Need help on Delta 1010 PSU mods

[mtl777]

Can you put insulating kits on the regulators ?
MULTICOMP|MK3306|INSULATING KIT, MICA TO-220 | Farnell United Kingdom

Great idea, thanks! 🙂 BTW, is this a common problem? The power supply of the 1010 is such a bad design!

Enough of that. Let's go back to the capacitor mods now. I mentioned that I added 0.1uF film on the output of all the regs. However, there is already a 1uF ceramic (SMD) on their outputs (C16, C21, C25, C62). Is it bad to mix film and ceramic in this case (like, maybe it might cause some bad resonances or ringing)? Is adding the 0.1uF films too much already? 😀

 

[Mooly]

Wouldn't say it was a common problem tbh, not unknown though.

If there are 1uf ceramics already there you don't need any others but if you have already fitted them then leave them in, they won't do any harm.

 

[mtl777]

Wouldn't say it was a common problem tbh, not unknown though.

Other than the ground loop problem, the capacitor problem due to overheating was real after all...

Capacitors going bad in Delta 1010

Delta 1010 capacitor repair

So it was well and good that I fixed the heat problem that causes this. 🙂

I think I know now why, in the past, the RMAA tests would become good after I replace the caps and after a month would become bad again. It's because every time I replace the caps, I put non-conductive thermal grease on the heat sink. This acts as a sort of electrical insulation so it disconnects the ground loop. After a while, some of the thermal grease evaporates due to the heat and there becomes progressively more direct metal-to-metal contact over time. The ground loop is re-established and the RMAA tests become bad again, and get badder and badder as the days go by. I replace the caps again and do the same thing, and the cycle repeats. The problem is never fixed. Man, talk about wrong diagnosis! 😀

If there are 1uf ceramics already there you don't need any others but if you have already fitted them then leave them in, they won't do any harm.

I already removed them, thinking that they might have something to do with the ground loop problem. I don't think I'll bother to put them back anymore. I'm sure there's gonna be a better use for them than this.

One last question. Of the regulators, I'm getting the biggest input ripple on U6 (1.27V), and the next biggest on U23 (0.64V). I'm thinking of adding a large ceramic (47uF or so) in parallel with the 2200uF pre-regulator filter caps C22 and C60. C22 and C60 are located close to the regulators (about 2 cm away), so lead inductance is not much of an issue. I'm thinking that the ceramics will help reduce the input ripple and share the burden of the heat with the electrolytics as well. Do you think this is a good idea?

Thanks a lot! 🙂

 

[Mooly]

Never ceases to amaze me how poor some PSU designs are... interesting seeing the bulging caps.

Think you are spot on with the grease diagnosis... well done... bit of a weird one that but it makes sense.

47uf caps won't alter the ripple much... but you try it and see. The reactance of 47uf is way to high to have much effect at line frequencies, the type or quality of the cap doesn't really come into it... you need lots of xc 🙂

 

[mtl777]

Think you are spot on with the grease diagnosis... well done... bit of a weird one that but it makes sense.

Thanks, good to know I was right. 🙂 And thank you for the mica insulator suggestion. I have done it now and it works great! My RMAA tests are looking superb!

47uf caps won't alter the ripple much... but you try it and see. The reactance of 47uf is way to high to have much effect at line frequencies, the type or quality of the cap doesn't really come into it... you need lots of xc 🙂

Even if it doesn't reduce the ripple, I'll still do it just for the heat spreading. 🙂

BTW, what is your preference for local bypass at an IC - ceramic or film? I'm also curious what are tantalum caps good for?

 

[qusp]

.....snip/ I'm also curious what are tantalum caps good for?

nothing 😛

you see them recommended in datasheets quite a bit, but i'm yet to enjoy the subjective results of them once. for me I generally stack films on ceramics for bypassing. with the small smd films around now, you can quite easily and neatly just solder one on top of the other.

 

[Mooly]

I tend to use small polys these days for decoupling.
Tants... hmmm... years back they were the cause of many failures, more through improper use I suspect as much as anything else. My Micromega CD player uses them for decoupling on all the opamps and logic... it's full of them... and they have been OK.

