| Tekko |
I was a little bit bored today so i started playing with a LM311P and a TC4422, this was the outcome after a few hours of playing around:
Video. |
|
|
| lumanauw |
Nice :D
Everything stays cool? At what rail voltage? |
|
|
| darkfenriz |
| do you drive output inductor directly from mosfet driver? |
|
|
| Tekko |
Yes. The thing get warm to the touch, rail volts is 15v regulated.
I settled for 1.52MHz since it got hot above 2MHz. |
|
|
| fumac |
good job tekko
the voltage is a bit lower. 15v
but you can run it at 2mhz is cool.
tell us how about the sound at mhz :)
fumac |
|
|
| Tekko |
The amp sound reather good, i don´t have a speaker that would make the actual sound justice, so i need to get RMAA and install a M-Audio Audiophile USB soundcard on the laptop to get a line in, since the laptop soundcard doesent have one.
I also need to find my T106-2 cores to replace this crappy -26 material one. |
|
|
| NeoY2k |
Great job, I'm really interested in these Mhz design as they really look simpler than hypex ones.
But what about performances? It's hard to find information about this technology. |
|
|
| Tekko |
My amp is ****, i tried to measure it but IMD + Noise peaked @ over 24000%, however thd was only 0.06%.
The imd problem could be due to that the amp generates a tone on low freqs, yet im not sure if it comes from the preamp or the poweramp, or if its due to that i dont have af error feedback yet. |
|
|
| Tekko |
The measurement errors turned out to be an unknown mic boost on the line input, the beeping coming from the amp on low freqs turned out to be the signal generator because it was also hearable in the laptop speakers.
Anyways i built a stereo version of my little amp today, pics are here. |
|
|
| NeoY2k |
Nice and small stuff! And seems to work well!
Do you have any schematics?
Does it includes an AF error feedback?
The only paper I found is
http://class-d.net/tech/
If you could give me some information on this :)
Thanks,
Nicolas |
|
|
| Tekko |
Schemo is here.
My MCD is based on fumac´s block diagram. |
|
|
| lumanauw |
Hi, Tekko,
The (+) and (-) inputs are different points from Fumac's block diagram? |
|
|
| Bender.ru |
Hi Tekko!
what is advantage of 1..2 mHz carrier freq. amp compared with usual 300..400kHz carrier freq. amp?
or it's only "science" experiment ;) ? |
|
|
| Tekko |
Lum, that have to do with that the LM311P output is inverting, if i use a TC4421, then the inputs would be the same as fumac´s block diagram.
Bender.ru, the higher switching frequency allows for more accurate reproduction of higher part of the audio band, class-d.net explains this. |
|
|
| Bender.ru |
| what about switching precise? :rolleyes: |
|
|
| lumanauw |
Hi, Bender,
I read Fumac's explenation here http://class-d.net/tech/ about the advantage of Mhz modulation. The loss (yellow triangle tip) seems can be applied if the system is forced clocked (triangle clocked).
Fumac, is this also applicable to self-oscillating classD? Isn't that in selfoscillating, it's self-adjusting towards errors? |
|
|
| Pabo |
lumanauw
| quote: | | is this also applicable to self-oscillating classD? Isn't that in selfoscillating, it's self-adjusting towards errors? |
You are correct to some degree. Saying that a higher frequency improves the high frequency reproduction needs to be proven, not just taken for a theoretical to practical translation.
Self oscillating amplifiers gets more loop gain the more it attenuates the switching ripple but at some point the propagation delay and switching nonlinearities will take over and cause more distortion. My experience is that going above 600kHz is difficult without generating very high levels of radiated emissions. Of course if one is only building for DIY it would be OK as long as the neighbours pace maker does not stop.
Saying that EMI decreases when the switching frequency is pushed above 1MHz is pure fantasy and based only on the assumption that if the residual decreases in amplitude the EMI decreases. If you switch twice as many times the radiated emissions double and if you want to use this higher frequency for improving the sound the DV/DT need to be shortened easily doubling the radiated noise again. |
|
|
| Eva |
| Fumac's claims are not based on scientific criteria. |
|
|
| latala |
I like the look of this! and it may be a usefull introduction to class D can you provide component values ?
I have tried without success to make a simple class d but all I get is smoke
Regards Trev |
|
|
| fumac |
| quote: | Originally posted by Pabo
lumanauw
You are correct to some degree. Saying that a higher frequency improves the high frequency reproduction needs to be proven, not just taken for a theoretical to practical translation.
Self oscillating amplifiers gets more loop gain the more it attenuates the switching ripple but at some point the propagation delay and switching nonlinearities will take over and cause more distortion. My experience is that going above 600kHz is difficult without generating very high levels of radiated emissions. Of course if one is only building for DIY it would be OK as long as the neighbours pace maker does not stop.
Saying that EMI decreases when the switching frequency is pushed above 1MHz is pure fantasy and based only on the assumption that if the residual decreases in amplitude the EMI decreases. If you switch twice as many times the radiated emissions double and if you want to use this higher frequency for improving the sound the DV/DT need to be shortened easily doubling the radiated noise again. |
hello Pabo
good question.
do you heard about the AES17 filter?
most of testing class-d by ap system must insert this filter between ap system and class-d AMP
why? because the EMI, the The residual amplitude will make the input stage of AP system err.
you can take a look at our test. we have not connet the AES17 (same like).
