| Workhorse |
Hi guys,
Could PLZ comment on , What s the best way to implement the Deadtime delay to avoid Shoot through in PWM Class-D amps.
1]Resistor+Capacitor+diode combination
2]Delay line using Cmos IC[I dont want to use CMOS ICs]
3]Anyother ???
ampman |
|
|
| Nelson Pass |
This is dependent on the output stage configuration.
What are you using? |
|
|
| peranders |
| We use at work Harris/Intersil HIP4081 (or 4082?) and this IC has built-in deadtime adjustment. You only have to change an external resistor. |
|
|
| Workhorse |
| quote: | Originally posted by Nelson Pass
This is dependent on the output stage configuration.
What are you using? |
I am using Convenetional Half Bridge topology using IR2110 Gate drivers with IRF540N mosfets.
AmPmAn |
|
|
| peranders |
| I suggest that you check out these HIP408x IC's. |
|
|
| Pabo |
| I wouldn´t think that you need additional deadtime with IR2113 because the internal deadtime is at least 30ns. I use HIP2101 which has about 3ns deadtime and I have applied a diode, resistor, capacitor circuit giving about 20ns deadtime. Instead of IRF540N I use 540Z which has a much faster body diode (decreases the shoot through rapidly) with a 20ohm gate resistor combined with a BAS32L turn off diode. This way I achieve 25ns rise and fall times with +/-40V supply voltage. |
|
|
| Workhorse |
| quote: | Originally posted by peranders
I suggest that you check out these HIP408x IC's. |
HIP408X are rated upto 80 Volts only , which is not sufficient for high power amplification where as IR2110 is rated at 500V
AmPmAn |
|
|
| Pabo |
| If you need a "high enough voltage" driver, check out the IR2010 which is a 200V device with 2A drive capability and about 30ns internal deadtime. |
|
|
| ChocoHolic |
Hi AmpMan!
Is there a reason why you do not use the IR21094?
IR21094 offers an adjustable dead time in the range between
0.54µs...5µs with just one resistor.
Do you need the possibilty to control the ON-time of the upper
and lower MosFet in separate? Then forget the IR21094.
But a traditional PWM-amp should be possible without fancy
conrol mechanisms and the drive signals of the upper and lower MosFets contain simply the inverted information + some dead time.
For this the logic of IR21094 should work fine.
Bye
Markus |
|
|
| ChocoHolic |
| ....or is already 540ns dead time unacceptable long? |
|
|
| Pabo |
ChocoHolic
More than about 100ns will cause much higher distorsion. |
|
|
| ChocoHolic |
Hi Pabo!
Thanks. Your answer matches exactly to some upcoming doubts.
I am thinking about a high power pwm-amp for a subwoofer with
two Sadhara drivers.
First ideas:
Full bridge, switching frequency about 70kHz, passive LC output low pass filter 25kHz/2nd order.
Rail voltage around 130V (regulated SMPS), PWM control with Unitrode UC 3823A, halfbridge driver IR2109, FET Drivers Unitrode UC 3710T, MosFets: Semikron MOSFET Modules SKM120B020.
Stop!
Don't ask me if I am going for an audio amp or some traction application...
I already downgraded the project from two TUMULTs to two SADHARAs... (hope I will manage to get them from Adire...)
:angel:
All comments welcome. I did not start up to now and this project will keep me busy the next year anyway.
I am not in hurry because that older beast still is working fine, but
the two sadhara will offer about ten times more volume displacement.... about 100times more acoustic power .... about 20db more max. output...
I must admit, it is more some sort of "research" or curiosity than a real demand :D .
Bye
Markus |
|
|
| Pabo |
ChocoHolic
I have designed a 500W/4ohm classd amp for subwoofer apps and it runs at 250kHz which was the frequency that I reached the lowest bias current. Below 250kHz the ripple current in the output choke became too high and above 250kHz the switching losses became too high. It is a fullbridge with +70V supply and it uses a simple voltage hysteresis regulation from before the output filter.
What I am trying to say is that it is not obvious that it is better to go down in frequency when you are designing a sub amp. If you want to limit the ripple current to a couple of amps you will have to use a 200uH inductor or something which has to handle 30A or so. That is a pretty hefty inductor!! |
|
|
| ChocoHolic |
Hi Pabo!
Yes, the output filter is one of the keys for success or not.
My first approach is a 2nd order filter with around 50uH and
1uF and a 10 Ohms (parallel to 1uF). If the two Sadharas would be pure resistors the HF current would still be very high.
4.6A in sum for the paralleled Sadharas, making 14W steady state losses in each speaker. Not immediately destructive, but not nice.
(Sadhara: each nominal 4 Ohms, DC: each 2.6 Ohms)
But the speakers Le (1.3mH each speaker) theoretically will bring this down dramatically to about 300mA in sum, making some milliwatt in each speaker. Do I miss something?
But I am thinking also about the resulting resonance modes of the
filter in combination with the drivers. I fear that they are in the audiable range.....
From this side I would prefer a simple inductor (200uH) in series to the paralleled speakers, no cap, and a 10 Ohms non inductive resistor paralleled to the speakers .
Losses in the speaker: some milliwatts.
Losses in the 10 Ohms around 30 Watt.
Choke: challenging....
...inductor for such currents is a certain sort of fun for itself....
or better "pretty hefty" ?
But don't worry about the magnetic...
...one of the parts, where I am confident that I will manage it... |
|
|
| ChocoHolic |
....but you might be right that also in my case the best trade off might be above 70kHz. I will have to investigate, which switching frequencies I can handle with that bulky MosFets, which I am thinking of.....
