80+ gold + graphics V/A: some doubts

[javitxi]

Hi!

Looking up at the chart of 80+ gold (90% efficiency) I've come across with some ATX12V PSUs that at first sight are quite spectacular and they are <400W:

NXP
HIPRO
Delta

At second round, if you look into PDFs, in the graphics V and A, some intensity are just too low or has glitches or some not sinusoidal wave form, so: how a PSU rated +80 gold can have such waveforms? NXP example

So then, I think that the best PSU, without taking care of 80+ thing will be one that has no glitches, sinusoidal waveform, and also that has as close as possible voltage and intensity waveforms. For example, a 80+ gold will be Enermax 450W

But now, we are looking for a PSU which all these things. Could it be possible? Yes, you are looking for a Delta PSU. For example:

DPS-235W
DPS-300W
DPS-350W
DPS-400W

So, Delta has OEM PSUs but.....what the heck? They are:

- better power efficiency than Corsair, Enermax, Seasonic, etc
- 12VATX form factor
- voltage and intesity sinusoidal waveforms
- non glitch on waveforms
- almost similar voltage/intesity waveforms
- <400W models with all these specs

I am a corsair VX-450W very happy user, but then, taking in mind all these points, why going to a Corsair, Seasonic, Enermax,...? For this comparision, please don't keep in mind quiet/silence issue (you can always replace the fan/mod the fan)

Many thanks in advance

Javier

[lm]

I'm sorry but I believe that you have completely misunderstood those graphs.

The reference voltage, shown by the blue curve, has its scale on the left and is identical in all those graphs you linked to, representing a perfect sine wave.

The current, shown by the red curve, has its scale on the right and we can clearly see that the different graphs have different scales!

Also, the current shown is on 50% load of each PSU, and those PSUs have different max loads! The higher the load, the higher the current is supposed to be.

Hence, to be able to make any meaningful comparisons between the graphs, you must scale the blue curves vertically so that their peaks are visually at the same height as peak voltage, and only then can you compare their form to the voltage curve. If you do this, you won't see much difference between them.

[speedboxx]

Alot of the gold lower wattage (<300W) psus seem to be manufacturered for OEM use and are nearly impossible to find anywhere for a reasonable price.

[javitxi]

I'm sorry but I believe that you have completely misunderstood those graphs.

Yes, I've misunderstood, sorry

Also, the current shown is on 50% load of each PSU, and those PSUs have different max loads! The higher the load, the higher the current is supposed to be.

So, why then at 50% load on these following two examples, the input current waveforms are not 99% sinusoidal (Hipro 250W) but also on the 2nd it has so many glitches (Enermax 500W)?

Hence, to be able to make any meaningful comparisons between the graphs, you must scale the blue curves vertically so that their peaks are visually at the same height as peak voltage, and only then can you compare their form to the voltage curve. If you do this, you won't see much difference between them.

Ok, thanks. Appart from this last point, if you scale voltage and current input curves and you find that they are the same (proportional voltage drain for same current drain ratio), this fact means that the efficiency graphic is going to be plane, am I correct? So this will be the case of an "ideal" PSU with losses (not 100% efficiency)

Alot of the gold lower wattage (<300W) psus seem to be manufacturered for OEM use and are nearly impossible to find anywhere for a reasonable price.

What about ebay, newegg, etc? I haven't searched but I think you will come across with some online shop it would deliver you the PSU at a nice price.

Many thanks for your replies

[Klusu]

am I correct?

No. Stop worrying, the waveform is not that important. There is nothing wrong with Hipro 250. Enermax 500 does not look good, should not be like this, I wonder what the cause is, still, it is probably good enough.

[lm]

Also, the current shown is on 50% load of each PSU, and those PSUs have different max loads! The higher the load, the higher the current is supposed to be.

So, why then at 50% load on these following two examples, the input current waveforms are not 99% sinusoidal (Hipro 250W) but also on the 2nd it has so many glitches (Enermax 500W)?

Power factor correction (PFC) is the feature in PSUs that is responsible for trying to make the PSU look a like a resistive load, i.e. perfect match between input voltage and input current curves with some scaling factor. The better the PFC implementation, the better the match. Without any PFC, the input current curve could look very different.

Hence, to be able to make any meaningful comparisons between the graphs, you must scale the blue curves vertically so that their peaks are visually at the same height as peak voltage, and only then can you compare their form to the voltage curve. If you do this, you won't see much difference between them.

