Update
Tested with lm319...Switching frequency has increased from 380 la 520 khz but distorsions have increased at 50 hz and 20 hz..no change at 15khz and above(also tested with aditional capacintace to obtain ~300khz)..so a faster comparator means more distorsions in my case....why???
Tested with lm319...Switching frequency has increased from 380 la 520 khz but distorsions have increased at 50 hz and 20 hz..no change at 15khz and above(also tested with aditional capacintace to obtain ~300khz)..so a faster comparator means more distorsions in my case....why???
Last edited:
Increased switching frequency shows the reduced propagation delay, so it means good.
Something is wrong with freqs at 20-50hz. For this circuit the imput frequency should cound nothing if it is much lower than switching freq. Isn't it because of your signal generator?
What is your power supply? How much is the supply decoupling capacitor?
Yes..something is wrong at frequencies<100hz..and >10kHz.
I tried adding filtering capacitance(from 10mF to 15mF)..no change,but added 1uF at output and i saw a small decrease in distorsions at 50 hz and 1A increase in current before distorsions at 18khz.
I then added 2.2uF but i saw a huge increase in distorsions ..it seems residual was to small so i decreasied switching speed but that generated oscilations with no load(1kohm dummy load) at high frequencies input,so i do not think it will work with 7.5uF output capacitance at 380khz...
..maybe 2-3uf ..
I tried adding filtering capacitance(from 10mF to 15mF)..no change,but added 1uF at output and i saw a small decrease in distorsions at 50 hz and 1A increase in current before distorsions at 18khz.
I then added 2.2uF but i saw a huge increase in distorsions ..it seems residual was to small so i decreasied switching speed but that generated oscilations with no load(1kohm dummy load) at high frequencies input,so i do not think it will work with 7.5uF output capacitance at 380khz...
..maybe 2-3uf ..
Yes..something is wrong at frequencies<100hz..and >10kHz.
I tried adding filtering capacitance(from 10mF to 15mF)..no change,but added 1uF at output and i saw a small decrease in distorsions at 50 hz and 1A increase in current before distorsions at 18khz.
I then added 2.2uF but i saw a huge increase in distorsions ..it seems residual was to small so i decreasied switching speed but that generated oscilations with no load(1kohm dummy load) at high frequencies input,so i do not think it will work with 7.5uF output capacitance at 380khz...
How do you measure distortion? Did you see clipping distortion, or what other non-linearity?
I am using my ears..i have very sensitive hearing..
I listen the difference between no load and load..i know exactly how a clean sine sounds..and those are big distorsions but rarely visible on the digital scope(some are visible especially at high frequencies).
I will make sound card distorsions measurements after i will be unable to pick up with my ears any change in signal in load/noload tests.
.I listen the difference between no load and load..i know exactly how a clean sine sounds..and those are big distorsions but rarely visible on the digital scope(some are visible especially at high frequencies).
I will make sound card distorsions measurements after i will be unable to pick up with my ears any change in signal in load/noload tests.
I am using my ears..i have very sensitive hearing.
I listen the difference between no load and load..i know exactly how a clean sine sounds..and those are big distorsions but rarely visible on the digital scope(some are visible especially at high frequencies).
I will make sound card distorsions measurements after i will be unable to pick up with my ears any change in signal in load/noload tests.
So you put a loudspeaker at output, and listen to it? But then what load and no load means?
Even if you ears are good, your brain is the most unrealiable source for comparison... so stay with oscilloscope, Fourier-transforms or other.
They are there
I made simulations and there are around -40 db.If you want i will post a file with alternating distorsion zones at -40 db so you can see if they are hearable for you.
-40dB distortion means 0,01%. If you want distortions lower than that with this simple schematic, then stop screwing with me
Just to be sure I made some test, generated some sines in audacity. Base is 50Hz, i can hear 150Hz down to -36dB, 100Hz down to -27dB. But this is also because ears sensitivity drastically increases from 50 to 100-150hz, and also because of headphone sensitivtiy incease...
I am usually satisfied with distortions lower than 0,1%. If you want distortions lower than 0,01% then you have to adjust lot of things: very good compator, fast gate drive, minimise dead time, imcrease feedback etc...
And you will need very good instruments.
I am sorry but i know that -40 db means 1 % distorsions(0.01 level)..for 0.01% -80 db it is needed.I would be happy with -60db distorsions
According to the manual of my THD meter (HAMEG), THD is the ration of the POWER of harmonics to the POWER of the base signal (or for small distortions the total signal).
So 10mV harmonics to 1V base signal, lets assum at 1Ohm is 0,1mW to 1W, so it is -40dB both at level, and both at distotions. So 0,01% distortion is -40dB
But just to make sure I searched wiki, and found that there are to mismatching definitions Total harmonic distortion - Wikipedia, the free encyclopedia
So you are right by the other definition. But to measure that distortion a digital scope's bitdepth is usually not enough, so measurements with soundcard is needed.
Last edited:
In case of low distortion (say < 1%), both methods will deliver almost identical values, because if the distortions are low, then the difference of magnitude of the fundamental vs the magnitude of the entire signal becomes neglectible.
But I am massively wondering why you say -40db would be 0.01% ?
In fact -80dB would equal to 0.01%.
xxdb = 20*log(U1/U2) = 10*log(P1/P2)
I guess you just mixed up some zeros.
In your example of U1=10mV and U2=1000mV the ratio is 0.01=1/100=1%
But I am massively wondering why you say -40db would be 0.01% ?
In fact -80dB would equal to 0.01%.
xxdb = 20*log(U1/U2) = 10*log(P1/P2)
I guess you just mixed up some zeros.
In your example of U1=10mV and U2=1000mV the ratio is 0.01=1/100=1%
Okay, so I can give you the following advices to reduce distortion
You said that decreasng dead-time decreased distortion. Try changing the gate-depleting dioda at gate drive to a PNP transistor. This would make turn off faster, then you can decrease the resistor. So overall gate driving will be better.
This looking like just being the transfer function of the output filter.
As I understood you have 10uH and 1.5uF? And 1 Ohms load.
At 18kHz this will deliver almost 3db attentuation.
Means behind the filter clipping must start at approx 70%...72% of the clipping limit in front of the filter.
80% efficacy, might have multiple reasons.
- Measurement errors.
- Slow or ringing gate drive
- Heavy overlap of ON-times (but you would see this in massive heating of the MosFets, even without any load)
In order to say something about the gate drive, we would need to see in one screen shot Ugs and Uds ( time scale 50ns/grid) of the lower MosFet during a switching event in the upper area of a positive signal half wave at high load.
Decreased dead-time decreses distortion somowhat but 20-50 hz distorsion is bad with low and ultra low gate charge mosfets(tested irfb5615) and i do not think adding a transistor will solve it,in my opinion the main problem is in feedback/filtering section.
ChocoHolic i have 10 uH and 1 uF output filter with 1 ohm load..so i have to change the filter to have 0 db attenuation at 16 khz(18 khz is to much) ..is that posible?
I will post gate drive wave forms tomorrow,eficiency is 80% at 1 ohm load and 90%(3% error) at 4 ohm load but using lower rds on transistors didn't improve it so power is lost in switching.
Thanks for your help
ChocoHolic i have 10 uH and 1 uF output filter with 1 ohm load..so i have to change the filter to have 0 db attenuation at 16 khz(18 khz is to much) ..is that posible?
I will post gate drive wave forms tomorrow
,eficiency is 80% at 1 ohm load and 90%(3% error) at 4 ohm load but using lower rds on transistors didn't improve it so power is lost in switching.Thanks for your help
Last edited: