Hi Terry,
I read the app-note and can appreciate your suggestion attempting to minimize the phase difference between voltage and current hence PF = 1 which can only be considered during the charge cycle when the diodes actually turn on.
Then again in my opinion that load would seem very capacitive or near inductive depending on the choice of the reservoir capacitor size, series resistance and series inductance as well as the complex signal being applied to the load.
We are back at the start - what capacitor? Or are you suggesting that this methodology could make the choice of capacitor size virtually irrelevant? Now that may be interesting.
I read the app-note and can appreciate your suggestion attempting to minimize the phase difference between voltage and current hence PF = 1 which can only be considered during the charge cycle when the diodes actually turn on.
Then again in my opinion that load would seem very capacitive or near inductive depending on the choice of the reservoir capacitor size, series resistance and series inductance as well as the complex signal being applied to the load.
We are back at the start - what capacitor? Or are you suggesting that this methodology could make the choice of capacitor size virtually irrelevant? Now that may be interesting.
Last edited:
This all is nonsence, Just use a normal transformer and use a Cap as bigh as you can get, you can place and you can effort.
Hi Nico,
with an ordinary rectifier-capacitor this is correct. But here is where a PFC boost converter is different. as long as the controller as running and the secondary voltage is higher than the rectifier drop (2*Vd), the boost converter is transferring power. When the rectified secondary voltage |Vs_pk*sin(wt| is small the boost converter draws a small current....and when |Vsec| is large it draws a large current such that the envelope of the boost FET current is (nearly) sinusoidal.
the problem with a rectifier-capacitor filter is that because the conduction angle is small (Andrew is quite right, its usually about 10%) the peak current is much higher than you would expect (IOW the VA rating is > the real power rating), so the voltage drop across the transformer leakage inductance is likewise larger.
making the rectifier capacitance larger reduces the 100Hz ripple, thereby decreasing the conduction angle, which raises the peak current even higher - if its say 10%, then doubling the capacitor will roughly halve the conduction time (a sine is pretty flat near the peak), and therefore double the peak input current. This in turn will roughly double the voltage dropped across the transformer leakage inductance, making the LF droop worse. its not nice to the diodes either.
for any particular transformer there is probably an optimum value of rectifier capacitance that minimises both the 100Hz ripple and the LF droop
with an ordinary rectifier-capacitor this is correct. But here is where a PFC boost converter is different. as long as the controller as running and the secondary voltage is higher than the rectifier drop (2*Vd), the boost converter is transferring power. When the rectified secondary voltage |Vs_pk*sin(wt| is small the boost converter draws a small current....and when |Vsec| is large it draws a large current such that the envelope of the boost FET current is (nearly) sinusoidal.
the problem with a rectifier-capacitor filter is that because the conduction angle is small (Andrew is quite right, its usually about 10%) the peak current is much higher than you would expect (IOW the VA rating is > the real power rating), so the voltage drop across the transformer leakage inductance is likewise larger.
making the rectifier capacitance larger reduces the 100Hz ripple, thereby decreasing the conduction angle, which raises the peak current even higher - if its say 10%, then doubling the capacitor will roughly halve the conduction time (a sine is pretty flat near the peak), and therefore double the peak input current. This in turn will roughly double the voltage dropped across the transformer leakage inductance, making the LF droop worse. its not nice to the diodes either.
for any particular transformer there is probably an optimum value of rectifier capacitance that minimises both the 100Hz ripple and the LF droop
Are you serious? 100Farad? For a 1 Watt amp everthink higher then 1000uF is overkill. Yes impedance is important, with high C the low impedance for low frequencies is guaranteed.
For high frequencies put paralel a 100nF low inductance C.
At least we have a "grote mond" that makes sounds that other Europeans can understand
I have not saw the schematic so I cant say anythink about it. If you give the scheme, I can give comments. However I think its just straight, bridge rectifier and huge Capacitor of 100.000uF or so. Thats all.
