Linear power supply: avoiding capacitively coupled ac stray voltage

[hibone]

Dear all

I put together a very simple power supply based on a "linear" transformer, a diode bridge and a cap.
Since I am meant to avoid noise, and ground loops, I need to shield the power supply the cable and all, and for that I tied the shield to the main earth line.

Problem is, the output being floating, it floats all over the place. If I measure AC voltage between the output and the shield, I read nearly  80Vac with respect to the shield, on the hot line and on the cold line as well. The output of the power supply is 60Vdc with no ac component whatsoever.

As such I'd like to prevent the AC stray voltage to reach all the way to the output, but I cannot tie the hot line nor the cold one to the earth line for various reasons, mostly related to the avoidance of ground loops and measurement issues when using a scope.

How can I solve this?

Should I put a 10+10 MOhm resistor series between the transformer leads and tie the center of the voltage divider to the earth line (thus before the rectifier)? 

Somehow it doesn't feel all that right to me.

[Kevin.D]

see my next post .which makes reference to your problem. You need to ac couple to earth to attenuate Common mode voltage.

[dom0]

Quality lab supplies have one (or more) electrostatic shields for this very reason in their mains transformer. This reduces the inter-winding coupling capacitance between primary (mains) and secondary (floating) to very low values, some transformers reach <1 pF.

You can either
(1) Shunt the secondary winding to ground with a small capacitor. This shunts some of the current flowing through the inter-winding capacitance to ground. On the other hand it increases ground-to-earth capacitance of the supply even further.
(2) Get a high-quality transformer with electrostatic shielding. These are quite rare used, but often go just as cheap as normal transformers. Obviously you can have a transformer custom made for you, but these shields cost extra and are quite a bit more expensive than a simple winding.
(3) Get a different transformer with less coupling capacitance. Standard toroidal transformers are very poor here (EI core with stacked windings are poor, too), while a EI core (or similar) transformer with separate winding chambers can be quite acceptable.
(4) Chain multiple transformers back-to-back. I have done this once for a supply, to decrease said capacitance and to increase insulation. Problem: Very uneconomic in time, space, cost and weight all at once, even for little power. So this is probably not an option for any lab supply.
(5) Cancellation by another transformer:

There's always the direct cancellation approach:  Use another small transformer (lowest VA possible) to provide a voltage opposite phase to the incoming mains, and of similar amplitude.  Couple that to output 0V via a carefully selected class Y1 capacitor to cancel as much of the stray voltage as possible.

[Gyro]

How can I solve this?

Should I put a 10+10 MOhm resistor series between the transformer leads and tie the center of the voltage divider to the earth line (thus before the rectifier)? 

domO has pretty much covered it. An electrostatic shield would be the ideal solution (some DVMs even have two shields, one to ground and one to the input guard).

Instead of 10M resistors to each side of the secondary, you would probably be just as well putting a resistor from the more earthly side of your DC output to ground - only you can decide how low you could get away with. On the up-side the transformer inter-winding capacitance isn't going to be very large (for a split bobbin one anyway) and it is only coupling at line frequency so the current you need to shunt to ground is actually quite small (higher for (Edit:) noise spikes though, a small cap in parallel with the resistor would help there).

I'm not clear if you mean "a simple power supply" in the bench PSU sense, if so, you could simply do what the commercial PSUs do and include an Earth terminal.

[hibone]

First, thank you all for your replies.

I need to measure the power consumption of micros (SoC usually). Usually the supplied voltage is around 3.3V, but the current can range from a few hundreds of mA to a fraction of the uA, depending on the micro.

As such I have to use a sensing resistor, paired with an instrumentation amplifier (e.g. AD621). Since the output of the IA goes straight to an oscilloscope, I have to shield noise pick ups while avoiding ground loops.

On purpose, I'd like to put together a power supply to power up the IA and, if possible, the micro controller as well. Thus I am avoiding switching converters, to keep the design on the simple side and to limit the generated noise.

Although I thought about putting a resistor on the DC side towards the earth, I wonder if noise can crawl up through the ground line and mess with the sensing. The same goes with everything comes to my mind.       

If possible, since I don't mean to build anything fancy such as a bench power supply, but rather a specialized tool, I'd like to avoid a battery powered solution and it would be great to keep the budget low, thus I am asking the forum.

[Pjotr]

If you want measure in the uA range (and below) simply use a battery. Your DUT will be powered normally by a battery anyway then I guess?