Ok, thanks. Appart from this last point, if you scale voltage and current input curves and you find that they are the same (proportional voltage drain for same current drain ratio), this fact means that the efficiency graphic is going to be plane, am I correct? So this will be the case of an "ideal" PSU with losses (not 100% efficiency)

No, efficiency diagrams are plotted on a plane where one axis is efficiency and the other is load. These V/A diagrams are from a constant load. We can't really infer any info of one from the other. Also, to be able to calculate efficiency at some load, we must know the output currents too.

However the power factor for this 50% load can be fully determined from the V/A graph. What happens is that more than the necessary power goes from the grid to the PSU and some of it goes back to the grid. This means maximum load of the grid is larger than it could be if there were no such oscillation of power.

As a consumer you usually don't pay for that in your utility bill, except indirectly as the power companies need to build better grids. Companies however usually do pay.

This last paragraph might not be 100% accurate:

As to the form of the efficiency curve on different loads: There's usually some constant loss from the PSU when it is not delivering any power at all, then some amount of added load will increase losses linearly. At higher loads, the PSU runs hotter and creates additional resistance which decreases the efficiency. At lower loads, the base loss of the PSU is so large part of the total loss that it has a large effect on the efficiency curve.

[javitxi]

Many thanks for your replies

@lm many thanks for the info, but I have some doubts/ things that I would like to discuss:

I believed that the PFC on PSUs were designed to correct the AC power factor that is delivered to your house. Also, they try to correct "noise" of the sinusoidal AC waveform to a "true"/pure sinusoidal to then, conver it into DC (for example and a very cheap solution, 4 diodes-bridge to have all the positive cicles and then a R and C in parallel to have almost DC).

I mean by power factor the following: if you pass your signal through Cs you put a= -pi/2 if your signal is v(t)= Acos(wt+a) versus i(t) and the otherway (+pi/2) in Ls (or maybe the other way). Adding the inductance and capacitance of the cables and other things, the energy is delivered in AC to your house. Depending on how many wattage are you paying for and if you are a company/"normal" person, you have a different power factor arriving. Then, I've read some companies put some complex circuits with Ls and Cs to correct this, trying to have v(t) and i(t) synchronous (for that conditions, the power losses are apx 0). By doing this, the electric company reduces the prices to companies (less wattage losses).

Little theory note:

Sinusoidal signals <=> AC, which you can, mathematical speaking, see as a sum of exponential, and exponential is the function for which you put it at the input of a system with a response h(t) [H(jw) if you do Fourier's Transform], your output will be A · exponential, where A is a constant determined by H(jw).

Trivia note talking about AC:

The logo from music group AC/DC guess from what it comes from: alternate current/direct current

However the power factor for this 50% load can be fully determined from the V/A graph. What happens is that more than the necessary power goes from the grid to the PSU and some of it goes back to the grid. This means maximum load of the grid is larger than it could be if there were no such oscillation of power.

For the little I know about fuses and noise (electrical) on the grid, the less glithces v(t) and i(t) inputs of your PSU of any electrical device, the better for the grid: you add less noise to the grid which means less noise when powering your devices. When the fuse is broken is to prevent this energy from the PSU to pass to the grid and then, interfere with your other "electrical feeded" devices. So then, if there is no fluctuation on the PSU of your devices the better it would be for this reason.

Furthermore, the more fluctuation is given back to the grid, the closer you are to the intervention of the fuses of your house. Maybe if you are not unplugged from the electrical grid, I don't really bother but, what about people depending on other sources (like solar panels i.e.)?

Also, I though if your equipment draws not the same current proportionally to voltage, it is because is not very efficient (like Enermax 500). Has this really nothing to do with efficiency?

This last paragraph might not be 100% accurate:

As to the form of the efficiency curve on different loads: There's usually some constant loss from the PSU when it is not delivering any power at all, then some amount of added load will increase losses linearly. At higher loads, the PSU runs hotter and creates additional resistance which decreases the efficiency. At lower loads, the base loss of the PSU is so large part of the total loss that it has a large effect on the efficiency curve.

I know that for R, the non-linear curve starts at 1MHz (so higher frequencies, lesser resistance, at least real one [Z = R + iX]) and also that with hotter temperature the R decrease (not increase). If R decrease at constant voltage working, the current increases. We know that voltage is a relative measure and current is the "real stuff" which has the electrons flowing, so if you put more electrons to your material that it can bare, it will break.