MCD is the only one class-d amp can Measurement same like class-AB AMP
we can test the THD at 80k bandwidth, no other class-d can, because they must insert a 20~20khz bandwidth filter (AES17 same like)
rg
fumac |
|
|
| fumac |
| quote: | Originally posted by Eva
Fumac's claims are not based on scientific criteria. |
yes eva, your are right :)
all my job is base on god,
one day ,two years ago,
when i Pray at a cave, : "God help me please, give me a class-d amp,give me a class-d amp..."
Suddenly, a pcb droped to my head.
then i got MCD,
here is the picture what happened that day
 |
|
|
| fumac |
| quote: | Originally posted by Eva
Fumac's claims are not based on scientific criteria. |
yes you r right, eva, :)
after one year ago,
i goto there again,
i Pray at a cave, : "God help me please, give me a class-d amp, more power full, more power full..."
Suddenly, a new pcb droped to my head.
then i got MCD-V,
 |
|
|
| fumac |
| quote: | Originally posted by Eva
Fumac's claims are not based on scientific criteria. |
yes you r right, eva,
after one year ago,
i goto there again,
i Pray at a cave, : "God help me please, give me a class-d amp, more and more power , more and more power ..."
not Suddenly agian, a new pcb droped to my head.
then i got mhzpower-4,
 |
|
|
| Pabo |
fumac
The AES17 filter is not there to filter out EMI. The AES17 filter is a software filter simply there to filter out the harmonic components outside the audible band.
The AUX0025 filter on the other hand is a filter provided by Audio Precision to filter out HF noise from the signal before it is sent into the AP.
The radiated noise I am referring to is at a very high frequency. If you are using fast modern MOSFETs it usually occurs around 100MHz. Have you measured radiated noise according to CISPR22 or EN55022? If so please provide measurement data to convince me.
I am sorry I started filing replies here as you seem to be on a level much too low to listen to critisism with an open mind. |
|
|
| fumac |
| quote: | Originally posted by Pabo
fumac
The AES17 filter is not there to filter out EMI. The AES17 filter is a software filter simply there to filter out the harmonic components outside the audible band.
The AUX0025 filter on the other hand is a filter provided by Audio Precision to filter out HF noise from the signal before it is sent into the AP.
The radiated noise I am referring to is at a very high frequency. If you are using fast modern MOSFETs it usually occurs around 100MHz. Have you measured radiated noise according to CISPR22 or EN55022? If so please provide measurement data to convince me.
I am sorry I started filing replies here as you seem to be on a level much too low to listen to critisism with an open mind. |
hi
S-AES17 Low Pass Filter
S-AES17 is a hardware filtering kit for optional installation in System Two, System Two Cascade or Cascade Plus that satisfies
the AES17-1998 specification for a standard low-pass filter for THD+N measurements of digital-to-analog converters (DACs) which exhibit high-level out-of-band noise.
APWIN 2.20 and later versions support the use of this filter. The filter can be set to either of two bandwidth selections: 20kHz or 40kHz. The S-AES17 Low-Pass Filter Option improves upon the capabilities of the earlier S2-AES17LP filter, which is no longer available.
please goto here
http://www.audiotech.com.tw/AES17%20Filter.htm?
aux0025 is a new filter for class-d
http://www.tti.co.uk/products-resale/ap/aux0025-1.htm?
we have no this two filters too, no any filter between class-d and AP system
we are testing our finished AMP to pass CE, and FCC, when it ready , i will post it for you
rg
fumac |
|
|
| Pabo |
fumac
Yes, I know what the AES17 filter spec says. I just thought you where referring to the one inside the audio precision software.
Anyway, I am not saying that it is impossible to switch at 1MHz or above. I am just saying that if the flanks are very fast and you switch at such a high frequency you will generate much more radiated noise. This noise is of course possible to handle but at a higher price and with more time spent to do it.
I would certainly like to see your emissions between 30MHz and 1GHz. Specify the measurement setup when you post the result so that it is easily comparable. |
|
|
| fumac |
| quote: | Originally posted by Pabo
fumac
Yes, I know what the AES17 filter spec says. I just thought you where referring to the one inside the audio precision software.
Anyway, I am not saying that it is impossible to switch at 1MHz or above. I am just saying that if the flanks are very fast and you switch at such a high frequency you will generate much more radiated noise. This noise is of course possible to handle but at a higher price and with more time spent to do it.
I would certainly like to see your emissions between 30MHz and 1GHz. Specify the measurement setup when you post the result so that it is easily comparable. |
this test need a Shielding room to test it, so we send our amp to a Certification company to test it , when it ready, i will post it for you
rg
fumac |
|
|
| Pabo |
fumac
I see what you where saying about the AES17 filter. I do not know how the Audio Precision modules is made in detail but I was assuming that 60dB attenuation from 20kHz to 24kHz with less than 0,1dB attenuation at 20kHz was mode in the digital domain.
I was wrong there and I will need to study how that is done. |
|
|
| fumac |
| quote: | Originally posted by Pabo
fumac
I see what you where saying about the AES17 filter. I do not know how the Audio Precision modules is made in detail but I was assuming that 60dB attenuation from 20kHz to 24kHz with less than 0,1dB attenuation at 20kHz was mode in the digital domain.