Cheers Markus |
|
|
| thoriated |
| Don't know how applicable this observation is, but when I was playing with open reel tape bias oscillators several years ago, I found that the audible fidelity subjectively improved as I increased the bias frequency up to at least 400 khz. If I was designing a high quality class D amplifier (I'm not at this point), I would probably accept some switching losses to run at a higher frequency than at the minimum for overall thermal losses if it improved achievable performance in other areas such as high frequency damping factor & distortion, frequency extension and phase linearity, etc. |
|
|
| ChocoHolic |
....things are getting interesting...
Up to now I am a beginner in switch mode amps.
But I got the speakers today ==> no way back.
DIY direction for my next year is clear. I will have to learn
about switch mode amps :hot:
I am faithfull that a subwoofer is less critical with respect to the switching frequency than a full range application.
If I consider that my max signal frequency that I want reproduce is around 50Hz, then 70kHz give a frequency ratio of 1400.
An amp for the full audio frequency range which must reproduce
20kHz and would operate at 1MHz (who can handle? :D In my percepting 1MHz is the entry to Voodoo Land).... well even you would run a full range amp at 1MHz the frequency ratio would only be 50.
So I felt quite comfortable with a factor of 1400 for my sub.
But I must admit: I have no practical idea in this field...
Let's see what will happen! |
|
|
| peranders |
| quote: | Originally posted by Pabo
I have designed a 500W/4ohm classd amp for subwoofer apps... |
Hej, Patrik.
We met earlier and you mentioned about some class D amp. Was it this? Can you show us some pictures and impress us? You surely impressed me last time we met :nod: Nice looking mono blocks! |
|
|
| thoriated |
| I could be off base, but I suspect part of the attraction of class D power amps is that they don't suffer from small signal output stage nonlinearities in the way that many class AB amps do. |
|
|
| Pabo |
Peranders
Hello there!
I have seen you around here, I sure liked your designs as well :)
The subwoofer amp is not the hi end amp I was telling you about. The hi end amp is called VCCA and is based on a more complex two loop design. We are continuously improving VCCA and have increased the linearity so that the THD is down at about 0,002% at 1W/1kHz and 0,01% at high frequencies. The interesting part is that he distorsion is allmost unchanged with power. For example at 50W/1kHz the THD is about 0,005%. Next DIY gettogether I will bring a stereo VCCA amp.
Thoriated
You may be right about that. |
|
|
| phase_accurate |
| quote: | Thoriated
You may be right about that.
|
I agree about that too altough care has to be taken for not messing this up again due to EMC etc.
OTOH deadtime has indeed some effect but IMO not as much as some people claim who directly compare the deadtime to crossover distortion in class-AB amps.
Another advantage of class - d: It drives reactive loads with ease. This is of course depending on frequency, feedback/output-filter topology and PSU topology. But at low frequencies they are a clear winner in this respect.
Regards
Charles |
|
|
| peranders |
| quote: | Originally posted by Pabo
Peranders
Hello there!
I have seen you around here, I sure liked your designs as well :)
The subwoofer amp is not the hi end amp I was telling you about. The hi end amp is called VCCA and is based on a more complex two loop design. We are continuously improving VCCA and have increased the linearity so that the THD is down at about 0,002% at 1W/1kHz and 0,01% at high frequencies. The interesting part is that he distorsion is allmost unchanged with power. For example at 50W/1kHz the THD is about 0,005%. Next DIY gettogether I will bring a stereo VCCA amp. |
Watch out Lars and the "ICE"-team, here is pabo :cool:
I can't wait.... to see your amp.... not a sneak preview...? |
|
|
| Workhorse |
| quote: | Only my mother calls me Per-Anders.
|
BUT AmPmAn will Call u
P-E-R-A-N-D-E-R-S THE FRIEND
I am testing my Prototype using 100nS deadtime , IR2110 , IRF540N with Full Bridge-Class-BD Trilevel modulation and i will post the photos too. Besides this i also want to look at PABO's amp also.
compliments
ampMAN:D |
|
|
| Workhorse |
| quote: | Originally posted by phase_accurate
I agree about that too altough care has to be taken for not messing this up again due to EMC etc.
OTOH deadtime has indeed some effect but IMO not as much as some people claim who directly compare the deadtime to crossover distortion in class-AB amps.
Another advantage of class - d: It drives reactive loads with ease. This is of course depending on frequency, feedback/output-filter topology and PSU topology. But at low frequencies they are a clear winner in this respect.
Regards
Charles |
Hi Charles Nice to see u too.
So whats ur suggestion on implementing the deadtime circuit choices. Secondly u mentioned about the reactive loads which are easily driven by class-D amps than class-AB.PLZ CommenT
Compliments AmPmAn |
|
|
| zic |
| but if I want to get the Delay by using Cmos IC,then what shoud I do? |
|
|
| phase_accurate |
The following is easily done in CMOS but a little unpredictable.
Regards
Charles |
|
|
| ChocoHolic |
| ...hm, during my last examination I came to the conclusion that adjusting the dead time with an RC-RD circuit for the gate drive is pretty fine. This allows to adjust the dead time with quite acceptable thermal drift and further on the external gate-source capacitor allows to control the dead time and also di/dt commutation speed as well. Furtheron it is providing a low hf gate drive impedance, which is reducing trouble from the MosFet's parasitic drain-gate-capacity. |
|
|
| IVX |
| Yeah, and why we shall purchase irs21955 for $5, if can buy simplest driver for $.5 and next adjust any dead time. :) |
|
|
|