[Gyro]

A bit sensitive then 

What about a float charged SLA battery (edit: followed by a linear reg) where you only isolate the charger (both leads) with a relay for the duration of the test. Ok, it's a battery, but at least you don't have to keep replacing it.

[hibone]

The problem with the battery is the voltage. I need about 30-40 V to supply the AD621. I'd have to use 4 transistor (9V) batteries as the bare minimum.

Sheesh, Bob Pease in "What's All This Femtoampere Stuff, Anyhow?" made it look quite simple. His concern was about cosmic rays, after all... 

I wonder what he did use to supply his femptoampere amplifer

[dom0]

Probably a LM317 taped to a wall wart ...?

[hibone]

Probably a LM317 taped to a wall wart ...?

I understand your sarcasm, but, nonetheless I am wondering if there is some sub-optimal solution to the problem. Since I don't have to record femto Ampere currents, maybe a cheaper, although less performing solution (with respect to shielded transformers) may be viable in my case.

For instance, a balun on the output may limit the noise and solve the issue. Some form of active guard may be possible as well, thus I am asking. 

[dom0]

His case is very special in that there is no external DUT: The circuit is a closed system with no analog connections to the outside world (he mentions a connection to a computer, so there is probably an ADC right there in the circuit, not shown, digitizing the high-level output of the circuit). Things like AC leakage into the circuit's ground don't matter. It really wouldn't surprise me if a wall-wart and a 317 is good enough to power this, although they wouldn't bother with that in a lab and just use any lab supply.

The situation you are in and the situation in the circuit described by him are just very different.

[RobK_NL]

Usually the supplied voltage is around 3.3V

The problem with the battery is the voltage. I need about 30-40 V to supply the AD621.

These two statements simply don't add up. Why would you need to supply your IA (which will happily run on 5V) with 30V if the circuit you're measuring runs on 3.3V?

[Gyro]

The power supply range for the AD621 is +/-2V3 to +/-18V. I don't understand why it needs a 30V supply voltage for measuring the supply current of a 3V3 device either. 

[Ian.M]

There's always the direct cancellation approach:  Use another small transformer (lowest VA possible) to provide a voltage opposite phase to the incoming mains, and of similar amplitude.  Couple that to output 0V via a carefully selected class Y1 capacitor to cancel as much of the stray voltage as possible.

[hibone]

---cut---
The situation you are in and the situation in the circuit described by him are just very different.

Thank you for your insight.

These two statements simply don't add up. Why would you need to supply your IA (which will happily run on 5V) with 30V if the circuit you're measuring runs on 3.3V?

The power supply range for the AD621 is +/-2V3 to +/-18V. I don't understand why it needs a 30V supply voltage for measuring the supply current of a 3V3 device either. 

The output swing of the AD621 depends on the power supply. Thus a higher supply voltage lowers the chances that I have to adjust the sensing resistor to the DUT load . Keep in mind that a low power SoC can draw a current spanning from 0.05 uA (sleep mode) to 0.05A (50mA when transmitting/receiving), and usually at full load (without transmitting) requires about 500 uA.       
 

There's always the direct cancellation approach:  Use another small transformer (lowest VA possible) to provide a voltage opposite phase to the incoming mains, and of similar amplitude.  Couple that to output 0V via a carefully selected class Y1 capacitor to cancel as much of the stray voltage as possible.

That's nice! Thank you for your suggestion.

[Kleinstein]

For cancelaton one should not use a very small transfomer: the small ones (e.g. 1 VA) produce quite some distorions and phase shift, as they are far from ideal. So use something in the 5-10 VA range. The no load power is often lower for a 5 VA transformer than for a cheap 1 VA one.


Just on the µA range, AC coupling is not yet so critical. Just make shure there is a good return path to the scope, so that coupling current flows to the scope ground.

I would consider using a transimpedance amplifier instead of shunt and INA - this can give less voltage drop (need an isolationg impedance) and better dynamik range, as the voltage at the current sensing resistor is in the 1 V range instead of 0.1 V. With 50 mA peak you likely need an extra transistor for more power.

[hibone]

For cancelaton one should not use a very small transfomer: the small ones (e.g. 1 VA) produce quite some distorions and phase shift, as they are far from ideal. So use something in the 5-10 VA range. The no load power is often lower for a 5 VA transformer than for a cheap 1 VA one.

I'll do that.
Also, since I have a few toroids I'll give a shot to a common mode choke as well.

Just on the µA range, AC coupling is not yet so critical. Just make shure there is a good return path to the scope, so that coupling current flows to the scope ground.

Allright. 

I would consider using a transimpedance amplifier instead of shunt and INA - this can give less voltage drop (need an isolationg impedance) and better dynamik range, as the voltage at the current sensing resistor is in the 1 V range instead of 0.1 V. With 50 mA peak you likely need an extra transistor for more power.

I am considering that, as well, although I'm worried that calibrating the thing won't be all that easy.

Thank you very much.

[Ian.M]

I'd start by disconnecting the transformer, shorting all the primary connection together and all the secondary ones together separately and putting it on a capacitance meter to get a rough idea of the inter-winding capacitance.  Try either half that driven with the full antiphase voltage for the cancellation cap or that value and half the voltage.    If you cant find a Y1 cap small enough, use a lower cancellation transformer voltage..   For initial experiments drive a 1:1 isolating transformer off a Variac to provide the antiphase voltage.

Also, try reversing the main transformer primary connections - you want the neutral end of the winding nearest to the secondary.  That alone may reduce the unwanted coupling by up to an order of magnitude.

[hibone]

I'd start by disconnecting the transformer [...]

I'll give it a shot as soon as possible.

Thank you all for your suggestions and expertise!

[T3sl4co1l]

Filter the inputs to your amplifier.  More than just AC line frequency will be present, and high frequencies are even harder to deal with.

Tim

[sarepairman2]

Why would these methods be used rather then just using a LDO and possibly some LC filter?

I have never see anything strange like active phase cancellation in any equipment... is this an old technique? Something appropriate for high power use (i.e. electrical distribution)?

I can see the electrostatic shield reducing design requirements... I have only seen it specifically called for in measuring nanovolt noise in a jim williams app note. But I think its more along the lines of him having it handy and deciding the pound the ruble and utilize the equipment he had rather then designing something else/trying to get management to sign off for more parts etc

of course medical transformers have it.. and I suppose having a shield reduces the electrical stress on other components in the design

In my head it seems that common mode/differential mode filter consisting of CM and DM inductors/X-Y capacitors is the best solution, preceding the transformer. and possibly other high frequency specific techniques (tapered inductors, ferrite) to prevent weird microwave shit..

are the solutions proposed in this thread "budget" solutions? why allow all this shit to enter into your transformer in the first place?? especially for a op amp supply! i would not consider these techniques unless I was trying to shield a washing machine!!

[Ian.M]

Various sorts of hum cancellation were used in old mains powered valve gear e.g a preset would be put across the heater winding with its wiper voa a cap to earth and adjusted to minimise capacitive coupling to the indirectly heated cathodes.

All the solutions proposed in this topic are severe kludges, apart from replacing the transformer with one that has a proper inter-winding screen.  The only reason to use them is when you need low common mode hum without drastically increasing the capacitance to earth, you *really* want to avoid using any switching regulators and you cant afford a screened transformer   Of course, before you bother with them, it would be worth checking what the scope picks up with a wide band E field probe, just lying on the bench!

Me?  I'd just use a 2 cell LiPO pack and a linear regulator!

[sarepairman2]

what is wrong with earth ground capacitance prior to the transformer?

[dom0]

Why would these methods be used rather then just using a LDO and possibly some LC filter?

I have never see anything strange like active phase cancellation in any equipment... is this an old technique? Something appropriate for high power use (i.e. electrical distribution)?

I can see the electrostatic shield reducing design requirements... I have only seen it specifically called for in measuring nanovolt noise in a jim williams app note. But I think its more along the lines of him having it handy and deciding the pound the ruble and utilize the equipment he had rather then designing something else/trying to get management to sign off for more parts etc

of course medical transformers have it.. and I suppose having a shield reduces the electrical stress on other components in the design

In my head it seems that common mode/differential mode filter consisting of CM and DM inductors/X-Y capacitors is the best solution, preceding the transformer. and possibly other high frequency specific techniques (tapered inductors, ferrite) to prevent weird microwave shit..

are the solutions proposed in this thread "budget" solutions? why allow all this shit to enter into your transformer in the first place?? especially for a op amp supply! i would not consider these techniques unless I was trying to shield a washing machine!!

This is about 50 (n*50) Hz common mode noise, not about RF noise coming from mains.