Rerurn to Romy the Cat's Site

Dear Romy, regarding your efforts to solve variations in mains power supply, there has been an interesting thread on RAT related to the use of toroids as power transformers that morphed into a discussion about the level of DC offset in mains power supply and possible solutions.  I thought you might be interested in the gist of the discussion because it contained much useful information, as follows.  Regards, Peter Foster.

---

Mains are polluted everywhere. Typical DC offsets are 0.8 to more than 3V (!) in Europe. I guess that North America is as bad - or worse.

---

So, aside from hair dryers on 'half power' what causes it?

** Asymmetrical loads on the supply - dummy. Don't much care what they are. Just depends what is going on around where you live.

---

DC offset is a big problem nowadays, with more and more switching power supplies w/o proper power correction in every home (it's even worse in offices). As toroids work very close to core saturation, even a small amount of DC offset wreaks havoc.

---

Mains DC offset exists in all countries - only a matter of luck of your local AC supply has no DC offset. The effect is dramatic on a toroidal power transformer.

Example: 300 VA, 240 volt 50Hz AC toroid (unloaded).

#1 Offset free supply:
I mag = 15mA rms / 25mA peaks.

#2 Supply with 1.2 volts DC offset:
I mag = 715 mA rms / 3.9 A peaks.

In case 2, the transformer is heavily saturating on one half of each AC cycle PLUS is drawing 171 VA - while OFF LOAD !!!!!!!! BTW 1.2 volt DC offset created by using a Black and Decker 1600watt ( handyman) hot air gun, set to half power.

---

Phil's heat gun set to half power could be a problem, as it likely uses a rectifier diode to conduct on say positive peaks of the powerline only.

** Correct.

And if your house wiring has enough Thevenin equivalent impedance (American powerline 15A outlets are around 1/4 ohm)

** That figure is around 0.5 to 1 ohm in Australian ( 240V ) homes.

Running the hi-fi on its own dedicated branch circuit from its own circuit breaker in the breaker panel greatly reduces that offset from that heat gun. There the powerline impedance is more like 1/40 ohm, for a 150 amp service. Or maybe 1/20 ohm if your power company cheaped out on wire size...

** Some folk have around 0.5 volts DC offset most of the time.

---

Dunno if there is any way to complain about it to the energy suppliers.

** There isn't because it isn't part of the tariff to have no DC offset. At least in the US, the tariff-the statutory rate you pay the utility they agree to in exchange for the monopoly on power-says what you are and are not entitled to.

---

Mains dc offset isn't a stable thing, and the dc offset is really just slow moving LF content. You'd see what LF shite there is in your mains supply by having a very steep cut LPF with a pole at say 20Hz to remove easily visible 50Hz and harmonics. An LCLCLCLC type filter will do it.

The mains is a noise source due to thousands of users all around you switching things off and on, and with so much continual switching, the result is a mains voltage that rises and falls with its varying shared load. There is a resultant content in all this noise which can be filtered out, and if you had a filter resonant at 0.1Hz, you would see a varying signal at this LF. It is so slow a signal wave that it acts like varying DC, but dc that varies is NOT dc anymore, eh.

The average load at your house is determined by the generator and wiring from generators to you, and all the transformed down high Z of the network. But at any given time, there are so many folks drawing power that there is the equivalent of a giant low value resistor shunting the mains supply. If the average input current is 4A to your house, the load of your house = 240V / 4 = 60 ohms. 4A might seem a lot, but 2 ppl in a house average around 1000 watts 24/7/365.

1,000 houses sharing the same mains wiring gives a load of 0.06 ohms, I = 4,000 amps. Noise generated across town won't appear near you because the wire resistance across town is considerable, and switching transients get smoothed out by inductive nature of the wires. With the shared R load being so low, the amount of L needed to attenuate HF isn't much. However, because the mains are effectively a huge informal antenna array, there is a largish RF content present.

If a negative swing lasting an hour before its swings the other way, don't ask the utility company for a refund. Every silly question has already been heard across the counters of power companies... Just wait until we all switch to electric cars and nuclear power stations. We will be seeing much more work for electrons to do.

---

This company in the Nederlands has a box incorporating the ( well known ) solution to the DC offset problem.

http://www.kempelektroniks.com/PowerDCXTerminator.aspx

Make sure to click on the CRO screen to see a neat simulated demo.

