Bipolar Supply, Unbalanced Load

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Does running a bipolar power supply off a center tap transformer cause noise if the load is unbalanced?

I'm building a bipolar power supply for a synthesizer project, and have some noise..
99% of the thing runs on positive 12v. (Only the filter needs negative 12)
My transformer has 2 secondaries, and I had them wired up as a center tap. That's when I could hear buzzing. I rewired the power supply with a single secondary, as just a positive supply, and it works perfectly w/ no noise.

Is that because the load is so unbalanced?

Would wiring the thing up as basically two power supplies (using the transformer secondaries separately), one positive and one negative, and then connect the grounds after the regulators be a better idea?

The supply is simple.. Transformer to rectifier, 2200uf cap, 7812/7912, 100uf cap.
(Not going for hi-fi, or anything here)

Thanks for any advice.
 
And by that you mean the I had the two secondaries wired incorrectly, right? Or do you mean another aspect of my grounding?

It's an Antek transformer. I went by their schematics, but that is definitely an area I'm not 100% clear on, as their schematics were different from others I've seen online (Antek schematics have no phase dot, either). But aren't the only options in phase or out of phase? They're definitely in phase, as I get 24v between the two secondaries... If I reverse the bottom secondary, I get ~0v. The way I have it is the two primaries are in parallel, and the secondaries in series, connected at the middle leads.. (bottom of first coil + top of second, blue+green wire connected, if you know about antek's lead coloring)
 
The usual mistake is to connect the CT to the star ground (or whatever ground arrangement is used), then the caps to the ground. This pushes charging pulses right through the ground connection.

Correct version is to connect CT to caps, then the other (smooth) end of the PSU to ground. Remember, with grounds you must not assume that a conductive connection implies equipotential, as there can be large currents flowing in some places.
 
Ok, this is what I did..... Center tap as you said, is where the caps connect, and is ground for the regulators. Basically like a star ground for all the power supply components. Then the output from that (+12,0,-12) all goes through the power cable (about 1' or 30cm long) into the synth. The connection from the star ground in the power supply goes through the cable and is the ground (0v) for the synth.

Is that correct? If not, where do I get ground? I thought the center tap was basically making ground.

Thank you for your help. I do have it working the other way, using the secondaries separately, a bridge rectifier on each one.
But, I would like to understand what I am doing wrong w/ the center tap version.
 
The CT provides current in dirty big short pulses. You need to keep those current pulses well away from your ground, yet there must be a connection from CT to ground somehow. The solution is as I said: CT to caps, caps to regs ground, regs ground to star ground. That ensures that the star ground only sees pure DC.

Always ground a PSU from the clean end, not the dirty end, and don't mix up the ends into one ground. A PSU should really use a ground bus, even if one end (the correct end!) then goes to a star.
 
Usually when people ask about grounding, I refer them to this excellent article.
Chapter 3 is relevant to this thread, but it's worth reading the whole article.

If you don't have the time to read it, here are two relevant images from that article. Also from it: "If signal reference current is routed through the power supply buss, this noise voltage will be impressed onto the signal by Common Impedance Coupling."
This may have been the reason why you had the noise.
 

Attachments

  • Image1.gif
    Image1.gif
    8.6 KB · Views: 245
  • Image2.gif
    Image2.gif
    5.8 KB · Views: 244
Last edited:
The second image (fig 3.2-4) shows you how not to do it. It is not a clean power supply - it uses the classic dirty CT noise injection which I have been telling the OP to avoid.

The first image shows why it is wrong. You ground a PSU at the clean/quiet end, not the dirty/noisy end or somewhere between.

This is the author's explanation why 3.2-4 is the right choice: "In figure 3.2-4, the internal ground buss is collapsed into a point, forming a local star ground. Bringing everything to a point forces us to make a connection to that point – no more multiple-point connection over which a noise voltage could form. Before, we had two connections: a high noise one connected to safety ground and a low noise one connected to signal reference. Note that now the high noise point is directly connected to the power common where it can be connected to safety ground to drain the AC leakage current, and the low noise point is directly connected to the power common where it can be connected to signal reference. "

Edit: I do agree with the dual rectifier approach: this forces you to connect the low noise side of the PSU to the star ground.
 
