Ok, I will try , soon.
Those will be my output coils for testing, 30uH each.
2xR50 cores winded by 2x200x0.04 litz in parralel.
Hope those output coils will not saturate so easy...
Your sincerely opinions about it?
And with one IRFB4227 (4 for full bridge), how manny watts is possibly to achieve without destroying it?
(overcurrent)
Those will be my output coils for testing, 30uH each.
2xR50 cores winded by 2x200x0.04 litz in parralel.
Hope those output coils will not saturate so easy...
Your sincerely opinions about it?
And with one IRFB4227 (4 for full bridge), how manny watts is possibly to achieve without destroying it?
(overcurrent)
DjLeco!
No. I*I*L=constant, so if you want 2 times more current, then you get 1/4 inductance.
In the amplifier you can check saturation by seeing carrier residual, but the amplifier probably won't survive if you drive it to 80-100 A output for a long period. This problem can be overcome 2 ways:
- you test the core with significantly higher number of turns, but with much smaller filter capacitance. (eg. 4 timer more turns, and 220 nF capacitor.) Then you can calculate it back to the original number of turn. (For example: 40turns*10A equals to 10turns*40A)
- you can apply narrow impulses to the amplifier. (Attached file.)
You must use dummy load, preferably a low resistance one. When the residual starts to increase, the core saturates.
...
That R and C doesn't provide dead time. You need a diode too!
No. I*I*L=constant, so if you want 2 times more current, then you get 1/4 inductance.
In the amplifier you can check saturation by seeing carrier residual, but the amplifier probably won't survive if you drive it to 80-100 A output for a long period. This problem can be overcome 2 ways:
- you test the core with significantly higher number of turns, but with much smaller filter capacitance. (eg. 4 timer more turns, and 220 nF capacitor.) Then you can calculate it back to the original number of turn. (For example: 40turns*10A equals to 10turns*40A)
- you can apply narrow impulses to the amplifier. (Attached file.)
You must use dummy load, preferably a low resistance one. When the residual starts to increase, the core saturates.
...
That R and C doesn't provide dead time. You need a diode too!
As I wrote, I have to wait for my friend, who will go to transsylvania in some (2-3) weeks. Other way the postal cost is very high.
DjLeco,
I don't wish to be nasty, so take this as friendly advise. You really need to do some serious homework before you try to design/build a class D amplifier.
Some of the feedback you are getting in this thread is useful, but you will never be successful unless you have some idea of what you are doing. For example, your question on how do I hook up a scope probe to measure current shows a very basic lack of understanding.
Rick
I don't wish to be nasty, so take this as friendly advise. You really need to do some serious homework before you try to design/build a class D amplifier.
Some of the feedback you are getting in this thread is useful, but you will never be successful unless you have some idea of what you are doing. For example, your question on how do I hook up a scope probe to measure current shows a very basic lack of understanding.
Rick
sawreyrw said:
Rick,the ideea is no one has born with alot of knowing...
Measuring the saturation current of an output coil in classD, is not so easy as you can think,so I prefear to ask the others about it.
As you can see,I'm not good to theory part, also not good in mathemtics, but at practice,I can give a real hard chalenge for alot of users of this forum...
So with my practice, and with some helps to understand couple things, I will get succeed!
In principle, classD (for me) is a perfect combination of classAB amplifier and SMPS
KiddingSo I'm good on class AB, ( 25 years ago, my first 100W bipolar AB class amplifier made) and works well now too.
Smps are not so heavy,and I'll give a try to classD...
I prefear to experimentate, research, improve,ask the others,instead stole and reproduce other amps,how others do...
Personally I own alot of Crown schematics, including XTI4000,ITech 6000,and many others, but the great satisfaction,is when you made yourself...
So, I will continue this project!
dudaindc said:
Please do!
Thank you, I WILL!
With a little help from everyone, little by little...
Today I will replace dead mosfets,and the output coils.
If someone is interesed (from the ones that knows better classD), I will make some scope measurements, in fundamental points that they will tell me, to observe exactly how the waveform looks.
If pushing hard, the mosfets failed once again, I will try to use Optos and gate drivers, even double the mosfets(8 instead 4; 2in parralel on each arm).
Waiting for your opinion, is enough to use 4 mosfets for 90Vdc supply and 2 ohms load, or I must double it?
Wish me luck!
For now you need to find why they die, no point od using more and having even more problems, that is for later. Your current shouldn't be that high to kill fet, and they should be heating pretty good at the limit
What you need to find, apart from taking measurements is, does amp die at 50% of power, does it die on 95% of power or does it die at cliping
What you need to find, apart from taking measurements is, does amp die at 50% of power, does it die on 95% of power or does it die at cliping
luka said:
Ok, I will check that.
Is any point in my base schematic,that you want to see the waveform?
Tell me, ans I will make some photos and small movies!
Ouroboros said:
Thanks for answer.
So If I put 2 pieces in parralel the saturation current will be aprox 50A, for 22uH?
And for 2 pieces in parralel and 39uH inductance, the saturation current will be around 35-40A?
And if I double airgap to 2mm,instead 1mm, what will happened?
Is any formula to calculate output coils inductance, for an known freqvency?
By example for 125 Khz and (or) 250 Khz mosfets switching freq.