I was wrong there and I will need to study how that is done. |
if you do with one well know class-d amp , when you use 2 modules in one case, and when you put the two modules nearby each other,
that will found the noise floor goto very high. sometimes beep.
i have no this class-d amp modules, all this is one of our oem user told us,
they just can place the two module at two side of the case.
but our modules can place nearby each other,
and the noise floor same like stand alone, MCD nevre beep.
perhap this will let you know about the emi of mcd.
after all, i will post the test after ready :)
nice to meet you, and the post of the god is not for you.
sorry for it
rg
fumac |
|
|
| fumac |
| quote: | Originally posted by Pabo
fumac
I see what you where saying about the AES17 filter. I do not know how the Audio Precision modules is made in detail but I was assuming that 60dB attenuation from 20kHz to 24kHz with less than 0,1dB attenuation at 20kHz was mode in the digital domain.
I was wrong there and I will need to study how that is done. |
more info about aes17 filter
Audio Performance Measurement
Audio measuring equipment with an AES17 brick wall filter, such as the Audio Precision AP2, are necessary. However a classic audio analyzer like the HP8903B can be used with appropriate pre-stage low pass filter is applied. The important consideration here is that the output signal of a Class D amplifier still contains substantial amount of switching frequency carrier on its waveform, which causes a wrong reading, and those analyzers might not be immune enough to the carrier leak from a Class D amplifier. Figure 9 shows an example of a filter.
here is the link
http://csoneone.bokee.com/6538503.html
pdf from IR
http://www.irf.com/technical-info/appnotes/an-1071.pdf |
|
|
| darkfenriz |
| I could agree that higher switching frequency can have some technical advantages, like lower aliasing and less tweeter heating for full range applications, but generally loses at efficiency and harmonic distortion. That's science! |
|
|
| fumac |
| quote: | Originally posted by darkfenriz
I could agree that higher switching frequency can have some technical advantages, like lower aliasing and less tweeter heating for full range applications, but generally loses at efficiency and harmonic distortion. That's science! |
hello darkfenriz
yes , that's science
show you a picture
you can find it at this link:
http://www.diyaudio.com/forums/show...25&pagenumber=1
rg
fumac |
|
|
| darkfenriz |
Seems you've done a nice work!
What are the mosfets?
Any special snubber techniques? |
|
|
| Bender.ru |
| Fumac, EMI & switching residual are two big differences :) |
|
|
| fumac |
| quote: | Originally posted by darkfenriz
Seems you've done a nice work!
What are the mosfets?
Any special snubber techniques? |
sorry, dont ask about this,
i'm making living on it, i have read about >3000 types mosfet datasheet, and test >100 types mosfet, then select <10 types for our project.
and i can show you the TDH+N, testing at 80k bandwidth 8 ohm load ,mhzpower-4 module
rg
fumac |
|
|
| Tekko |
Heres some graphs of my little amp:
Link |
|
|
| NeoY2k |
Wow, I don't know if the amp does, but at least people on this topic are overheating a bit xD
Well, your amp is nice but shows a quite small SNR don't you think?
Fumac, would you please fix your website? I'd be interested in your amps, but I need to have a look at SNR & THD measurements... as well as prices of the amp.
If it works well, and is simple in the sense of few components (at least does it looks like), it should lead to a fair cost. Which is very interesting ^^ |
|
|
| Tekko |
The snr is prolly due to that the amp itself have quite low gain so i have a preamp on it that gave quite high gain.
maybe the AF error feedback resistor and audio input resistor should be equal in resistance, or the input resistor shoiuld be smaller, or maybe ther preamp opamp should be included in the feedback loop, or it could be that its a low power amp running off only 15V as well as beeing single rail and having input and output dc blocking caps, what do i know.
Its a work in progress.
Heres the scopes noisefloor: Link |
|
|
| NeoY2k |
Ok, great work anyway :)
I'll keep reading your thread ;) |
|
|
| Eva |
The extra loop gain allowed by a higher switching frequency is employed to correct the higher dead time error also due to the higher switching frequency. There is no advantage in that respect. THD performance is same as 350Khz UcD. Snubber losses and switching losses increase three fold, though.
The clmaims on Fumac's site derived from the triangle-wave figures are wrong.
I don't find anything interesting on those pictures showing square waves on an CRO.
Carrier residual is something inherent to switching amplifiers and there is no need to suppress it.
There is no point in amplifying components above 20Khz, anything above 20Khz entering an amplifier is usually ambient EMI hash and should be filtered. |
|
|
| Iyremenko |
Eva, you are being too kind.
Fumac's relentless sales pitches consistently demonstrate a failure to grasp the most basic concepts of modulation theory.
When you start by ripping off "perfection" though, it is easy to "improve" it.
"sorry, dont ask about this,
i'm making living on it, i have read about >3000 types mosfet datasheet, and test >100 types mosfet, then select <10 types for our project."
That's PATHETIC. You have ripped off others and yet aren't even willing to discuss your choice of mosfet? Like we can't find out .... if we really cared to?
Clearly you don't have much to go on.:rolleyes: |
|
|
| RX5 |
| quote: | Originally posted by Eva
The extra loop gain allowed by a higher switching frequency is employed to correct the higher dead time error also due to the higher switching frequency. There is no advantage in that respect. THD performance is same as 350Khz UcD. Snubber losses and switching losses increase three fold, though.
|
so this means its better to stay in the 350Khz UcD "zone", right?? :D |
|
|
| Pabo |
RX5
I would say that if you are able to design an output stage which has much smaller parasitic components (series inductance and parallell capacitance) you can start switching with much higher DV/DT without causing higher ringings and radiated noise. If the propagation delay and DV/DT is improved by a factor of two you can move the filter resonance by the same amount, double the fsw and get a self oscillating amplifier with 6dB higher loop gain at 20kHz hence less distortion. It would also have less phase shift (which is useless but anyway).