Last edited:
Ok, thanks.
I do have it built on breadboard for now, all the components stuck in w/ untrimmed legs, and while I called my grounding scheme "star", it was messy. I was assuming the unbalanced load was the issue, but it was just a guess. Since everyone is thinking it's ground related I'll rebuild it neatly.

Well, with a breadboard you almost automatically end up with a bussed ground.
Would you be willing to try to build the PSU as in fig. 3.2-4 from post #9 and see if that does the trick?
 
The author's explanation is wrong. He has directly injected charging pulses into his ground. If everyone reads and follows this misinformation it might explain why so many people get their CT grounding wrong. He seems to be more worried about a small amount of noise/interference from the mains supply than the much larger problem of charging pulses.

The charging pulses are the source of the noise/interference.

I don't agree with you and that became even more apparent to me when I noticed the images attached to post #9 were of the PSU alone, not of the whole device. The power common is not yet connected to anything in that image and the "star" is local, as the text I quoted says. I'll explain below why I disagree.

Always ground a PSU from the clean end, not the dirty end, and don't mix up the ends into one ground. A PSU should really use a ground bus, even if one end (the correct end!) then goes to a star.

It is the bussed ground that is the cause of noise, not the cure for it. See fig. 3.2-1, the noise from the bussed PSU affects the signal reference.

A current through a resistance creates a voltage, Ohm's law. The lower the resistance of the path the current pulses must travel through, the better.
Fig. 3.1-1 shows the "bussed" ground that creates a higher voltage noise (created by the current pulses) than necessary. Fig. 3.2-4 minimizes this and avoids the connection across a noise voltage (noisy side to safety earth and quiet side to signal reference).
With earthed appliances (as we DIYers are forced to make because we cannot guarantee our creations are compliant to double insulation standards ) this is even more important because this noise voltage can drive a groundloop. From the same article: "Loops aren't bad, it depends on what's in the loop. Unless there is a voltage generator to drive a current around the loop, or radiated current into the loop, it is merely a parallel path. Consider the parallel shields of a left and right channel stereo cable."

The more I think about it, the more fig. 3.2-4 seems right. Esp. if you realize that the star drawn there is not the main starground. What seems contradictory is that the PSU star is both the low and high noise point. Left of the star is the high noise point (here flow the current pulses), right of the star is the low noise point (the current pulses don't flow here). In the star itself the currents in the loops left and right of it obviously flow together, but since this local star is just one very small very low resistance point, the current pulses can't create appreciable noise that would affect power common and signal reference.
 

Attachments

  • Image4.gif
    Image4.gif
    10.7 KB · Views: 143
  • Image3.gif
    Image3.gif
    11.5 KB · Views: 141
Last edited:
jitter said:
It is the bussed ground that is the cause of noise, not the cure for it.
No, it is the connection between dirty ground and safety ground which causes the problem. Safety ground should be connected to the clean ground. The solution is very simple, and to be honest I am getting tired of how many times I and others have to keep saying it: keep charging pulses in tight loops away from grounds, and connect the ground to the clean end of a PSU.

Connecting the transformer CT and all the caps to a star, whether it is a PSU star, safety star or (even worse) signal star just creates problems. You can minimise the problems by getting the connections in the right order at the star, but much better to avoid the problems altogether by doing it right: CT-caps ground should not go to the star.
 
No, it is the connection between dirty ground and safety ground which causes the problem. Safety ground should be connected to the clean ground. The solution is very simple, and to be honest I am getting tired of how many times I and others have to keep saying it: keep charging pulses in tight loops away from grounds, and connect the ground to the clean end of a PSU.

Connecting the transformer CT and all the caps to a star, whether it is a PSU star, safety star or (even worse) signal star just creates problems. You can minimise the problems by getting the connections in the right order at the star, but much better to avoid the problems altogether by doing it right: CT-caps ground should not go to the star.

an simple image will help us :)
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.