Multi-Tap Transformers Versus Voltage Regulators?

[stryd_one]

Man, that is just weird. The vregs should be outputting 5V (+/- 0.notmuch) or nothing at all. It's like you've somehow made it into a constant current supply... ?

[m00dawg]

Yeah it's got me stumped too. It's the same basic design as I posted previously so it's either my poor prototyping skills or something is actually screwed up in the design. Not sure which :) But I'm thinking at this point, it could be the design. I dunno though. Quite a bit of AC hum was leaking through so maybe I do have something jacked on the prototype I just didn't see.

The printed version of this circuit should be in from Batch in a few weeks I think. I may try an alternate design in the interim but, either way, will keep everyone posted! In the meantime, if anyone has any thoughts, ideas, or flames, I'd be happy to hear there :)

[m00dawg]

No disregard needed - there was some additional information not in that previous six pages :) So it was a good read, or, if nothing else, a nice refresher.

Seems like the regulators are getting plenty of power from the xformer. You made me think of a good point, though - I haven't checked the voltages on the outer and center taps when I notice the drop on DC voltage. I would assume it would be normal or the regulators would stop providing power, but it never hurts to check! :)

[MRE]

well dang.. I was too fast on the delete button then.

Well, glad you got something out of it, and an idea for troubleshooting your current situ..

Ive looked over it, and Im a bit stumped.

still trying to figure out the configuration to be honest. It is .. um.. unconventional. At least, its not a standard 'taught' design ;)

[MRE]

Might have been covered already (I admit I didn't read ALL the posts in the thread)...

Have you checked the input to the lower voltage regulator (I think this is your 5 volts)?

Are you certain it is not AC there?

I *think* you still need one diode here, because you are getting a push-pull effect on the input of the 7805.

Yes, the top half still provides the ground reference for the lower half. That may, in effect, be the problem?

In the lower half of the center tap, current flowing 'downward' would supply the VR, in reference to the bridge's ground.

But when current is flowing 'upward' isn't it pulling current to supply the centertap to bridge connection?

Also, if the bridge is supplying DC to the top half on both directions, and the lower half is supplying it only on half the cycle (one direction) then it is reasonable to assume that you will have less overall current available on the lower half. I DID notice the difference in caps, but I think they would have to be considerably larger to hold that 5 volt regulator stable (remember that *IF* I am correct, you are only supplying it with +current for 1/4th a cycle, and robbing it of current for 3/4ths).

EDITThis would imply that even with a diode to fix the problem, you are still supplying it only half the available cycle, which would mean your caps would have to work awfully hard to supply a core, an LDC, and several IO boards. Your 5 volt supply absolutely needs to be full-wave rectified. If anything, it should be on the top end of the circuit.end edit

So, is it possible that the caps on the right side of your 5 volt regulator are simply holding the slack for that LED every cycle, while the regulator is in fact dropping out most of the time?

That would explain an overall apparent voltage of less than 5 volts, and the regulator would *appear* to be on. It pops in just long enough to charge the cap bank. They discharge over the period of the cycle (likely down to about 2 volts in reality due to the LED. Right about the time they give out, the regulator jumps back in to save the day.

All of this on a DMM will look like the average of the upper and lower voltage limits during the cycle (whatever the REAL values may be).

You really need to scope the input and output of the V-reg to know for sure what is happening.

I dunno. That's just the impression I got from looking at the schematic. Something is fishy.

[MRE]

PS- That diode would be going between the transformer and C2.

Quite honestly, I think even if it is NOT the problem, it would still be prudent to have one there.

[m00dawg]

Thanks MRE! That's been most helpful, and I see what you're getting at. In fact, it makes a lot of sense. When I had smaller caps on the 5V side, when the load was high enough, I started to see the 5V LED start to flicker. I thought it might be because of AC, but if the regulator is actually shutting down, that could explain things.

Using a diode on the 5V side has been discussed. I thought about putting one on the schematic, but in the circuit simulator, it didn't seem to make much difference. It's easy to add in though so I'll see if Radio Shack has one or just order a few of these guys. I think there is AC hum leaking through. The SID testtone app outputs a noticeable hum. If this was coming from the 5V supply, I assume it will still hum right?