Think about it: why then are you cooling your CPU (i.e)? You have your CPU working at a constant voltage with some current. If you reduce the R of the chip, the current will increase. If the current is so high increased, the chip "pops" Also, think about this "pop" thing with PSUs and time: the electrolite capacitors decrease its Z through time because of voltage (electrons) fluctuation. If your PSU starts to fail with a heavy ussage, probably the electrolite liquid inside the capacitors is not the accurate and then, there is more current delivered to the circuits from what is designed for (V = Z · I)

Note: also with other working temperatures, they might appear some Cs that are not on your theorical models, some properties of the material change, etc

Hope to have expressed myself correctly. I'm trying to improve my English hehe

Javier

[Klusu]

...power factor that is delivered to your house... you have a different power factor arriving... the fuse is broken is to prevent this energy from the PSU to pass to the grid and then, interfere with your other "electrical feeded" devices...
Has this really nothing to do with efficiency?
..the non-linear curve starts at 1MHz (so higher frequencies, lesser resistance, at least real one [Z = R + iX]) and also that with hotter temperature the R decrease (not increase). We know that voltage is a relative measure and current is the "real stuff" which has the electrons flowing...

PF is not delivered. PF is made by your house. The fuse is there to prevent damage if too much energy flows FROM the grid. Really nothing to do. Non-linear? LCR circuits are called "linear". 1MHz? Could be any number, depending on LCR values. Losses can be imagined as a serial resistance (more resistance - more losses) or paralel (less resistance - more losses). Voltage is as real as current. Voltage has the electrons flowing.

[javitxi]

PF is not delivered. PF is made by your house. The fuse is there to prevent damage if too much energy flows FROM the grid. Really nothing to do.

Ok, but I though that the fuse is not only to prevent too much drawed energy, but also to prevent that some amount of energy is delivered back to the grid.

Non-linear? LCR circuits are called "linear". 1MHz?

Please, re-read what I've said. Resistor (and at this point I'm only talking about resistors), the non-linear curve starts at 1MHz, and if you have buyed a resistor of i.e. 1 KOhm with 1/4W of power dissipation, at 1MHz or higher frequencies, its V or I vs frequency curve starts to be not linear marking that at these frequencies, the R of its Z is going down due to some internal capacitance/inductance appearance (generally some internal capacitance due to its construction).

Remember (if you have studied/know some little deep electronics) that every single component in this real world is not linear, but!! we take the range of working frequencies from which the component is linear (understanding linear as its performance with V or I versus frequency)

Also, if you are talking about LTI Systems, electronics circuits are linear on its range of working frequencies but only on these frequencies.

In the example of R, when you are working at frequencies higher or equals than 1MHz, in the resistor appears some internal capacitance, caused by the two films that are linking the resistance film to the two "cables" (I don't know the name in English for the two "legs" of any component: BJT transistors has 3, non adjustable Cs, Rs, Ls has 2, etc).

Furthermore, have you ever designed a Wien Oscillator? If you have done it (if not, please do it and convince yourself), you have checked why on it there is no one x MOhm resistance and also, why you don't go further than 100-500 KHz as its frequency working point? If don't please, check it

Could be any number, depending on LCR values.

Also capacitors and inductances has its working frequencies. Have you asked yourself why there is electrolitic, polymer-based, ceramique, tantalium, etc capacitor? And also, you should know that a cheap filtering solution to the power of your circuit, is by putting a ceramique, polymer and electrolitic capacitance in parallel to filter the power. The Bode-diagram of the combination of the 3, makes a plain (or linear, whichever you call) response.

Losses can be imagined as a serial resistance (more resistance - more losses) or paralel (less resistance - more losses). Voltage is as real as current. Voltage has the electrons flowing.

Ok about losses. Please, take in mind that every single measure in this world that is between two points its a relative measure! How far are you from your house? You have already taken a reference point that is your house!

Anyway, how many voltage do you have between two points of a conductor? Yes, its 0V.... and that means that no I is flowing? Maths don¡t think so [(Z =0) = (V=0) · (I= whatever) ] So, this lead as to the definition in Wikipedia.

Precise modern and historic definitions of voltage exist, but (due to the development of the electron theory of metal conduction in the period 1897 to 1933, and to developments in theoretical surface science from about 1910 to about 1950, particularly the theory of local work function) some older definitions are no longer regarded as strictly correct. This is because they neglect the existence of "chemical" effects and surface effects. A particular lesson from surface science is that, to get consistency and universality, formal definitions must relate to positions or (better) electron states inside conductors.