---

The above link raises the question, how is DC offset reduction achieved in the 1.2 kg item offered?

** By simply AC coupling the PT primary to the AC supply. The weight must be all in the (lead) alloy box.

---

Do they just have a pair of large value series capacitors in series with mains active and neutral?

** I would say back to back caps shunted by a string of diodes. That's the usual way to block DC.

---

But does not an ordinary isolation transformer remove the dc at the sec?

** Of course it does. But if you run a toroid isolation transformer close to saturation, you've just moved the problem.

---

The above link raises the question, how is DC offset reduction achieved in the 1.2Kg item offered? Do they just have a pair of large value series capacitors in series with mains active and neutral?

If there is a varying level of dc, it really means you have an added LF wave between active and neutral. The neutral is at ground potential as it is tied to earth at the house earth connection. Hence the jittery level of the mains wave shown on the CRO.

But does not an ordinary isolation transformer remove the dc at the sec? And in any case, the very low F "dc offset" is small and should not have a serious effect on the maximum Bmax of the PT.

But mains voltage does change in amplitude from minute to minute, and raw Vdc after rectifier diodes changes continually, so depending on mains 50Hz voltage variations, and music caused dc load variations on the PSU from largely class B amps, the transformer and rectifiers are seeing a varying load at all times.

Transformers should be designed to take a 10% extra input voltage indefinately. I forget the exact words of the Australian Codes, but imported stuff is supposed to comply.

Many trannies would hum quite badly and perhaps thermal out if they were forced to accept 264Vrms instead of the usual 240Vrms we are supposed to have.

Chinese made trannies designed for 220Vrms don't fare well, as do US mades with 230V connection but rated for 60Hz.
And here we have 250Vrms measured most days. It isn't a nice sine wave like the one shown in the CRO pic at the above URL. What we have here is a triangular wave with flat crests which have a total length of about 1/4 of the whole wave cycle lengh. It means the equivalent of maybe 10% odd order harmonic distortion.

And BTW, I loath toroidal trannies because most are so poorly made. Good ones are well varnished internally, potted, and don't hum because sufficiently high TPV are used to keep Bmax less than 1 Tesla, and thus be able to pass tests for over-voltage without saturation effects making them noisy or hot.

These good toroids are of course larger and heavier than what we see in the crummy budget audio gear which is 95% of what is foisted onto the public. Makers don't like paying the costs when once you get above a certain quality criteria, welded E&I cored PTs are the better thing for cost and performance.

---

The majority of commercially designed transformers are designed to run with Bmax at whatever figure gives the wanted rise in T, and with about 3A per square mm for current rating in the wire. Toroidals are run at maybe 1.3Tesla, and right on the onset of saturation.

Current spikes can be seen in the wave form with no load. So if B is increased 10% with a 10% mains voltage rise, the spikes are larger. However, the core material can be such that it still won't get hot.

With load, the current wave is dominated by the load current, and the spikes due to saturation effects are not so obvious.

Presence of DC, say 1V would to me seem like a heck of a lot of DC and very unusual for mains supplies, but if present, the dc current through a low resistance P winding could cause Fe magnetization and more serious saturation current spikes. But any other E&I transformer could also suffer similarly, depending how close to saturation is it running. But because so many mains users have transformers connected to the mains at any given time, and because the resistance between active and neutral is so low, the presence of Vdc generated by stray devices connected tends to be very low.

The simple way to make sure any tranny stays cool, does not hum mechanically, and can sustain some DC and not hum badly with a rectifier on the sec feeding large C with Si diodes is to design the tranny with B < 0.9Tesla, which means most makers would have to increase copper turns about 30%. Bean counters alweays say farkin no to this, and to get them out of their silly mindset, should have a decent sized E&I shoved where the sun don't shine. If the turns are increased, there has to be more room for them, and the wire must be thicker to maintain copper resistance and the copper losses. OR the Afe core size must be increased 30% which reduces B 30% for the same turns per volt.

There majority of toroids available are wound as a cheap lightweight alternatives to old fashioned E&I trannies, but I am very reluctant to use any of them because they all have Bmax far too high, and although they rarely get hot, they can be noisy when rectifier output is needed.

If they are designed with B < 0.9 Tesla, the weight/size and cheapness advantages dissappear because such a toroid is a special order.