As for the caps, my design called for one 6800uF and one 2200uF cap on the 5V before the regulator. Do you think that isn't enough? The simulator said it might be but, then again, it's a simple app so I just sort of went with it. I originally had a 2200uF and 47uF and that seemed to be much too low.

If I were to take a guess, I'm assuming your suggestion is to just use the outer taps, fully regulate and smooth that into DC and feed it to the two regulators directly? I was hoping to build something more heat efficient but now I'm thinking that this approach may be the best option.

Thanks again for all the help and insight!

[MRE]

I will have to think over the other questions.. but as to the question about just using the outer taps:

That is the "typical" design. Center taps are great if you need a negative voltage supply.

Other than that, most (if not all) designs I have seen were either multi-coil (custom and expensive) or simply took the outer taps and applied them to both regulators.

You are right, it is not efficient for the 5 volt supply, and you will need a hefty heat sink.

Another (bulky) option would be two transformers, matched to the desired output.

I can re-post all that info about transformer selection if you'd like.

I'm not saying your design is not viable... just unusual (and certainly creative).

I will dig up the calculations for those caps. I am afraid though that I would only have calcs for full wave and half wave.

In your case, the only way to know for sure that you have the right ones might be some experimentation (as any suggestions I could give would be based on doubling the half-wave calculation and the regulator data sheets).

Do you have an o-scope?  Not some home built ADC/PIC etc kit that is only good for 5 volts.

I'm talking about something that you can look at those transformer outputs with.

[MRE]

As for the caps, my design called for one 6800uF and one 2200uF cap on the 5V before the regulator. Do you think that isn't enough? The simulator said it might be but, then again, it's a simple app so I just sort of went with it. I originally had a 2200uF and 47uF and that seemed to be much too low.

How did you arrive at these numbers? What equations did you use?

[m00dawg]

How did you arrive at these numbers? What equations did you use?

Magic :) Well actually no, I did this in part because it seemed like more capacitance was needed on the 5V side. I thought it might be half-wave though - you're thinking it's more like quarter-wave or something like that? Basically, I knew 2200uF and 330uF were good pairs for a standard full-wave rectified setup, so I basically assumed I needed to at least double it. So, I got the biggest reasonably inexpensive cap that would still fit on a reasonably sized board - so the pair ended up being 6800uF and 2200uF.

That's the idea anyway. For my current board, I used a 6800uF but still have the 330uF in place. I can switch that out, but it sounds like that still may not be enough?

As for your previous question, alas, I do not have an oscope, but I might know someone I can borrow one from. I'll go ahead and check on that and see if that might be doable.

[MRE]

I would advise leaving the caps alone for the moment.

Regardless of their values, they are unlikely to be the culprit for the overall problem.

Once THAT is solved, you can always fine tune your caps. The O-scope will be a big help doing that.

While it is true you would want to increase the 'storage tank' capacitor (the larger value one) when you change from full wave to half wave, you generally can leave the smaller cap alone. (In fact, you might find it works just fine at the first value you came up with)

The larger cap is there to serve as sort of a battery. It charges up and maintains a high enough voltage to keep the regulator going during the downswing of the transformer.

The smaller cap, is more of an filter for transient frequencies. These can happen anywhere, but often during zero crossing (imagine having home lighting kits bursting data during that time for instance).

By the way, via spec sheet of the 78xx regulators, immediately before the regulator, you should have a .33uF cap. It is only REQUIRED if there is some distance between the reg and the PSU filter, but still, a good idea. Likewise, immediately after, you should have a low value cap no lower than .1uF.  Anything lower can clunk it out.

These additions *may* help with your ac hum.

Anyway, to the gremlin at hand:

I'm not sure, but I think you have a quarter wave here. I've never designed a psu this way, so I don't know exactly what it will do. I might dig out Electronics workbench and make a sim of it just to see what it looks like.

I would suggest the following steps:

1 - throw that diode in  transformer --[>|---(caps)---regulator        Do this first, as you may be damaging that reg as you use it.

