Transformer+rectifier+filter expected voltage?

[mmagliaro]

This is for the electronics-minded.
I am puzzled and frustrated by a basic thing that I have done many times before.

18 VAC transformer secondary (120v primary, as usual)
Measures 18V on a DMM, so we can reasonably assume, like always, that the meter reads RMS.

Put a full-wave bridge rectifier and a 2000 uF cap across the transformer, and now I get about 30.5 VDC.
Even with a 100 ohm resistor across it (so it's not floating), I still get about 29.5V

I would expect the peaks in an 18VAC output to be 1.414 x 18 = about 25.4V, and therefore it should be impossible
to get higher than that after I rectify and filter it, no matter how big the capacitor is.

So what am I missing here?

[peteski]

Interesting.  I would have expected the same values as you do.

What voltage values will you see when you create this circuit in your simulator program?

[mmagliaro]

Peteski, somehow I was sure I'd hear from you!
About the simulator, I had not thought to try that.  I did, and yes, I get exactly what  I expect.   If I set the AC source to be 25.4v peak, I get 18v RMS, and the filtered DC with the load on it is about 22 VDC.

But I think I have found the issue.  As it turns out, I am NOT getting 18 VAC on each side of the center tapped secondary.  I'm getting 22 (44VAC across the whole thing).    22 VAC RMS would be 31.1 peak, so there you go.  That explains why I'm getting about 30+ DC when I filter it.  When I get the load up to abou 0.8 A, it drops to 26.  I would be that if I push it all the way to the transformer's rated output of 1.56A, it would drop to about 18.  It is probably built to deliver its rated voltage at full rated current.

[peteski]

Peteski, somehow I was sure I'd hear from you!
About the simulator, I had not thought to try that.  I did, and yes, I get exactly what  I expect.   If I set the AC source to be 25.4v peak, I get 18v RMS, and the filtered DC with the load on it is about 22 VDC.

But I think I have found the issue.  As it turns out, I am NOT getting 18 VAC on each side of the center tapped secondary.  I'm getting 22 (44VAC across the whole thing).    22 VAC RMS would be 31.1 peak, so there you go.  That explains why I'm getting about 30+ DC when I filter it.  When I get the load up to abou 0.8 A, it drops to 26.  I would be that if I push it all the way to the transformer's rated output of 1.56A, it would drop to about 18.  It is probably built to deliver its rated voltage at full rated current.

Yes, transformer voltage ratings are at full current it is designed to supply.

I guess since you didn't provide more details, I assumed that it was a single-winding secondary, rated for 18VAC, and a 4-diode bridge rectifier.
Sounds like you have a center-tapped secondary where the center tap is common, and use a 2-diode full-wave rectifier (one diode on each leg of the winding).

[mmagliaro]

I actually used a full-wave bridge (4 diode) across the two outer legs on a center-tapped secondary, and then use the center tap as the common ground.

[peteski]

I actually used a full-wave bridge (4 diode) across the two outer legs on a center-tapped secondary, and then use the center tap as the common ground.

Ah, that gives you a dual-polarity (positive/negative) voltage output.  I would recommend using a separate filter cap for each leg (tied to the common ground).

[mmagliaro]

Ah, that gives you a dual-polarity (positive/negative) voltage output.  I would recommend using a separate filter cap for each leg (tied to the common ground).

Yup... did that.  Having dual polarity output was the whole point of this.  I just want to be able to run an LM317 regulator off one side and take a negative voltage reference off the other side so I can pull the LM317 down to true zero volts to get a regulated output that goes from 0 ... 12 (or whatever) instead of 1.25v... 12

[Point353]

Having dual polarity output was the whole point of this.  I just want to be able to run an LM317 regulator off one side and take a negative voltage reference off the other side so I can pull the LM317 down to true zero volts to get a regulated output that goes from 0 ... 12 (or whatever) instead of 1.25v... 12

Or, you could put two diodes in series with the output of the regulator.

[mmagliaro]

Using two series diodes ruins the current regulation precision of the circuit, a lot more than you'd think.  I even tried it (and it is something that circuit weenies on line always recommend against, and now I know why).  In my present situation, with the output set  on 3.050 volts, at 10mA load, if I put a 60 mA load on there, the output drops to 2.850 volts.  That's 0.2 volt variation!  Measuring the voltage drop across those two diodes, that is where ALL the error was coming from.  Diodes don't really drop a steady 0.7 volt.  That's, of course, a nominal value.   Their voltage drop goes up a little as the current increases.  It can vary from 0.6 all the way up to almost a full volt.

As it turns out, in this case, it was going up from 0.65 to about 0.75, and since there are 2 diodes, that's where my 0.2 volt variation was coming from.  Getting rid of the diodes got rid of that variation.  It also didn't just affect it at low currents (my first hopeful "guess").  i.e. if you load down the regulator with a fixed resistor so you're always drawing 200 mA or more, you'd think those output diodes would be locked in to whatever voltage drop they had and they wouldn't vary.  But they still do.