Historically, one of the big advantages was that the leakage current is very small, so you could use really high value caps like 100uf in timer circuits along with resistors in the meg ohm range... but that's all ancient history, modern electros suffice for 99% of applications, for the 1% they don't we use a different circuit and method nowadays anyway.

 

[mtl777]

nothing 😛

you see them recommended in datasheets quite a bit, but i'm yet to enjoy the subjective results of them once.

Yeah, I haven't heard much good said about tantalums. 😀

for me I generally stack films on ceramics for bypassing. with the small smd films around now, you can quite easily and neatly just solder one on top of the other.

So my idea to put a film in addition to the ceramic was not so crazy after all. 😉 Do you find that combining film and ceramic works much better than just one or the other? Are there no problems with oscillation and the like?

 

[mtl777]

(Regarding RC snubber for AC supply) That's it, one at each end.
Is it better ? If it reduces a noise problem then yes... it's not going to make it worse anyway. This is why I keep saying you must measure and see what's going on. If there's no problem to begin with (and there probably isn't) then they will have no noticeable effect anyway.

I have installed the RC snubber in the 1010. I don't see any difference. With or without the snubber, my scope is showing a nice sine wave on the AC input. Maybe I'm not using the scope correctly or it's just not capable of showing tiny noise imperfections in the AC (it's only a cheap 25 MHz scope). Or maybe there is no noise problem to begin with. 🙂 Anyway, I'll just leave the snubber in there.

I tend to use small polys these days for decoupling.

Hmmm.. I may still put back and try again those 0.1uF films on the output of the regs.

Tants... hmmm... years back they were the cause of many failures, more through improper use I suspect as much as anything else. My Micromega CD player uses them for decoupling on all the opamps and logic... it's full of them... and they have been OK.

Historically, one of the big advantages was that the leakage current is very small, so you could use really high value caps like 100uf in timer circuits along with resistors in the meg ohm range... but that's all ancient history, modern electros suffice for 99% of applications, for the 1% they don't we use a different circuit and method nowadays anyway.

I think that tants are usually resorted to when space is at a premium, as they offer large capacitance in a small size. I just might try them, 22uF tants in parallel with C22 and C60 (pre-regulator filter caps for U6 and U23). They're the largest value I could fit in the underside of the board. However, I read that they have high ESR and low ripple current rating. On the other hand, the 2200uF electrolytics that they will parallel with are low ESR and high ripple rated, so maybe they will balance out. My intention is not so much to reduce the input ripple to the regs (I know they won't do much) but to share the burden of the heat. Do you think the tants will be fine here and not cause my 1010 to sound bad?

Thanks! 🙂

 

[mtl777]

More info on the 22uF tants:

25V @ 85 deg.C or 16V @ 125 dec.C rated voltage
-55 to +125 deg.C temp range
1.6 Ohms ESR
8% DF
4.4uA DCL
no ripple current rating published

The measured DC voltage across the 2200uF cap C22 is 12.7V and the measured ripple voltage across it is 1.6V. The current is 0.2112A as calculated from the formula I = Vr x C x F. I don't know how to calculate the ripple current, though. Will the 22uF tant in parallel with C22 be able to handle the ripple current and not fail? I'm concerned because I've read that tants are not good with ripple. I would have preferred a ceramic instead of tant but couldn't find a large value ceramic in the local store nearby.

Thanks again! 🙂

 

[Mooly]

I certainly wouldn't parallel any large electroylitic with a tant, it's just asking for trouble.

Tants have real limitations in use, such as the rate of rise of the applied voltage, and maximum voltage that can be applied in the presence of any ripple component. That value can be as low as 25 to 30% of the rated working voltage of the cap.

Tant's don't like heat, vibration, shock, any mechanical stress... and any transient voltage spikes and it's curtains.

They are going to go bang sooner or later... forget it.

 

[mtl777]

They are going to go bang sooner or later... forget it.

Wow, fireworks! 😀 Thank you for saving me from a potential catastrophe! 🙂

Well, how about a 10uF 25V Rubycon ZLG electrolytic which has 230mA rated ripple? That, plus a 1uF 100V X7R ceramic in parallel with the 2200uF electrolytic?

 

[Mooly]

The Rubycon is OK if you want to fit it, same for the ceramic.