A problem that can occur is that if the output filter has a too low filter cap the amp stability gets affected by driving a capacitive load. |
|
|
| Eva |
The impact on distortion of dead time and any timing error on switching transients is proportional to the switching frequency.
For example, 10ns of dead time produce 10 times more THD at 1Mhz than at 100Khz. |
|
|
| lumanauw |
| What about 20khz phase shift/open loop in Fumac's website? 400khz shows -80deg, 1mhz shows -19.5deg. How important is this to audible sonics (when the loop is closed)? |
|
|
| fumac |
| quote: | Originally posted by Bender.ru
Fumac, EMI & switching residual are two big differences :) |
this is just your thinking.
switching residual is one part of emi.
there are not same thing, but they have relationship.
if class-d designer don't care about switching residual,
then you can design a class-d without output filter, or you can design the output filter at the speaker box.
but , why no one do this ?
a part of THD come from the output filter.
diffrent thinking will give you diffrent answer. |
|
|
| fumac |
| quote: | Originally posted by Eva
The extra loop gain allowed by a higher switching frequency is employed to correct the higher dead time error also due to the higher switching frequency. There is no advantage in that respect. THD performance is same as 350Khz UcD. Snubber losses and switching losses increase three fold, though.
The clmaims on Fumac's site derived from the triangle-wave figures are wrong.
I don't find anything interesting on those pictures showing square waves on an CRO.
Carrier residual is something inherent to switching amplifiers and there is no need to suppress it.
There is no point in amplifying components above 20Khz, anything above 20Khz entering an amplifier is usually ambient EMI hash and should be filtered. |
eva, dont just talking the msg from book, make a very low thd class-d amp just running at 100k,
good job is better than good words.
as your thinking , you can make a class-d amp, with very low thd and full range (20~20khz)and just running at 100k , also including low EMI,
so , please show us your low thd class-d amp.
as you say, that's a very easy job to you,
don't stop at reading a book, do it please
rg
fumac |
|
|
| fumac |
| quote: | Originally posted by Iyremenko
Eva, you are being too kind.
Fumac's relentless sales pitches consistently demonstrate a failure to grasp the most basic concepts of modulation theory.
When you start by ripping off "perfection" though, it is easy to "improve" it.
"sorry, dont ask about this,
i'm making living on it, i have read about >3000 types mosfet datasheet, and test >100 types mosfet, then select <10 types for our project."
That's PATHETIC. You have ripped off others and yet aren't even willing to discuss your choice of mosfet? Like we can't find out .... if we really cared to?
Clearly you don't have much to go on.:rolleyes: |
as you can find out the mosfet, ok don't ask me please.
our mosfet is drop from God,
but i can told you a truth (i have told others at this forum):
Most of the mosfets can run at 1MHz, you can take a look at the datasheet of mosfet, the manufactury tested the mosfet at 1mhz"
rg
fumac |
|
|
| IVX |
| quote: | Originally posted by Eva
The extra loop gain allowed by a higher switching frequency is employed to correct the higher dead time error also due to the higher switching frequency. There is no advantage in that respect. THD performance is same as 350Khz UcD. Snubber losses and switching losses increase three fold, though.
|
:) Oohhh noooo, Fumac told me, that his loop gain is extremely low and equal 10 (I guess 10 is mean 20db, though maybe I'm wrong). So his 1-2mhz amp must have in fact no dead time (probably =<1nS) if .05% with 10-20db loop gain yet (maybe some kind of sw.error noise shaper implemented, a'la Westlake?). I've listened class D amp (tda8939 based) with up to 2.5mhz (or little bit less) switching, and I see no reason to use it, sound is the same, THD+N worse, EMI 100% should be worse too, and of course lower efficiency. Residual switching easy to get smaller in BD operation H-brige (just couple of UcD +1 cap 47nF). |
|
|
| Iyremenko |
Pabo, phase could be useless if all you make is subwoofer class d. Though in the context presented by fumac, yeah.
Ivan, you just invented fumac's next product line. Though, that would not be BD ;)
Fumac I dare say, you are unconvincing with your dropped from god references, be them jokes or not, it is a very weak, and incorrect standpoint. One wonders if that is your best.
I have zero need to know your mosfet, but I think you should be able to discuss it openely. If you find that secretive, proprietary information, I think you'll sink real fast.
What has god given you for EMI performance ?
:dodgy: |
|
|
| Eva |
For a given circuit, the amount of EMI energy radiated is proportional to the switching frequency (to the number of switching events per second), so a given circuit operating at 1Mhz will produce 3 times more EMI than at 330Khz.
Instead of showing us square waves, you could as well show us the carrier residual at the output of the amplifier with no input signal and a dummy load connected to one of the supply rails to force some current to flow and drive the output stage into non-resonant hard switching.
The amount of hash in the resulting waveform tells a lot about the quality of any class D amplifier... ;)
Phase is not relevant, group delay is. A reasonably flat group delay is always ok. |
|
|
| fumac |
| quote: | Originally posted by Iyremenko
Pabo, phase could be useless if all you make is subwoofer class d. Though in the context presented by fumac, yeah.