No problems, Nico, I've still at the job! Tom's transformer model is something I've been dissecting, it's somewhat different to what I've seen before, and I'd like to get a good sense of how it works.
In the meantime Terry has been offering up excellent material in regards to the transformer, there are all sorts of "tricks" out there to squeeze the last bit of performance out of how these parts do their job. But, ultimately there is always a physical limit: a simple way of looking at the first stage of a linear PS is that the transformer is a converter of energy from one form to another, the form in this case being voltage level. So you can never get more energy out than goes in, no perpetual motion machines here! And the VA rating of the transformer, the nominal maximum energy delivered in a period of time is the key parameter here. That rating is not an absolute maximum, if you have a 300VA transformer you could use tricks to force it to deliver 500VA, say, but you would be pretty stupid to do so, because the transformer would be tremendously stressed, it would overheat enormously, hopefully blow its internal fuse, or cook itself to a cinder, or at worse set your house on fire!
At a very simplistic level, a 300VA transformer if perfectly incorporated into a supply could allow the amp to deliver somewhere near 300W continuously, the voltage sag would be engineered away by being super clever in the circuit design. But is it worth it? Probably easier all round just to buy a transformer well beyond what's apparently required, rules of thumb come in nicely now.
As Terry stated, the leakage inductance and winding resistances of the transformer are what makes the voltage sag badly when running at high power. In Tom's transformer the leakage inductance is 1.4mH and secondary winding resistance is 0.29R; if replaced with an ideal transformer the voltage droop disappears. Throwing vast quantities of C at the "problem" doesn't help the continuous power drain issue, the transformer can't refill the caps fast enough, the amp will always start clipping.
But, lots of intelligent C will help ride through the musical peak demands of normal music with minimal twitching, modulation. The trick, as Nico says, is to work out just the right combination to get the job done as well as possible -- good ROI.
Frank
In the meantime Terry has been offering up excellent material in regards to the transformer, there are all sorts of "tricks" out there to squeeze the last bit of performance out of how these parts do their job. But, ultimately there is always a physical limit: a simple way of looking at the first stage of a linear PS is that the transformer is a converter of energy from one form to another, the form in this case being voltage level. So you can never get more energy out than goes in, no perpetual motion machines here! And the VA rating of the transformer, the nominal maximum energy delivered in a period of time is the key parameter here. That rating is not an absolute maximum, if you have a 300VA transformer you could use tricks to force it to deliver 500VA, say, but you would be pretty stupid to do so, because the transformer would be tremendously stressed, it would overheat enormously, hopefully blow its internal fuse, or cook itself to a cinder, or at worse set your house on fire!
At a very simplistic level, a 300VA transformer if perfectly incorporated into a supply could allow the amp to deliver somewhere near 300W continuously, the voltage sag would be engineered away by being super clever in the circuit design. But is it worth it? Probably easier all round just to buy a transformer well beyond what's apparently required, rules of thumb come in nicely now.
As Terry stated, the leakage inductance and winding resistances of the transformer are what makes the voltage sag badly when running at high power. In Tom's transformer the leakage inductance is 1.4mH and secondary winding resistance is 0.29R; if replaced with an ideal transformer the voltage droop disappears. Throwing vast quantities of C at the "problem" doesn't help the continuous power drain issue, the transformer can't refill the caps fast enough, the amp will always start clipping.
But, lots of intelligent C will help ride through the musical peak demands of normal music with minimal twitching, modulation. The trick, as Nico says, is to work out just the right combination to get the job done as well as possible -- good ROI.
Frank
Nicely put Frank! and thanks for extracting the numbers (silly question - how many watts is the nominal load? I know its audio, but from an analytic perspective Pout is much more useful).
1.4mH = 0.440/0.528 Ohms at 50/60Hz, which is 1.5/1.8 times larger than the resistance.
