Power Supply Resevoir Size

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I think that secondary resistance is just measured with an ohm meter.

Attached is an image of my original spice transformer simulation circuit, which contains all of the transformer measurement instructions.

But the much better model that I created, after that, with the help of Terry Given, is SCALABLE, i.e. you can change the output voltage, the Volt-Amps rating, the mains voltage, and the mains frequency, and the model parameters are all scaled automatically to reflect the changes! An image of that LT-Spice model is also attached. (Note: I am not sure how the accuracy varies, as the model is scaled.)

Besides using the model in LT-Spice, I also implemented the scalable transformer model in MS Excel, as part of the power supply simulation I created. If you download the spreadsheet and open it, click on the Transformer tab at the bottom, to get to the transformer model. It, too, includes the transformer schematic and the measurement instructions.

MS Excel 2007 version:
http://www.diyaudio.com/forums/power-supplies/216409-power-supply-resevoir-size-167.html#post3287619

MS Excel 97-2003 version:
http://www.diyaudio.com/forums/power-supplies/216409-power-supply-resevoir-size-166.html#post3282260

Cheers,

Tom
.
 

Attachments

  • Transformer_meas.jpg
    Transformer_meas.jpg
    337.1 KB · Views: 326
  • transformer_meas_pu.jpg
    transformer_meas_pu.jpg
    460.7 KB · Views: 317
I think that secondary resistance is just measured with an ohm meter.
...

Thanks for links.

I asked because I wondered if Rp is taken into account, but that was the version linked to from another thread, and now I'm disoriented :confused:

Anyway, AFAIK the primary resistance should be in series with the primary current, so that it is effectively reflected to the secondary, as is primary capacitance. Measuring the secondary with DC doesn't take this into account, and your model appears wrong in this respect. I might be missing something though.
 
LTspice will easily demonstrate what happens to the mains power waveform, which does become badly distorted across the secondary, when real behaviours of transformers are included.

I use another version of SPICE 3f4, and developed a model some years ago intended for valve output transformers.

The model allows the transformer to be specified in term of physical parameters, and allows cores and windings to be connected via a bus. This was quite educational but ultimately I ran out of steam when I got to hysteresis.

A static BH curve can be approximated but not very well. With considerable difficulty and cheating, hysteresis can be modelled at one frequency. If there's an LTSPICE model that does all this for all frequencies I'd love to see a listing.

The issue is important but often ignored. For example, a commercial general purpose toroid, to take a worst case, will typically be nudging the top end of the BH curve, close to saturation, when lightly loaded. In such a state it produces its maximum mains harmonics and its inductance is furthest from linear. Those who deliberately oversize transformers in pursuit of lower regulation are not necessarily improving matters. There's no substitute for purpose-built quality.

I don't understand why all DIY amps aren't class A, or shunt regulated. Some people get their kicks from saving energy, I guess.

Maybe it's not as significant as I suspect. That's why it would be nice to see a simulation.

P.S. This is related to my other point above about Rp. Because it is in series with the primary inductance and passes all primary current, the voltage across the primary inductance, and hence the magnetising current, drops as secondary current increases.
 
Last edited:
Thanks for links.

I asked because I wondered if Rp is taken into account, but that was the version linked to from another thread, and now I'm disoriented :confused:

Anyway, AFAIK the primary resistance should be in series with the primary current, so that it is effectively reflected to the secondary, as is primary capacitance. Measuring the secondary with DC doesn't take this into account, and your model appears wrong in this respect. I might be missing something though.


All models are wrong. It's just a matter of how wrong they are.

Did you also see all of the AC measurements that were made?

Anyway, it's not my model. There is a link to an OnSemi appnote PDF file, right on the schematic. Christophe Basso justifies the model fairly well, in the appnote, in my opinion.
 
All models are wrong. It's just a matter of how wrong they are.

Did you also see all of the AC measurements that were made?

Anyway, it's not my model. There is a link to an OnSemi appnote PDF file, right on the schematic. Christophe Basso justifies the model fairly well, in the appnote, in my opinion.

Models can be close enough if used intelligently, for the intended purpose. They can be very wrong otherwise.

Regulation is an important consideration in your use of the model. If it fails to take into account primary resistance, then regulation will be underestimated. It's your job as the model user to check whether the difference is significant.

Just trying to assist your understanding.
 
Thanks. I do understand the limitations of engineering modeling and simulation.

The preparation of the model was with the assistance of Terry Given, who is an experienced professional magnetics expert.

The model includes the primary and secondary resistances and leakage inductances, and the primary magnetizing inductance.

I believe that it is sufficiently accurate for simulation of linear power supplies for audio power amplifiers, as long as it isn't scaled too far away from the initially-measured parameters.

But it is always the user's responsibility to validate and verify any model they choose to use.
 
Last edited:
Member
Joined 2004
Paid Member
Plastic-
For the denser of us, what exactly is wrong with the model? What is the scale of error the "incorrect" model will create? While we are all content to accept Ohm's law as an absolute most of the simulation work is not totally accurate and in many cases a simplification enables meaningful answers quickly. Most electronic systems can be reduced to problems that can be solved with a slide rule and yield working systems.