2 - check with an o-scope both before the diode, and at the regulator.

nununu - pretend that is a normal 60hz sine wave at whatever output voltage your transformer provides (both outer terminals).

nnnnnn - full wave rectification  (after the bridge on the top side circuit, referenced to ground, REST OF FILTER/REG CCT DISCONNECTED)

n_n_n_ - half wave rectification (lower end after diode. rest of circuit disconnected. If you get this, you can still use it with a large cap) You can actually test for this easily. Just use some clip leads to attach one end to the transformer lead, and the other to the diode. Now clip the scope to the opposite end of the diode. No need to connect to the circuit. Use the ground reference built from the top side.

n__n__ - quarter wave rectification (lower end after diode. rest of circuit disconnected. If you have a long period after the wave, or it is considerably smaller peak than you expected, you will need to redesign your psu. It will be quite difficult for a cap to keep the regulator stable)

WITH the filter and regulator circuits connected (i.e. oscope probe at the regulator), you will get a different wave form... It is hard to do in ascii art, so I will try to describe them: Imagine if you cut just the very tops of a sine wave. Now as you come over the top and crest the wave, rather than falling down the other side, draw a diagonal to the next peek, just below the top, and start again. looks kinda like a rounded sawtooth waveform, but the right side is stretched way out. That effect is caused by the discharging cap.

That is the ideal waveform going into the regulator.

If you get a peak, with a long fall-off that eventually hits zero.. you have a problem. If it stays at zero any length of time before the next wave comes, you have an even bigger problem.

[m00dawg]

By the way, via spec sheet of the 78xx regulators, immediately before the regulator, you should have a .33uF cap. It is only REQUIRED if there is some distance between the reg and the PSU filter, but still, a good idea. Likewise, immediately after, you should have a low value cap no lower than .1uF.  Anything lower can clunk it out.

These additions *may* help with your ac hum.

Yep, I actually do have those. In fact, they are right in line what you suggest (.33uF and .1uF) on either side of each regulator. Still am hearing a good bit of hum though. I haven't heard back about borrowing and o-scope but your suggestions have been quite helpful! Hopefully I can track one of these guys down and put your suggestions into practice.

In the interim, I may just build a more standard setup, or modify an existing design I already have built so that I can at least work on testing my MB-6582 while I wait. Of course, I'll keep everyone posted on the results!

[MRE]

good luck to you

[/tilted/]

hiya.

Sorry if I've missed something, but from what i have seen, it looks like you have a wire missing.

So... if you are set up like this:

[tt]

Trafo            | PCB        |         |Synth

(9v tap)---------|*-  -9v out*|---------|*

(9v tap)---------|*-  -0v out*|---------|*

(5v tap)---------|*-  -5v out*|---------|*        [/tt]

Then you are missing a connection.

If the transformer is a single unit, providing 9v-0-9v with a 5v tap, ie: 9v-5v-0-9v or 9v-0-5v-9v, then you'll need to connect your center tap (0) to GND on the PCB.

If you have two transformers, one 18v CT (9v-0-9v) and one 5v (5v-0), then you need to connect BOTH 0v taps to GND on the PCB.

Without this connection, the PSU will still work, but will be somewhat unstable, as the ground reference is 'floating', and is dependant on load taken by the 9v side. It should be relatively stable, since the load on the 9v-0-9v side is inherently symmetrical in this design.

I think this *might* go some way toward explaining your symptoms.

EDIT: PS: Only make this connection if my impression of your PSU setup is accurate, and the connection is not already there. Making this connection if not needed may void your warranty, steal your car keys, burn your house down... etc. /tilted/ takes no responsibility for explosions or chip fatalities... etc.

[seppoman]

oh not again please, I thought this thread was already beyond the point that this is NOT about a bipolar supply ;) so the center tap is definitely not to be connected to GND. This is about using a regular center tapped transformer, i.e. no extra 5v winding, to generate +9v and +5v. My suggestion on which this schematic is based is surely unconventional and there might be a reason I don't yet understand why this would only work well with balanced loads. If this is the case, changing to half wave rectification is probably a good idea (i.e. put only a single diode on each the center and + output, continue to use the "-" output as GND and go on from there).

S

[m00dawg]

Wish I could find a multi-tap transformer that did offer +9 and +5 though. Either way, since some of the schematics are a bit old and buried in previous pages, here's the latest and greatest which may help the conversation a bit. Note that the center tap is X1-3. This was to make it easier to route the wires on the board itself.