For 125 Khz, I used 33-39uH for each coil, and the residual carrier on output is under 100 milivolts.
If I decrease output coils value to 22-25uH,keeping the 125 Khz switching, the residual output measured on speaker,rises above 200-250 mv, and in idle, the output coils and mosfets temperature rises a little ; output coils a little more than mosfets, around 30 degrees Mosfets and 40-45 degrees output coils.
If I rise up the carrier to 250 Khz, keeping the 22-25uH output coils value, the residual output carrier measured on speaker is under 20 mv(looks optimum), and everything including coils are damn COLD.
But it looks like a little warmer mosfets on heatsink during audio on amp,instead same power at 125 Khz..
There is my output filters schematic and values (125 Khz switching, and 33-39uH coils value).
An externally hosted image should be here but it was not working when we last tested it.
C1=C2= 1,5uF/250Vcc nepolarised.
C3 = 10uF/250Vcc nepolarised.
In idle(no audio on input), if I disconnect the speaker, with C3 at minimum 10uF, (C1 and C2 same value keeping unchanged)is ok, if I lower C3 value, the amp start to ring itself(autooscilation).
For 250 Khz and 22-25uH value, C3 must be minimum 30uF to avoid autooscilation with no speaker...
What is wrong?
Course on amp input, the audiosignal is in 20-300Hz range, from crossover 24db/oct.
Otherwise the C3 will be a shortcircuit, on output for frecv more than that...
I would be happier to have two coils, each of half the inductance, placed in a series connection. So instead of a 20uH inductor which would saturate at 25A (say), then if you wound a 10uH inductor on the same core, it would saturate at 25A x 1.414, so approximately 35A. As the RM cores have a closed magnetic field, there will be no mutual inductance coupling, so two 10uH coils in series will give 20 uH.
Ouroboros said:
So in this case, replace the both coils from output with 4 coils (2 series on each bridge output) of 10uH to simulate 2x20uH?
And saturation current in that case will be 35 amps per total bridge?
And If I used in parralel(2 each side of 20uH, to simulate 1x stronger 20uH), 2x20uH total?
No! If you put two inductors (of the same value) in parallel, then the inductance will be halved.
So two 20uH parts in parallel will give you 10uH.
NOTE. This is assuming that there is no magnetic coupling between the two inductors.
So two 20uH parts in parallel will give you 10uH.
NOTE. This is assuming that there is no magnetic coupling between the two inductors.
Ouroboros said:
You are right!
Tested!
Now , let's take just for pure example :
Presuming the inductance needed is 18uH for 1/2 of fulbridge amplifier (L1 on high side) ,so the fullbridge will have 2 coils of 18 uH(36uH total)
Litzwire will be 400x0.08mm in both cases.
All cores are the same(R50) in both case.
First picture represents a 18uH coil ,realised by 2x9uH (R50 core) seried cores:
An externally hosted image should be here but it was not working when we last tested it.
Second picture represents 18 uH coil, realised by 2xR50 cores parraleled:
An externally hosted image should be here but it was not working when we last tested it.
So, on our opinion,wich is the best choice?
What I presented here, represents just ONE COIL from the fullbridge amplifier, so will be doubled for fullbridge amp.
If you use toroid, then paralleled cores are the better, but it's impossible to do this with RM cores (wich was in the original question).
But you can parallel two (or more) complete chokes made with RM cores.
To be clear: paralleling (or serializing) chokes, and paralleling cores are different things! On your picture on the top there are independent cores on the top, the chokes in series, and below there are paralleled cores, with single winding. They are electrically equivalent, as you can see.
1-1 pcs of IRFB4227 is usually enough for 45 A peak if everything is OK, but needs quite good thermal contact for safe operation. I quess there is some other problem too.
But you can parallel two (or more) complete chokes made with RM cores.
To be clear: paralleling (or serializing) chokes, and paralleling cores are different things! On your picture on the top there are independent cores on the top, the chokes in series, and below there are paralleled cores, with single winding. They are electrically equivalent, as you can see.
1-1 pcs of IRFB4227 is usually enough for 45 A peak if everything is OK, but needs quite good thermal contact for safe operation. I quess there is some other problem too.
Pafi said:
Yes in second picture, it was 2 parralled cores, not coils!
So in this case I have doubled (teoretically) the section of material, am I right?
And you tell me 1-1 pks of IRFB is enough for 45A peak, so I must understand , in principle the final stage cannot fails from here...
I will give a try with those doubled cores bobbins presented on post 141.
Im intending to believe mee too, core saturation, to be the reason of output devices fails...
No problem for heat transfer between case and radiator, I have used thermal paste and mica foil,and good screew strenght.
So no one told anything for output coils calculations(inductance) for an known frecvency..
The inductance and capacitance in the output filter are a function of the expected load capacitance. Values which are optimal for an 8R load, will not be optimal for a 4R load (and vice versa).
A standard passive filter design program will help you determine the C and L values if you (for example) want a second-order L-C filter, going between a source impedance of 0.1R and a load impedance of 8R, with a -3dB frequency of 35kHz.
A standard passive filter design program will help you determine the C and L values if you (for example) want a second-order L-C filter, going between a source impedance of 0.1R and a load impedance of 8R, with a -3dB frequency of 35kHz.
- Status
- Not open for further replies.