In this case small differences in the available supply voltage before clipping from different models is a minor issue. The AC supply can easily vary more than 15% regardless of the load. I don't know if the model adjusts for the heat in the transformer or the change in drop on the diodes with heat. But I suspect the difference in max output from the details in the model will be at or below the threshold of hearing.
 
Plastic-
For the denser of us, what exactly is wrong with the model? What is the scale of error the "incorrect" model will create? While we are all content to accept Ohm's law as an absolute most of the simulation work is not totally accurate and in many cases a simplification enables meaningful answers quickly. Most electronic systems can be reduced to problems that can be solved with a slide rule and yield working systems.

In this case small differences in the available supply voltage before clipping from different models is a minor issue. The AC supply can easily vary more than 15% regardless of the load. I don't know if the model adjusts for the heat in the transformer or the change in drop on the diodes with heat. But I suspect the difference in max output from the details in the model will be at or below the threshold of hearing.

My point is simply that the primary winding resistance should be in series with the magnetising current, and as far as I can see in gootee's simulations, it isn't. The significance of this depends on judgement and the particular application of the model, as always.

I have tried to explain the implications. Being not especially knowledgeable or clever, I may not have been clear. I might have missed something in the model so I might not be entirely correct. I just spotted what seems like a mistake and wished to engage with the issue, thanks.

The application to which the model is put here is related to regulation, i.e. the extent to which loading reduces voltage. That in turn relates to winding resistance. The effective AC resistance of a transformer winding is not just the static resistance of that winding, as might be measured at DC with an ohmmeter, but also includes the reflected resistances of the other windings, in parallel or series as appropriate.

For a simple mains transformer, assuming mains resistance to be insignificant, the effective resistance of the secondary is equal to

Rs + Rp/A²

where A is the turns ratio. The model in question looks to me like it would give just Rs. Numerically the difference may or may not be significant, but conceptually it is of crucial importance to understanding how transformers work.

Intuitively, considering a transformer is reversible, a model would treat Rp and Rs in exactly the same way, but the model in question treats them very differently. That may work for some purposes but not others, and certainly confuses the issue conceptually.

The objection more serious to my mind is that the mistake promulgates the myth that bigger is better when specifying transformers, and possibly that toroids are better than EI, simply because regulation is lower, so there is less droop at peak demand. This ignores the perils of under-loading.

A commercial general-purpose mains transformer is likely to be close to saturation when unloaded or lightly loaded. This is especially significant for toroids because of their relatively sharp saturation characteristics. If mains voltage is higher than rated, the condition is exacerbated. This is not generally important because the rated load current will move its operation along the BH curve to its linear region.

One consequence is that an under-loaded transformer is likely to distort more than one at its rated load. Distortion results in mains harmonics which, as they get higher in frequency, may be less well removed by subsequent filtering and more intrusive into the audio. Class A, or shunt regulation, removes the problem by drawing a more constant current from the transformer. The simulation in question however concentrates on the effects of varying current demand.

Now, the mechanism that moves the operation of the transformer from its saturation to its linear region is the voltage drop across the primary inductance, produced by the primary resistance as a result of current drawn from the secondary. Put another way, magnetising current is reduced by the primary resistance in proportion to current drawn from the secondary. If the model does not place primary resistance in series with magnetising current, then this effect disappears.

In summary, the model underestimates regulation and is conceptually misleading. I think both are significant, but others may think otherwise. Genuine engagement with the issue is welcome.

On the more general issue, a good audio mains transformer should always operate in the more linear region of the BH curve. This it achieved not just by being bigger, but by having a different ratio of iron to copper.

Finally, the simplest version of my model was aired here. Note that the winding model is universal...regardless of whether primary or secondary. It failed to illicit a response, other than directing me to LTSICE current practice, which is actually cloaked in mystery AFAICS, with the originators long gone.
 
In addition to the current entering the primary because of the secondary load, there is the core exciting current which flows in the primary whether the secondary load is connected or not.
This current is drawn by the primary core reactance and the equivalent core-loss resistance and is multiplied by N1/N2 when it is referred to the secondary side.
It has two components, the magnetizing component which flows 90 degrees lagging behind the induced voltage and the core-loss current which is in phase with the induced voltage.
Ordinarily this current is small and produces negligible voltage drop in the winding.
- Electronic Transformers and Circuits Reuben Lee

The statement from the book seems contradictory, on the one hand saying magnetizing current flows whether loaded or not but it is also multiplied when referred to secondary. So is this basically saying magnetizing current is unrelated to the load completely.
 
isn't magnetising current a very small portion of the load current? does it really matter?

i agree on the size of the traffo though...

Magnetising current is at its maximum when load current is zero, and minimum when fully loaded, so its proportion varies enormously. For consideration of distortion, it's not really the proportion that counts, but rather its absolute value, because that determines where on the BH curve the iron is. The risk with a big transformer providing current for an amp running with low idle current is that the quiet passages will suffer most from mains harmonics.