@MRE

Good news! I should be able to get my hands on an O-scope! The bad news is that it may be a few weeks out :P So until then I'm just going to use a fat heatsink on the 7805 and build a more conventional setup.

@seppoman

I'd be ok doing half wave rectification but where does the negative come from? The other outer tap or?

Thanks everyone for all the help! I dunno about you guys, but this has been an enlightening learning experience.

[MRE]

have you considered a switching regulator for the 5 volt supply?  higher part count, a bit more expensive, and a little more board space, but in the long run, MUCH more efficient.

[m00dawg]

have you considered a switching regulator for the 5 volt supply?  higher part count, a bit more expensive, and a little more board space, but in the long run, MUCH more efficient.

I sort of discounted that because of the heated debate on the forums about whether or not that's a good idea for the MB-SID stuff (due to the fact that the SID is analog and switching supplies tend to spew RF). I've seen people say they work well, though, but it's always been after ignoring the pushback of people on the forums :) Sort of makes sense I guess. The SID uses 5V too so it seems like it certainly can't help noise, but heck I'm willing to give things a shot :) Part of the point of this project is to make something that's reproduceable for others who can't or don't want to use a busted ass C64 PSU for powering their beautiful SIDs.

On a related note, I swapped out the regulators of my old power board so I could output 9V and 5V from the same taps. It actually works pretty well, although I'm still seeing a voltage drop in proportion to load (it's just not as big). I think it's specific to the MB-6582 board, the cable to do, or something along those lines. I need to try it again, but I checked the voltages on the board with a PIC and 2 SIDs stuffed and was getting like a lower voltage reading from the board than I was when reading it off my power board. So I think there is still some grounding weirdless. Although, I don't hear ANY AC hum, so that's a good sign.

[m00dawg]

Oh I should add the voltage drop isn't enough to make things not work. It's around 4.5-4.9 or so I think (but I'll need to try that test again this evening to double-check that). I haven't added the BankSticks back into the equation, though, but I did stuff another PIC without much difference. The LCD backlight remains quite bright, which is a good sign. Just not sure why I'm not closer to 5V on the 5V rail. The 9V rail dropped only very slightly - from 9.02V to 9.00V.

[m00dawg]

Ok so I did another check and the voltage is about 4.9V with 2 SIDs, 1 PIC, LCD, and 8 Banksticks. Later today, I am slowly going to start adding more PICs and SIDs and see how things go. I did play around with it in it's current state and it worked like a champ. Filters work on the SIDs (*whew*) and no skipped notes or anything like that and it's fairly quiet. There's some hiss here and there (it's barely noticeable) but no hum. The heatsink on the 7509 and 7809 seems to work well. While I still want to figure out the center-tap stuff, but since I have to wait a while in order to get access to an o-scope, I think I may get a 9V transformer instead of my current 12V and go from there.

Again, will keep anyone that happens to care posted :) Thanks again for all the help on this one!

[/tilted/]

OK. I've had a really good look now.

I never assumed this was a bi-polar supply, but I was (and am) struggling with two main points of this design.

1) GND reference is coming off a rectifier. Interesting. Not sure it will work.

2) 5V rail (pre regulator) seems to have no rectifier at all.

These two points suggest to me that respectively

(1) The GND will have a lot of AC on it, and not really be a GND. (actually, a GND can be pure AC for all I care, so long as everything else is stable in respect to it.)

(2) The 7805 will drop out / heat up a LOT, and not really be able to stablise voltage, due to lack of rectification. Also, capacitors on this line will be ineffective, as they will be charged and then violently discharged in equal amounts each half-cycle.

Please tell me I have it all wrong. I just cannot see how this design is supposed to work, because it just doesn't look right to me, and I'm really not sure what is being achieved by not using a more conventional design. That said, I'm keen to try it, and I will do so in the next day or so. (I also own a scope, so I'll post any results I get.)

[seppoman]

Please tell me I have it all wrong.

You have it all wrong :P

No not really. I know this is quite a crazy concept and as I said, maybe it's better to use half wave rectification, and I didn't say I was sure it'd work under all conditions. But it's not that crazy after all:

1) GND reference is coming off a rectifier. Interesting. Not sure it will work.

think about the way e.g. the Core module does its PSU section. GND reference from a rectifier is actually the norm for unipolar designs.