Ivan, you just invented fumac's next product line. Though, that would not be BD ;)
Fumac I dare say, you are unconvincing with your dropped from god references, be them jokes or not, it is a very weak, and incorrect standpoint. One wonders if that is your best.
I have zero need to know your mosfet, but I think you should be able to discuss it openely. If you find that secretive, proprietary information, I think you'll sink real fast.
What has god given you for EMI performance ?
:dodgy: |
i will post emi test pictures here, please wait
we need to pay a proffession company to test it .
at my lab now, i can told you about using the MCD
MCD can work very well with TV set and FM radio.
even mcd is not installed in a cover box.
i have no instereting at high power damp(>1000w).
because i have test many class-h amp(1000w ~2000w to 8ohm made by my friends)
i found , class-d is not good at high loading (8ohm) and class-h is very good at high load and huge power output.
(full audio range 20hz~100khz)
we do resaerch at class-d , so we know which using is good for class-d,
class-d good at : bass boost (even high power to 10kw is ok), full range, to about 1kw 4ohm is ok, also can drive 1ohm to 2ohm
class-h, very good at high power, high load, 1000w 8ohm~ 3kw 8ohm, when you test a good class-h, you will find: 400w rms to 8ohm thd<0.01% at 1k ,
then you will know , which is better for PA system.
so , we will stop class-d rearch when our 1000w 4ohm version (full audio band )is finished. we have no any instereting at bass-boost class-d amp.
a real class-d designer must to know about which is the best of class-d and which is worst of class-d.
forget about the 1kw 8ohm, full range(at least 20~100k),will not better than class-H now,
class-d need a super fast mosfet. then will be more good at high power and full range audio
about our new project:
after that, we will goto SMPS, infact now ,we are testing the first test version of smps for hifi audio ,
250w rms and 600w peak, we have tested it , the noise floor is about 50mV, this is a bit high for hifi audio, so need to do with it
we have a diffrent thinking at smps, we will have the fastest speed loop of reg, this is one of the way to make good sound, and dynamic.
this version is just running at 350k, but we want a mhz level smps.
you can take a look at : vicor, they are making very good high power DC-DC working at 1mhz.
rg
fumac |
|
|
| fumac |
| quote: | Originally posted by Eva
There is no point in amplifying components above 20Khz, anything above 20Khz entering an amplifier is usually ambient EMI hash and should be filtered.
|
hi eva, i have told you about tannoy
ok , let me post the link again here:
http://www.tannoy-speakers.com/s.ph...itle=ST200&s=27
ST200
Introduction
Always an innovator where quality sound reproduction is concerned, Tannoy has continued this tradition by being at the forefront of the development of WideBand™ technology. All Prestige models except Autograph Mini are compatible with
Tannoy’s SuperTweeter™ designs, providing the opportunity to extend high-frequency response to above 50 kHz
, thereby providing all of the bandwidth required for today’s wide bandwidth digital recording formats.
you can ask tannoy , for what , why need a speaker from 18k~50k, they have three types about this :)
 |
|
|
| fumac |
who can tell me why?
are they stupid or are we stupid?
rg
fumac |
|
|
| fredos |
Hey Fumac...Have you ever heard about group delay and impulse response of an amplifier....No need to go to 100Khz for this with switching amplifier...This is mostly related to output filter. In other hand, analogue amplifier need a wide frequency response to have good group delay and impulse response...Think about this! Eva is totaly right about too hight switching frequency...
Fredos |
|
|
| Tekko |
| Sheesh guys:hot:, cool down or i´ll have this thread locked.:smash: |
|
|
| Pafi |
fumac!
In linear mode amps the wide freq. response is a "side-effect", not primary purpose. In some of them there is a filter to limit it, but 100...200 kHz is still harmless.
(For example now I make a linear mode current generator for 0...100 Hz signals, and despite of the low signal frequency every stage have to have n*100 kHz BW at least, because of the high precision requirements.) |
|
|
| Iyremenko |
Hi Fumac,
Do you mean to tell me, that as "RF guy", you can't devise a simple measurement that would give you an idea as to the EMI performance of your module???
I wish I could say I'm speechless, for your sake, I'll pretend to be.
:dead: |
|
|
| Eva |
| An oscilloscope picture of carrier residual is already a good indicator of EMI levels, and I think that someone already published that picture a few munths ago... It was showing really bad performance with lots of ringing. |
|
|
| megajocke |
| Ouch, that triangle wave "missing parts" stuff hurts my head... (Hint: How high do the harmonics of a 20kHz fundamental triangle wave go?) ;) |
|
|
| Eva |
| Triangle waves (and square waves) contain infinitely high frequencies :D:D:D He has sooo much to learn... |
|
|
| Pafi |
| Pardon? What is this discussion about? |
|
|
| lumanauw |
| Sawtooth (even it looks like triangle wave) seems more destructive in HF harmonic contents than triangle or square wave. |
|
|
| phase_accurate |
The original topic was about fast switching stages. IMO they make sense for switching amps using delta sigma modulation.
PWM amps don't need them IMO.
A sufficient bandwidth (i.e. around 40 kHz) can be achieved with lower switching frequencies.
Flat group delay, low EMI, good carrier suppression, flat amplitude response, low THD, low IMD ....... can be achieved with low switching frequencies and well-known topologies.