The right way for a toroid to be wound is with finely woven cloth insulation material placed over the core to prevent sharp edges biting the wires, and then all layers have no overlapping turns, and have a layer of insulation tape wound on between consequent layers of wires. Once completed, the transformer is soaked overnight in varnish, and drained of excess then baked to cure the varnish to make good solid bond between core and layers of wire. The woven material allows the wet varnish to soak into the tranny; plain polyester film does not.

Proper construction as described is never ever done with budget toroids bought by furquit hobbyists trying always to avoid costs.

I'd always use toroids if I could find a source of cheap decently constructed types, but none are to be found whose cost is less than good E&I.

The design spec of say 0.85 Tesla will allow a mains tranny made for 220V 60Hz to be used with 255V and 50Hz, where the B would then be 1.11 Tesla. If designed initially for 1.3T, an adverse condition could give 1.8T, and serious saturation problems.

Audio gear should never have Bmax in power trannies higher than 0.85 Tesla. Bean counters never agree though. Hence all the shit we see that I will NEVER buy.

Chinese electrical items often seem to be rated for 220V only, and suffer badly with the 255V we see hear sometimes.. Chinese soldering irons get way too hot here and last only a month if you are lucky and need a box with resistors inside to drop the voltage at the iron. Then they last more than 18 months.

Some chinese amps don't like the 255V applied to the power trannies, because the B+ rises too high, but most have cheap E&I cored trannies which appear to take the missmatch of mains V OK.

---

Toroids don't have an air gap, even the most carefully stacked E-I laminations have a residual air gap, are you telling me that doesn't change their characteristics?

Indeed the toroid spiral strip core does have a different characteristic to fine and well stacked E&I lams. Not only is there a microscopic air gap in E&I lams, or C-cores, but there is a change in grain direction of the Is and Es where they butt together.

C-cores don't have such a grain direction change. But they do have a cut and join. The effect of the butt join with E&I and cut and polished join in C-cores reduces the maximum µ available. With the GOSS toroid core, perhaps max µ = 40,000, and if the same material is used in E&I lams perhaps max µ = 17,000, and with C-cores could give perhaps above the 17,000.

NOSS material is non oriented grain steel with Si content but is never usually used for toroids, although it could be, not used for C-cores, but is used in cheap E&I cores and has a max µ when stacked close of around 3,500. NOSS if used for a toroid might lift the max µ to 5,000. All these steels are "special", and some more special than others, but they behave differently depending on the app. The frequency of saturation is only very slightly higher with GOSS than NOSS, but GOSS saturates more sharply than the NOSS. GOSS cores have much lower magnetizing currents than NOSS, and hence the GOSS runs cooler.

GOSS with high µ is now routinely used in E&I mains transformers where the cores are not interleaved at all but all Es are just butted to all Is, and the two bundles are machine welded. This saves an enormous amount of Joe's time spent stacking in the Es and Is to interleave them fully. People don't like doing the job Joe does; its very boring to spend all day stacking laminations. Some cores get machine stacked. but welded is cheaper. The resulting µ of the welded cores is well below what could be achieved with maximal stacking, but its adequate to get losses to be low enough.

Even ARC use mains trannies with welded cores.

---

As promised yesterday, here are some more measurements as per your suggestions...and some others as well:

The Plitron toroid that I am using for the testing is an 80va unit wound for 20-0-20 volts at 2.0 A rms, and having a single 115 V rms rated primary.

I characterized it first with no DC offset, just to see how close to the limit it is designed in terms of saturation. The meter used for exciting current measurement is a Fluke 87 type III ("true" RMS...it measures only the true RMS of the AC component of the waveform. To obtain the true RMS, you need to obtain the DC component separately, square it, add it to the square of the AC component that the Fluke 87 measures, and then take the square root of the sum.) The exciting current waveform is observed using an HP465A AC only current probe. The peak value of the exciting current is designated Ip.

Voltage adjusted with 1.5 kVA variac:

Vin (60Hz) Imag Ip
volts rms mA rms mA peak

105 8.1 10
110 9.2 12
115 10.6 18
120 13.2 24
125 17.8 50
130 31.5 100

The exciting current waveforms start off at 105 volts in as what look like slow rise time square waves. This is due to the large amount of 3rd harmonic in the Imag. You don't normally see this behaviour in EI transformers as the flux density cannot be run this far into the knee of the BH loop without experiencing unreasonably high core loss. But with all the flux in the toroid being in the "easy" direction of core magnetization, you can really honk the flux right up there without excessive core loss, and that's exactly what Plitorn did with this design.