Skin & Proximity Effect:
if the wire diameter is < 5mm we can completely ignore skin & proximity effect, regardless of the number of layers in the xfmr at 50/60Hz - this is because for phi/delta < 0.5 Fr = Rac/Rdc = 1 for any no. of layers (phi = effective conductor thickness = equivalent rectangular conductor with no gaps).
problem is we dont draw sinusoidal current - not by a long shot! skin depth is ~9mm at 50/60Hz, but at the (predominantly odd) harmonics it drops to:
3rd ~ 5mm, 5th ~ 4mm, 7th ~ 3.4mm, 9th ~ 3mm
so as long as the wire diameter is less than say 2mm you can safely assume constant R up to the 9th harmonic. whereas for a really big xfmr with say 4mm wire thats only true up to the 3rd harmonic - above that the resistance will increase - but not proportional to sqrt(Fharmonic) - it will rise faster than that due to proximity effect (the no. of layers in the winding).
I thinks its fairly safe to ignore this completely (unless you're planning on driving say 2R speakers so have a stupidly low DC bus) and assume R is constant.
The voltage dropped across the xfmr resistance & inductance is the vector sum of all the harmonics. the resistive portion is just Ipeak*R, and the inductive portion will look similar but shifted in time (its the derivative, but the current is sinusoidal-ish so the derivative will be much the same shape).
with Xl = 1.5...1.8xR the vector sum is dominated by the inductive drop (the R supplies 20...15% of the total) so a first-order analysis can simply ignore the resistance.
with a 15% conduction duty cycle (1.5ms conduction time @ 50Hz) the crest factor (IEEE def. CF = Ipk/Irms) is 3.53. But the ratio of peak to 1st-harmonic-peak is 5.0. So the voltage drop across the leakage inductance is FIVE TIMES larger than that you would expect given the load WATTAGE.
So an excellent 1000W transformer with 2% leakage will drop 2% volts when running at 1000W, but it will drop 10% when running at 1000VA with D = 0.15. Hmm, I'll see if I cant jack up a pretty curve....
1.4mH = 0.440/0.528 Ohms at 50/60Hz, which is 1.5/1.8 times larger than the resistance.
Skin & Proximity Effect:
if the wire diameter is < 5mm we can completely ignore skin & proximity effect, regardless of the number of layers in the xfmr at 50/60Hz - this is because for phi/delta < 0.5 Fr = Rac/Rdc = 1 for any no. of layers (phi = effective conductor thickness = equivalent rectangular conductor with no gaps).
problem is we dont draw sinusoidal current - not by a long shot! skin depth is ~9mm at 50/60Hz, but at the (predominantly odd) harmonics it drops to:
3rd ~ 5mm, 5th ~ 4mm, 7th ~ 3.4mm, 9th ~ 3mm
so as long as the wire diameter is less than say 2mm you can safely assume constant R up to the 9th harmonic. whereas for a really big xfmr with say 4mm wire thats only true up to the 3rd harmonic - above that the resistance will increase - but not proportional to sqrt(Fharmonic) - it will rise faster than that due to proximity effect (the no. of layers in the winding).
I thinks its fairly safe to ignore this completely (unless you're planning on driving say 2R speakers so have a stupidly low DC bus) and assume R is constant.
The voltage dropped across the xfmr resistance & inductance is the vector sum of all the harmonics. the resistive portion is just Ipeak*R, and the inductive portion will look similar but shifted in time (its the derivative, but the current is sinusoidal-ish so the derivative will be much the same shape).
with Xl = 1.5...1.8xR the vector sum is dominated by the inductive drop (the R supplies 20...15% of the total) so a first-order analysis can simply ignore the resistance.
with a 15% conduction duty cycle (1.5ms conduction time @ 50Hz) the crest factor (IEEE def. CF = Ipk/Irms) is 3.53. But the ratio of peak to 1st-harmonic-peak is 5.0. So the voltage drop across the leakage inductance is FIVE TIMES larger than that you would expect given the load WATTAGE.
So an excellent 1000W transformer with 2% leakage will drop 2% volts when running at 1000W, but it will drop 10% when running at 1000VA with D = 0.15. Hmm, I'll see if I cant jack up a pretty curve....
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.