 

[mtl777]

The Rubycon is OK if you want to fit it, same for the ceramic.

Thank you! 🙂 Meanwhile, I have put back the 0.1uF film caps on the reg outputs. Can't really see much difference on the scope or in RMAA, but on listening test it seems a little more 3D. I think I'll keep the film caps in there.

I'll do the Rubycon and ceramic parallel with 2200uF later when I get back from Vegas for a much needed vacation. Time to have a different kind of fun! 😉

Oh, one question before I go. The resistors I used for reducing the input voltage to the regs have a temperature coefficient of 250 ppm. Is that good enough considering that these resistors are running very hot? I'm wondering if it's worth the trouble of replacing them with 100 ppm - long wait and hassle of placing an online order and all?

Thanks again! 🙂

 

[Mooly]

Tempco is totally unimportant. A non issue.

 

[mtl777]

Hi, I'm back! Thanks for your reply on the TC! Glad to know it's a non-issue. 🙂

I forgot that adding the resistors (to reduce the input voltage to the regs) and installing them with very long leads to place them outside the board and draw the heat away from the board has increased the length of the path from the 2200uF pre-reg filter cap to the regulator a lot. The length of the path is now 4 inches compared to only 2 cm before adding the resistors. After reading this thread, I'm now having second thoughts about adding the 10uF Rubycon and 1uF ceramic in parallel with the 2200uF. The way I understand that thread, the common practice of bypassing a large electrolytic with a small cap (or combination of small parallel caps as I'm planning to do) is bad when the distance between the large electrolytic and the IC (in my case the regulator) is long. Is that true, or maybe it's true only if the bypass cap is very small, like, 0.1uF or less? But since I'm going to use 10uF electrolytic plus 1uF ceramic, will it be large enough to not be a problem? If it's still too small, what about 10uF electrolytic plus 10uF ceramic, or what do you suggest?

Now, if the bypass capacitance required in order to not cause a problem is so large that I won't be able to fit it, then I would be contented with putting a 1N5818 diode across the terminals of the 2200uF electrolytic just to be able to spread the heat a little bit. In that case, which terminal, (+) or (-), should I connect the cathode of the 1N5818 to?

Thanks! 🙂

 

[Mooly]

You are worrying a lot over all this 🙂

All you need ensure is that the regs are decoupled close to the regulator pins with a small cap. That kills all all problems of oscillation. The only problem with the longer lead length, is that the ability of the main cap to supress hf problems is reduced due to the inductance/resistance of the extra wire.

If you want to add diodes, the cathode (stripey end) goes to the + terminal of the cap so it's reverse biased and non conducting.

 

[mtl777]

All you need ensure is that the regs are decoupled close to the regulator pins with a small cap. That kills all all problems of oscillation.

There is already a small 0.1uF cap right on the input pin of the regs so that's taken care of. No worries of oscillation then.

The only problem with the longer lead length, is that the ability of the main cap to supress hf problems is reduced due to the inductance/resistance of the extra wire.

You mean due to the longer lead length, the ability to suppress HF is already reduced even if the bypass is not added? Will adding the bypass further reduce this ability? All I really want to confirm is, will adding the bypass be harmful or not? (Note: I will now be using two 15uF ceramics = 30uF as bypass. Browsing the online store and looking at data sheets, I have just found that this is the largest I can fit for reasonable cost.)

If you want to add diodes, the cathode (stripey end) goes to the + terminal of the cap so it's reverse biased and non conducting.

Thanks! And just to confirm, the 1N5818 will be fine for this and will not introduce any unwanted side effects, right? 🙂

 

[Mooly]

Diode is OK across all but C14 and 17

Longer lead length... if the cap were 100% ideal and perfect, after 4 inches of lead, then at the end of that lead it isn't perfect. That's why we add the small decoupler at that point.
You can add as much bypassing as you want... but is it doing any good ?

 

[mtl777]

Diode is OK across all but C14 and 17

Thanks, I'll keep that in mind. 🙂

You can add as much bypassing as you want... but is it doing any good ?

Well, as long as it's not doing any harm, it can at least help with the heat and reduce the ripple a little. I guess the only way to find out is to try it, so try it I will! 🙂