Regards
Charles |
|
|
| lumanauw |
| What if the question is flipped. How low oscillating frequency is the minimal for descent full range classD? 150khz? 250khz? |
|
|
| Tekko |
| well that is completely up to each individuals ears, some may say a 40khz d-amp is fully good, some may want a GIGAHERTZ one, some cannot stand any class d and would only accept a class a, ab or b amp. |
|
|
| IVX |
| quote: | Originally posted by Tekko
well that is completely up to each individuals ears, some may say a 40khz d-amp is fully good, some may want a GIGAHERTZ one, some cannot stand any class d and would only accept a class a, ab or b amp. |
indeed to say 1MHz or even 1THz it is easy, try to find sound differences in blind trial, is much more complex. |
|
|
| fractuured1 |
| Doing an engineering project using a board with a tc4422 on it, the driver started to smoke. Now it's dead. If you guys lived in the boondocks and wanted to find another tc4422, where would you look? I'm ready to resort to disassembling items in my house to get one out. This project is due Friday, so it's too late to try FedEx. Any suggestions? I appreciate any tips! |
|
|
| Pafi |
| I'd try to build it from discrete components. |
|
|
| Tekko |
| I got the TC4422´s from ebay. |
|
|
| fumac |
hi eva, just reading about this post
http://www.diyaudio.com/forums/show...25&pagenumber=2
i agree with you about this :
| quote: | Originally posted by Eva
Such manufacturers are not actually manufacturers but bare amateurs without a clue about circuit design finding their way to earn big money.
It's all a load of B.S. They will be bare amateurs until they become able to develop their own high performance class D stuff from scratch.
|
as you say:
| quote: | Originally posted by Eva
The impact on distortion of dead time and any timing error on switching transients is proportional to the switching frequency.
For example, 10ns of dead time produce 10 times more THD at 1Mhz than at 100Khz. |
i think , you have very good knowledge from books,
so dont just stop at the talking at internet.
you need to make a so much low thd class-d amp to proof your Theory
from our mhz-power class-d module
we got a very low thd+n :
lowest is :0.0035% at 1k and 0.006% at 10k (1.5w 8ohm)
as you say, you need to make a class-d running at 100k or 150k then can get thd+n same like :
0.00035% at 1k and 0.0006% at 10k (8ohm) (10x better than ours )
so please show your project to us ,
if you can't make it ,
or just tell me who else can make a class-d like that.
Don't just stop at the books,
Don't just stop at talking at forums
Do a good job and show to us please.
as you day, i need to learn more, yes, i need to learn english and do more researching at class-d.
we never stop at the devolopments.
too busy this days, because our new baby was born 3days ago.
show you our new baby :
thanks
rg
fumac |
|
|
| IVX |
| BTW, at 1.5W your amp just working without dead time at all. Indeed learning is the best way for you, don't worry about english, I'm sure you can get right chinese books even.:D |
|
|
| fumac |
| quote: | Originally posted by IVX
BTW, at 1.5W your amp just working without dead time at all. Indeed learning is the best way for you, don't worry about english, I'm sure you can get right chinese books even.:D |
ok
at 50w rms to 8ohm output :
mhzpower-4 : thd+n=0.009% at 1k, thd+n =0.01% at 6k (80k test bandwith)
then you can make a 100khz 10x better than us :)
so you need to proof: at 50w to 8ohm
thd+n=0.0009% at 1k and thd+n=0.001% at 6k (20k test bandwidth,
because you can't test 100k class-d at a 80k bandwidth at AP system ,you need a 22k filter pluse in,
same reason, you can't test true THD+N at 10k)
ok please show to us
rg
fumac |
|
|
| soongsc |
| I think it would be interesting to see how audio amplifiers will perform at 0.1W into 8 ohm load. Is there any reason why manufacturers don't publish data at such levels? |
|
|
| Pafi |
fumac!
Eva is right, but you have to understand that she was talking about THD before feedback. When you have feedback, THD goes lower, and because higher switching freq allows higher loop gain, THD returns to similar level as with lower switching freq.
So basically very high switching freq has not good, nor bad influence on THD.
P.S.: THD is not THD+n! You can't expect 0,00035 % THD+n at -20 dB, because noise of any device is already higher, but Eva was talking about only distortion! On the other hand: 0,00035 % THD at this level is achievable (however very difficult to measure). |
|
|
| fumac |
| quote: | Originally posted by soongsc
I think it would be interesting to see how audio amplifiers will perform at 0.1W into 8 ohm load. Is there any reason why manufacturers don't publish data at such levels? |
because the noise floor ,
we can guess the noise floor at the 10mW point of Curve of thd+n vs output power
if an AMP have a low thd+N at low power output (10mW)
that means the noise floor is also very low.
if an amp have a high thd+n at low power output
that means the noise floor is high .
the low floor noise can give us very good dynamic.
we can find the Curve of thd+n vs power same like a "V"
the mini No. of THD+N at the Curve of thd+n vs power
the left side is N>THD, the right side is THD>N
this is the datasheet of mhzpower-2
you can read the THD+N of mini power ,5mW at 8ohm or 10mW at 4ohm
http://www.class-d.net/mhzpower-2.pdf
rg
fumac |
|
|
| fumac |
| quote: | Originally posted by Pafi
fumac!
Eva is right, but you have to understand that she was talking about THD before feedback. When you have feedback, THD goes lower, and because higher switching freq allows higher loop gain, THD returns to similar level as with lower switching freq.