Now that I've used a negative supply and gotten rid of the series diodes, the output is within a few millivolts from 10mA to 250mA  (that's just the range I tested it on)

[peteski]

Max, if you want this type of accuracy, why not get one of the lab power supplies from eBay?  They go down to 0 Volts, and you can use them in constant voltage or constant current mode.  The one I bought has backlit digital readouts for both current and voltage. It has fcoarse and fine knobs for current and voltage adjustments.  I picked up a Techpower unit that can supply from 0-30V and 0-5A.  I don't recall the price but I think it was around $50.

[mmagliaro]

Peteski... BUY one??? Are you mad?  When I've got parts and junk lying around?   

Seriously, though, I already had this thing built, with nice meters and switches on it, and the only thing I never liked about it was the poor regulation.  Once I realized I could use a center tap transformer (which I also had lying around) and a negative voltage reference, to get it very precise, I wanted to figure it out for myself.

In case anybody cares, here's a typical example of what the circuit looks like.  Typically, the data sheets will show a precision reference zener diode in there, instead of the two standard diodes I used.  Note, this is NOT the same as putting two diodes inline wtih the output.  The lower voltage connection (obscured by the railwire source label) is -30v in my case, but that is purely arbitrary.  The key is just to get the lower end of the two diodes at about -1.4v  (voltage below ground), and the 2.7k resistor is just a convenient value to draw enough current through the diodes so they turn on and drop "about" 0.7v.  The trick to this is that the path through the diodes and the 2.7k resistor is very stable.  The current doesn't vary with the load or output voltage of the regulator, so the two diodes drop a very constant amount in this case.  A precision zener reference diode would be nicer, but these are pretty darn good.

This circuit gives you "close to zero" output at the lower end, a few millivolts.  And you can tweak the R2 and R4 if you are really determined to get a perfect zero output.

The two C1's are mislabeled.  Should be C1 and C2.  0.1 and 1.0 uF are typical values to prevent oscillation in the regulator.
V+ needs to be about 3v higher than the highest output you want, so for a 12v max output, V+ needs to be at least 15. 

[Point353]

Using two series diodes ruins the current regulation precision of the circuit, a lot more than you'd think.  I even tried it (and it is something that circuit weenies on line always recommend against, and now I know why).  In my present situation, with the output set  on 3.050 volts, at 10mA load, if I put a 60 mA load on there, the output drops to 2.850 volts.  That's 0.2 volt variation!  Measuring the voltage drop across those two diodes, that is where ALL the error was coming from.  Diodes don't really drop a steady 0.7 volt.  That's, of course, a nominal value.   Their voltage drop goes up a little as the current increases.  It can vary from 0.6 all the way up to almost a full volt.

As it turns out, in this case, it was going up from 0.65 to about 0.75, and since there are 2 diodes, that's where my 0.2 volt variation was coming from.  Getting rid of the diodes got rid of that variation.  It also didn't just affect it at low currents (my first hopeful "guess").  i.e. if you load down the regulator with a fixed resistor so you're always drawing 200 mA or more, you'd think those output diodes would be locked in to whatever voltage drop they had and they wouldn't vary.  But they still do.

Now that I've used a negative supply and gotten rid of the series diodes, the output is within a few millivolts from 10mA to 250mA  (that's just the range I tested it on)

Which diodes did you use in series with the output of the regulator?

What is the minimum regulated output voltage you require?

What precision do you need for the output voltage and over what load range?

[mmagliaro]

1N4001 (100v 1A)

I want regulated output from 0 - 12v

Current regulation: vary by no more than 10 mV between 10mA and 250mA

---------------
The drawing shows bigger 5400 series diodes, but I actually only used 1n4001's in there.
I am getting excellent regulation, until I get near 12v, and that's only because my input is 15V and the transformer
sags a little under load, so it can't sustain 15VDC.  But up to about 10.5V, it's rock steady within 3 mV even up to 1/2 amp.

[Point353]

1N4001 (100v 1A)

I want regulated output from 0 - 12v

Current regulation: vary by no more than 10 mV between 10mA and 250mA

---------------
The drawing shows bigger 5400 series diodes, but I actually only used 1n4001's in there.
I am getting excellent regulation, until I get near 12v, and that's only because my input is 15V and the transformer
sags a little under load, so it can't sustain 15VDC.  But up to about 10.5V, it's rock steady within 3 mV even up to 1/2 amp.

A 1N5401 diode would exhibit much less change in V_F as a function of I_F over that range of load current, as compared to a 1N4001.
A maximum value of 10mV load regulation is a rather tight spec, and the datasheet for the LM317 doesn't guarantee that level of performance.

[mmagliaro]

A 1N5401 diode would exhibit much less change in V_F as a function of I_F over that range of load current, as compared to a 1N4001.
A maximum value of 10mV load regulation is a rather tight spec, and the datasheet for the LM317 doesn't guarantee that level of performance.

True (about the 1N5401).  I noticed that in the performance curves on the data sheets.  But I tried it with two of those in series and still got too much variation... in the tenths of volts, instead of mV or even hundredths.  With the diodes where I have them now, the current through them varies so slightly that their voltage drop variation is miniscule.