2) 5V rail (pre regulator) seems to have no rectifier at all.

From the theoretical point of view, it sure is rectified, because it's on the center tap and the rectifier on the outer taps makes sure that GND is always below and +12V is always above that point.

(1) The GND will have a lot of AC on it, and not really be a GND. (actually, a GND can be pure AC for all I care, so long as everything else is stable in respect to it.)

yes, that's the point, as long as everything else is stable in respect to it. GND is only a question of the point of view. Think of the transformer plus rectifier as if it's a one-dimensional bird that is in flight and the position of its wings and body are voltages on this one axis. If you nail the bird's body to something fixed, like in a bipolar design, the wing's tips will move in positive and negative direction. But if you nail one wing's tip to something fixed and the bird still does the same movement, the body AND the other wing will move away from that point in the same direction. Strange picture, I know ;) So in that picture, it doesn't matter which point you define as GND, GND IS the reference so it can't have any AC in reference to itself.

(2) The 7805 will drop out / heat up a LOT, and not really be able to stablise voltage, due to lack of rectification. Also, capacitors on this line will be ineffective, as they will be charged and then violently discharged in equal amounts each half-cycle.

see above - in an ideal world, it is rectified.

The only question that'd maybe spoil the concept is: Would unequal load distribution in conjunction with real world properties of the used parts (capacitance/inductance) lead to situations where this bird image doesn't fully apply? My suggestion is based on mere theory and is just an experiment of thought, not a recommendation based on real experiences.

So if it turns out it doesn't work well in all cases, the other idea is to use half wave rectification, i.e. use only one diode on center and one outer tap and use the other outer tap as GND.

S

[/tilted/]

Eeewwww.

I think that bird analogy belongs in this thread.  :)

Perhaps then, the problem is that we are nailing the bird's wing to a rotating tyre?

It feels to me that the stability of the GND in this design is dependant on how much current is being drawn through the bridge and the 9v reg.

Personally, I'd have gone for the old two bridges, pretend the transformer is 0-6-12v (for example), tie GND to 0v tap approach. But that's just me.

[m00dawg]

It feels to me that the stability of the GND in this design is dependant on how much current is being drawn through the bridge and the 9v reg.

Personally, I'd have gone for the old two bridges, pretend the transformer is 0-6-12v (for example), tie GND to 0v tap approach. But that's just me.

Hmm...that's interesting you mention that because the problem I have been having with this design is that, the more power it pulls, the less voltage I get. I also get a lot of AC hum. Seems to me that crazy ground explanation you gave previously may be panning out? I'm not sure though.

As for using two bridges - I tried that and failed :) I was never able to get 6V and 12V (it was just one or the other) when using two bridges. I have no problem using that approach, but how do I hook those bridges up with the xformer? Can you even do full-wave rectification this way? Do they both share the same outer-tap (which would be GND) or? I tried multiple times doing it this way so any help in that regard would be appreciated.

[/tilted/]

Okie dokie.

I made a prototype of this:

on breadboard, with the following differences:

1: The transformer is a type 2155 (15V, 1A, multi-tap, taps used: 0v, 6V3, 9V4)

2: Some small differences in Electro caps (6800uF is 3x 2200uF, 330nFs are a 470nF (7809) and a 560nF (7805) - all I had at hand)

3: none of the post-reg caps are mounted yet, so I can get a better look at the regs themselves.

Some interesting results.

With the entire 5v section removed, 9v supply works as a normal unipolar supply (no surprises here). nice smooth output, even with no additional smoothing caps.

With the 5v section inserted, the 9v reg input is a little lumpier, and looks more like smoothed half-wave, than smoothed full-wave.

The 5v regulator input looks like truncated AC, rather than smoothed regulated AC (ie, more like a square wave, rather than a sawtooth-ish wave). This might explain the AC you described on your 5V rail.

The 5v regulator output has more noise than would normally be the case, with positive and negative going ripples from the baseline voltage. This noise is not huge (ca 5mV) -mind you, these tests were done with no load, but also with no additional smoothing caps.

With the addition of a single diode before the 7805:

The noise on the output of the 5v reg is still there, but the negative ripples are gone (as would be expected), but the positive ripples are still there, in a spacing consistant with half-wave ripples.

Pictures to come, once I find a place to put them.