So basically very high switching freq has not good, nor bad influence on THD.
|
yes , eva is right, infact i agree with her talking from books
but , can you tell me , why most of good class-d amp is running at >300khz?
and not running at 100k?
as she said, she can make a 100k class-d amp 3~4x better than 400k,
ucd-400 THD+N :0.004% 20~20K 1W 8OHM.
EVA need to make a 100k D amp same like :
THD+N :0.001% 20~20K 1W 8OHM.
ok show it to us , we are waiting
| quote: | Originally posted by Pafi
P.S.: THD is not THD+n! You can't expect 0,00035 % THD+n at -20 dB, because noise of any device is already higher, but Eva was talking about only distortion! On the other hand: 0,00035 % THD at this level is achievable (however very difficult to measure). |
yes you are right, thd is not thd +n
infact thd+n is higher than THD.
"very difficult to measure " is not "no one can measure "
rg
fumac |
|
|
| soongsc |
| quote: | Originally posted by fumac
because the noise floor ,
we can guess the noise floor at the 10mW point of Curve of thd+n vs output power
if an AMP have a low thd+N at low power output (10mW)
that means the noise floor is also very low.
if an amp have a high thd+n at low power output
that means the noise floor is high .
the low floor noise can give us very good dynamic.
we can find the Curve of thd+n vs power same like a "V"
the mini No. of THD+N at the Curve of thd+n vs power
the left side is N>THD, the right side is THD>N
this is the datasheet of mhzpower-2
you can read the THD+N of mini power ,5mW at 8ohm or 10mW at 4ohm
http://www.class-d.net/mhzpower-2.pdf
rg
fumac |
Looks more and more interesting!
Looks like the Power supply section takes the transformer outputs rather than DC? I've found that transformer performance really has a very very significant effect on class-D amp performance, why not just design a SMPS instead? |
|
|
| soongsc |
| quote: | Originally posted by fumac
yes , eva is right, infact i agree with her talking from books
but , can you tell me , why most of good class-d amp is running at >300khz?
and not running at 100k?
as she said, she can make a 100k class-d amp 3~4x better than 400k,
ucd-400 THD+N :0.004% 20~20K 1W 8OHM.
EVA need to make a 100k D amp same like :
THD+N :0.001% 20~20K 1W 8OHM.
ok show it to us , we are waiting
yes you are right, thd is not thd +n
infact thd+n is higher than THD.
"very difficult to measure " is not "no one can measure "
rg
fumac |
I think we should make the tone less provoking. Eva has shared much of her work on this forum. I have learned a lot from her posts.
Anyway, I think GedLee mentioned some techniques to measure through the noise which involved syncronizing and averaging data to reduce noise effects, hope we see some tools out there that do that in the future. |
|
|
| Pafi |
fumac!
| quote: | if an amp have a high thd+n at low power output
that means the noise floor is high |
Not really. This means that either noise or distortion is high.
And actually noise floor is a different thing, it is used to describe spectral density of noise.
In your datasheet once you say 1 MHz sampling freq, then 1 MHz switching freq, however they are not the same! In one period a ClassD amp takes 2 samples! 1MHz=2MS/s!
| quote: | EVA need to make a 100k D amp same like :
THD+N :0.001% 20~20K 1W 8OHM. |
This was what I was talking about : she don't need, because she told that about only power section, not the entire amp with feedback! |
|
|
| fumac |
| quote: | Originally posted by soongsc
Looks more and more interesting!
Looks like the Power supply section takes the transformer outputs rather than DC? I've found that transformer performance really has a very very significant effect on class-D amp performance, why not just design a SMPS instead? |
about smps +class d ,
many people like to try it ,
but now , most of the smps +class d at the world can't get a very good measure to us.
because the noise floor is higher than translator + class-d.
this is also happened at smps + class-AB
you can easy download the datasheet of one Famous smps+class-d module.
and you can take a look at the thd+n vs power.
then you can get the answer.
rg
fumac |
|
|
| fumac |
| quote: | Originally posted by Pafi
fumac!
Not really. This means that either noise or distortion is high.
And actually noise floor is a different thing, it is used to describe spectral density of noise.
In your datasheet once you say 1 MHz sampling freq, then 1 MHz switching freq, however they are not the same! In one period a ClassD amp takes 2 samples! 1MHz=2MS/s!
This was what I was talking about : she don't need, because she told that about only power section, not the entire amp with feedback! |
thanks for your Advice, and we will change the word: samples to switching freq.
but , she need to proof her saying, because she said : 100k is 10x better than 1mhz.
just remenber, don't make a bassboost :) or just stop at talking!
a class-d amp for hifi, not just a class-d technoledge can do,
and no one just heard about the "only power section" (as you said)
we need a full design of hifi audio, not just a "power section"
we need to know about that the Target is hifi audio,
the Target is not just class-d technolege.
talking about any AMPs is not just talking about one side of it , it is a full section. you can't seperate them just like one setion or the other setions.
we need to balance all the setions then we can get a very good amp.
we are waiting
talking is quite easy then doing
thanks
rg
fumac |
|
|
| Pafi |
| quote: | | she said : 100k is 10x better than 1mhz |
Where? In what aspect?
| quote: | | no one just heard about the "only power section" (as you said) |
I don't agree with you fully, because there are some amps without feedback, where you can hear the quality of power section directly, but this is not really important. Analyzing power section alone is part of a method of understanding and designing amp. The fact that nobody ever seen a real circle (in mathematical meaning), doesn't mean that PI is meaningless thing!
| quote: | | you can't seperate them |
Separation is one of the most useful thing in science and engineering.
| quote: | | we need to balance all the setions |
Without separation you can't balance consciously. |
|
|
| fumac |
| quote: | Originally posted by Pafi
Where? In what aspect?