So, by saying the current "has the real stuff" is to imagine every reading person on this fourm to imagine a high flow of whatever as the current thing, and voltage as a balance in which you put how many electrons are flowing at one point related to another point.

If you haven't convinced yourself, please, read about electromagnetics, difference in potential, electrostatic, etc.

....and there was my "little" opinion

Javier

[lm]

I believed that the PFC on PSUs were designed to correct the AC power factor that is delivered to your house.

No. Power factor is a property of a load (so it can have different values even for the same device). The PFC circuitry in PSUs tries to make this power factor close to ideal (1).

There might be additional circuitry in the PSU to filter out noise coming from the grid, and it could even be that these are not completely separate from the PFC part of the circuit, but it does not change what PFC means.

You can see power factor vs load level graphs on all the PSU review articles on SPCR and most other sites.

I mean by power factor the following: if you pass your signal through Cs you put a= -pi/2 if your signal is v(t)= Acos(wt+a) versus i(t) and the otherway (+pi/2) in Ls (or maybe the other way). Adding the inductance and capacitance of the cables and other things, the energy is delivered in AC to your house. Depending on how many wattage are you paying for and if you are a company/"normal" person, you have a different power factor arriving. Then, I've read some companies put some complex circuits with Ls and Cs to correct this, trying to have v(t) and i(t) synchronous (for that conditions, the power losses are apx 0). By doing this, the electric company reduces the prices to companies (less wattage losses).

The machinery of the company has some power factor when its operating. Again, it's a property of the load, not a property of the grid. The further the factor is away from ideal, the more the company pays for "nothing", as they are billed by their "apparent power", which is just the power going in, even though the true power the devices actually use is smaller. Then it makes sense to add circuits that improve their power factor and make their utility bill smaller.

For the little I know about fuses and noise (electrical) on the grid, the less glithces v(t) and i(t) inputs of your PSU of any electrical device, the better for the grid: you add less noise to the grid which means less noise when powering your devices.

Sure, but doesn't have that much to do with fuses.

When the fuse is broken is to prevent this energy from the PSU to pass to the grid and then, interfere with your other "electrical feeded" devices. So then, if there is no fluctuation on the PSU of your devices the better it would be for this reason.

Fuses are used to prevent catastrophic failures of electrical devices, e.g. a short circuit. The fuses are specified to break on such amount of current which the device won't ever draw during normal operation, to prevent further damage: Overheating, fires, injuries to people. The idea is that the fuse breaks so fast that the damage to the device is minimized.

Furthermore, the more fluctuation is given back to the grid, the closer you are to the intervention of the fuses of your house. Maybe if you are not unplugged from the electrical grid, I don't really bother but, what about people depending on other sources (like solar panels i.e.)?

Correct, and might be a concern for off-grid applications.

Also, I though if your equipment draws not the same current proportionally to voltage, it is because is not very efficient (like Enermax 500). Has this really nothing to do with efficiency?

The connection is very indirect. Imagine a single inductor: In ideal world, it has power factor 0 but is 100% efficient. This counterexample should disprove your claim.

However it's likely that companies who put good quality PFC circuits in their PSUs also make them efficient and vice versa.

the electrolite capacitors decrease its Z through time because of voltage (electrons) fluctuation. If your PSU starts to fail with a heavy ussage, probably the electrolite liquid inside the capacitors is not the accurate and then, there is more current delivered to the circuits from what is designed for (V = Z · I)

A good reason to go for all-solid caps design.

[javitxi]

Many thanks lm for PFC, fuses, etc info.

Everything is clear now

Javier

Offtopic: By the way, have you any info about the new techniques of building solid caps and if they have widen its working frequencies? Have they introduced some nano-techniques/materials?

[Klusu]

..the R of its Z is going down due to some internal capacitance/inductance...

Wrong. R stays the same. And what is "V or I vs frequency curve"?
LCR circuits are linear, meaning: V and I are proportional (not exactly a definition). Let's take your 1 kΩ resistor with some C (0.5pF) between it's ends. At some frequency (300Mhz) it's Z begins to decrease (btw, linearly). Not 1MHz. No need for Fourier etc.
When you say:

...I know that ...

, remember:

I've misunderstood