The significance of the regulation error can be guesstimated perhaps. Imagine for my convenience a transformer with 240V primary, 24V secondary, and a secondary resistance of 1 ohm. What is the primary resistance likely to be? There'll be ten times the turns of wire, perhaps of a quarter the x-section area, which comes out at 40 ohms. Divide that by the square of the turns ratio and you get 0.4 ohms. Effective secondary resistance is then 1.4 ohms, which to me seems significantly different from 1 ohm. Worth measuring, I would say. Needs to be in any mains transformer model, surely?

I should continue to stress, since the originator of the SPICE model isn't here, that I might have missed something.
 
Plastic,

The Basso/OnSemi model that I used is correct, as far as it goes, even if it is difficult for some to understand.

The reference to the paper from which it came was given. Here it is again, in ready-to-click link form:

http://www.onsemi.com/pub_link/Collateral/AN1679-D.PDF

The complete model was also posted, for anyone to simulate, if there were doubts.

If you have a model that also models the core, et al, then why not test both models with the power amplifier that was also modeled, earlier in this thread, to see if there are any significant differences, and let everyone know the results?
 
Questions about: Decoupling forced to work well via diode isolator approach.

I've got a whopper of a question!
Data:
In this application (scroll down past the chat to see photos) http://www.diyaudio.com/forums/chip...e-no-lossy-emitter-resistors.html#post3495575
MBR1645 is enough forward voltage drop for 220uF or less, but not more
MUR, MR or FR is enough forward voltage drop for 440uF or less, but not more
10A05 is enough forward voltage drop for 660uF or less, but not more
*above examples work somewhat better if using either less capacitance or more voltage drop than the maximums given--for example 6A05 and 440uF works thoroughly, in practice.

First question:
Proportionately to the above examples, how much amplifier board capacitance is then supported by the voltage drop of KBPC3502's (one series each rail, and used as if a series pair standard diodes for voltage drop)?

Second question:
How much voltage drop is needed to support this same application if 15,000uF is used?
It would be a huge amount of voltage, but a percentage might be even more interesting.
 
It would be very nice if I could get a copy of Chan's paper without joining Yahoo. Otherwise I'll have to wait until I go to the library to download it. If it is just a reworking of Hymowitz, it may not be worth using.

This paper:

http://www.onsemi.com/pub_link/Collateral/AN1679-D.PDF

was intended for modelling transformers in switch-mode power supplies. Further, its main interest is leakage inductance. An SMPS transformer typically has windings of very low resistance. Leakage is much more important. They work at high frequencies, and core characteristics are very different from mains transformers.

Early in the ramble, the author puts the primary resistance in series with the primary, but later switches it to parallel with no explanation of why, AFAICS. He gets the behaviour near enough, perhaps, because the regulation due to resistance is negligible, and the relationship between Cp, Ll, and Rp is maintained, so filter characteristics may be correct or near enough.

As it stands, it is not appropriate for modelling a mains transformer, where for nearly all purposes leakage is insignificant and resistance is important. It has very limited value as a design tool even for SMPS, because flux is crucial there and it isn't included.

I have to admit I regret binning my model. It's much better and works for everything because it simply arranges the transformer equations into a computational model. If other SPICE engines can do recursion like mine can (I was surprised and amazed, just trying it for amusement) then I recommend it. It allows flux to be graphed easily, so you can see if it's too great even if you can't see the consequences because the BH curve is linear.

The tan function used in other models for the BH curve can be easily adopted, and I dare say Chan's work could be built in too. I doubt either approximation would be useful, however, outside of narrow limits. We are interested in the details of distortion, and a generalised approximation is of no value IMHO.

I binned it because simulation has very little value in designing a linear power supply. I prefer to copy best practice and tweak to suit.
 
Magnetising current is at its maximum when load current is zero, and minimum when fully loaded, so its proportion varies enormously. For consideration of distortion, it's not really the proportion that counts, but rather its absolute value, because that determines where on the BH curve the iron is. The risk with a big transformer providing current for an amp running with low idle current is that the quiet passages will suffer most from mains harmonics.

The significance of the regulation error can be guesstimated perhaps. Imagine for my convenience a transformer with 240V primary, 24V secondary, and a secondary resistance of 1 ohm. What is the primary resistance likely to be? There'll be ten times the turns of wire, perhaps of a quarter the x-section area, which comes out at 40 ohms. Divide that by the square of the turns ratio and you get 0.4 ohms. Effective secondary resistance is then 1.4 ohms, which to me seems significantly different from 1 ohm. Worth measuring, I would say. Needs to be in any mains transformer model, surely?

I should continue to stress, since the originator of the SPICE model isn't here, that I might have missed something.


i understand what you are saying but.....
isn't the % regulation spec of a transformer supposed to cover this?
do we really need to be this precise or meticulous?
can things be as simple as possible?
after all readers who are also diy'ers are not going to be interested in the
pure theory of things at the mathematical levels,
imho they will want to build practical amps that work without
being saddled with maths....;)
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.