By the time you reach 115 V the Imag waveform shows some "tailing up" just prior to its falling edge, indicating a closer approach to saturation. Note the relationship between Ip and Imag. At 105 Vin, Ip is only 1.23 Imag. By the time 115 V is reached, Ip is now 1.70 Imag. At 130V in, Ip is 3.17 Imag.

Although opinions (and measurement standards) in the industry may vary as to the exact onset of saturation, there is no doubt by any definition, that I have ever seen published, that we are soon to be in trouble if our input voltage exceeds 130 V. This toroid by the way was sold for use in North America where nominal AC line is typically 120V. It is not unusual to encounter 130V in some areas during light load periods.

Now for the offsets:

Offsets are measured using a low pass filter consisting of a 100H inductor and 20uF polypropylene cap. The voltages are measured with a Beckman DVM.

There is no variac used during these tests as its series impedance in addition to that of the AC line tends to greatly increase the DC offset. We do not want to exaggerate the problem.

1) Measure offset using 40 W, 120 V incandescent bulb with series 1N5406, measure Imag.

Connect the 40 W bulb and diode on same branch circuit. Voffset = 47mV dc
Ioffset = 47mV/3.6 ohms = 13.06mA dc
Imagac (13mAdc offset) = 28.5 mA rms , Ip (offset)=100mA peak

Total Imag = sqrt(Imagac^2 + Ioffset^2)= 31.4mA rms

And remember that from our previous measurements on the toroid:

For 120 V rms in, Imag (no offset) = 13.2 mA rms, Ip (no offset) = 24mA peakSo for the single 40 W bulb, the peak current increases by a factor of 4 and the RMS current by a factor of 2.4.

2) Now let's make this look like a cap input filter load, still with a half wave rectifier and try to keep the load POWER the same as in the simple half wave rectifier and 40W bulb example.

This will require our 1N5406, a filter cap of 1500uF (to keep ripple across the load insignificant) and a load resistor of ~1445 ohms. After building this up, here are the results:

Voffset = 30mVdc, Ioffset = 30mV/3.6ohms = 8.33mA
Imagac (8.33mA offset) = 20mA rms, Ip (offset) = 100mA peak
Total Imag = sqrt(Imagac^2+Ioffset^2) = 21.7 mA rms

At first this does not look like as bad a case as the 40 W bulb and diode, which also looks like a 20W load but with a much longer diode conduction angle. The peak current however in the toroid primary is just as bad, so the threat of saturation is just as serious...particularly if the AC line voltage were to increase to 125 or 130 V under this condition.

If I had a better current probe, like one of the modern Tektronix units, I could better examine all of the peak currents and give some better answers. Using a shunt resistor to measure currents here is not an option as an isolation transformer in the AC line has sufficient impedance to really exaggerate the DC offset, and floating the 'scope is not an option for the sake of safety.

3) Now what about if the DC offset source is on another branch circuit on the same leg of the 120-0-120 V supply?

Using the original 40W bulb setup on another branch circuit, the effects on the toroid were reduced substantially.

Voffset at transformer = 12mV dc, Ioffset = 3.33mA
Imagac(3.33mA offset) = 14.3mA, Ip (offset) = 42mA
Total Imag = 14.7mA

Examine the results in the first table and compare.

4) Hair Driers etc.

I have two hair driers here. One is a less than 2 years old and causes no DC offset on either of its two speeds. The other is a Braun drier that it still in excellent shape that I have had from my university days in the early 70s. It still works well but produces 500mV of offset on the AC line when used on its low speed. Judging by how long mine has lasted, I'll bet there are still quite a few of these out there, owned no doubt by parsimonious old farts like myself.

I also have however, quite a recent model of heat gun, a Balck and Decker Heat and Strip Model 9756. This thing, on its low speed produces 1.2 V dc of offset, causing the Imag of our toroid to exceed 1 amp rms! Even when used on another branch circuit, the Imag in our toroid hits 140mA rms with peaks of 600mA. So asymmetrical loads on the AC line can be a real problem if a toroidal transformers is used on the same branch circuit. Stating that such loads don't exist, or should never exist, is just wishful thinking.

---

Rerurn to Romy the Cat's Site