overcurrent value. This is not uncommon in PSUs and will help the devices overheating when any short is present on the outputs for a long time. The problem I think is that the Power One's foldback is too severe. It detects an overcurrent on start up (because of all those capacitors), reduces the current limit accordingly, and then, because the amount of current taken by the modular exceeds this new current limit, it never switches back on.

You can disable the foldback current circuit and convert it to standard current limit, but to do so, will require the circuit diagram of the power supply and knowledge of how these things work. And you'll have to mount your PSU on a decent metal plate as it will get a lot hotter that it used to do when a genuine short happens on the output.

Alternatively it is possible to add a resistor-capacitor combination to the current limit circuit which slows the normally fast acting foldback circuit, but doesn't disable it for large steady state current overloads. But again this requires the circuit diagram of the power supply and the knowledge to hack it safely.

Another option told to me by a Power One applications engineer (but not really recommended) is to slightly turn up the current limit trimmers on the PSU board. This is not really a good idea since the current limit may end up being set so high it doesn't trip early enough to protect the power supply. So this only needs to be turned up as much as it needs to switch on reliably.

Personally, I'd look into those power entry capacitors. If that doesn't work, look to using two smaller power supplies. Or grab one of Graham's power supply solutions.

[KSS]

Sequencing power on is a known and useful way to help with inrush issues.

CGS even has a PCB designed to do so. Never used, so my comment isn't about its goodness or badness.

Only that power-on sequencing can be useful.

[deepblackjoe@gmail.com]

Well I have finished reassembling the two cases, with the 3a in one and 3ag in the other, with a blacet ps800 handling about a third of the modules in with 3ag, the hungriest ones. The 3ag still has some false starts, but maybe I'll just live with it for now. I have yet to connect any of their commons, but it all sounds real good, less crosstalk than before and just clearer.
The only thing out of the ordinary that happened is with the Haible SOST. With just one in and one out, messing around for about five minutes, it was sounding great and then a sound crept in like wind blowing on a microphone over it and stayed, louder than what's supposed to be there. All functions are intact behind this garbled staticky whoosh, which has remained with subsequent power ups. Could not having the commons connected have pushed some chip into constant oscillation? Maybe some trimming would help but I don't get it. I wasn't touching anything when this happened. When I unplug the input, the sound decays away like the TDA's are ok. Hmm. Everything else works perfectly, so maybe it's not related to this new power scheme.

[Graham Hinton]

I have yet to connect any of their commons, but it all sounds real good, less crosstalk than before and just clearer.

Good. You need to establish a good common though.

it was sounding great and then a sound crept in like wind blowing on a microphone over it and stayed

That is usually the sign that an op amp has failed.

Could not having the commons connected have pushed some chip into constant oscillation?

No, not having compensation capacitors does that.
Have you looked at the output with a scope?

[deepblackjoe@gmail.com]

Just had the headphones on for about an hour, and have been unable to recreate my troubles. Must have been some kind of user error, something with the stepped LFO and the different invert stages. Thanks for the ideas, it led me to some good study. And the busbars idea altogether!