I don't agree with you fully, because there are some amps without feedback, where you can hear the quality of power section directly, but this is not really important. Analyzing power section alone is part of a method of understanding and designing amp. The fact that nobody ever seen a real circle (in mathematical meaning), doesn't mean that PI is meaningless thing!
Separation is one of the most useful thing in science and engineering. |
hi Pafi
please read this topic again,
many topic about feedback and no feekback
also at clas-AB amps
but to now , noone agree each other.
does feedback really make the sound worst than no feedback?
feedback is one technoledge at morden design. it is used any where,
if continue talk about feed back, will begin another fighting.
dont forget, most of moden class-d have feedback loops.
and THD not just from deadtime , also will product THD from the output filter.
perhaps we need to let tekko keep on his topics.
sorry tekko ,
i think you can make your class d better and better.
just keep trying. you are same like us at that days.
rg
fumac |
|
|
| Pafi |
fumac!
Sorry, I can't read the whole topic again for this!
Feedback is neccessary, I agree, I told just: 1.: still there are some ClassD amps without it, 2.: we can separate it in mind for analysing purposes. |
|
|
| soongsc |
| quote: | Originally posted by fumac
because the noise floor ,
we can guess the noise floor at the 10mW point of Curve of thd+n vs output power
if an AMP have a low thd+N at low power output (10mW)
that means the noise floor is also very low.
if an amp have a high thd+n at low power output
that means the noise floor is high .
the low floor noise can give us very good dynamic.
we can find the Curve of thd+n vs power same like a "V"
the mini No. of THD+N at the Curve of thd+n vs power
the left side is N>THD, the right side is THD>N
this is the datasheet of mhzpower-2
you can read the THD+N of mini power ,5mW at 8ohm or 10mW at 4ohm
http://www.class-d.net/mhzpower-2.pdf
rg
fumac |
A little calculation says 5mW is 49db below 400W, is the dynamic range of power amps that bad? Such that noise has effect on test levels 49db below max? |
|
|
| Tekko |
My stereo amp is now undergoing a major rebuild power rail wise. From beeing a single supply amp, it is beeing rebuilt to accomodate split supply, just a quick test without changes to the board reveals that it works much better than running single supply.
Soldering station, here i come:smash: |
|
|
| fumac |
| quote: | Originally posted by soongsc
A little calculation says 5mW is 49db below 400W, is the dynamic range of power amps that bad? Such that noise has effect on test levels 49db below max? |
you calculation wrong,
at 4ohm:
the thd+n at 10mW is 0.015%,
and thd+n at 400W is 1%
not just can output 5mW mini.
calculat again please
rg
fumac |
|
|
| fumac |
| quote: | Originally posted by Tekko
My stereo amp is now undergoing a major rebuild power rail wise. From beeing a single supply amp, it is beeing rebuilt to accomodate split supply, just a quick test without changes to the board reveals that it works much better than running single supply.
Soldering station, here i come:smash: |
yes, seperate power supply is better than single one.
because the low pssr of class-d
rg
fumac |
|
|
| Pafi |
fumac!
soongsc's calculation is not wrong, just he calculated a different thing from you (didn't). 10*lg(400/0.005)=49 indeed. (The question is what did he want to tell with this. Maybe that THD should be specified at low power too.)
| quote: | | because the low pssr of class-d |
Class-D don't have PSRR. It's only a Class! Amplifiers have. Eg. "MHz amplifiers" have low PSRR, some others (eg. Hypex UcD) have higher PSRR. |
|
|
| Tekko |
| My MHz amp hums, even tho its behind dual regulated supplies, yes the power is regulated twice, first to 2x15v then to 2x8v, but it still hum when i activate bass boost on the preamp. |
|
|
| fumac |
| quote: | Originally posted by soongsc
A little calculation says 5mW is 49db below 400W, is the dynamic range of power amps that bad? Such that noise has effect on test levels 49db below max? |
| quote: | Originally posted by Pafi
fumac!
soongsc's calculation is not wrong, just he calculated a different thing from you (didn't). 10*lg(400/0.005)=49 indeed. (The question is what did he want to tell with this. Maybe that THD should be specified at low power too.)
|
i don't know what are you saying
do you caculate a dynamic range as mr. song?
that's right? so funny
-----------------
Audio
The dynamic range is defined as the difference between the minimum and maximum amplitude a given device can record. For example, if the ceiling of a device is 10 dB and the noise floor is 3 dB then the dynamic range is 4.85 dB, since 10 dB−3 dB = 4.85 dB (recall that care must be taken when adding numbers in the decibel scale).
------------------
from wikipedia
http://en.wikipedia.org/wiki/Dynamic_range
Mr song:
you need to caculate the max output voltage and the noise floor level,
then you can get the dynamic range.
5mW, not the noise floor,
5mw also not the minimum amplitude at MCD.
you can find the noise floor picture or No. at our datasheet.
we have posted more test pictures than others.
mhzpower-2 we got a noise floor at 22uV (tested by ap syetem. )
22uV to 4ohm is 121pW, ----this is the noise power output.
you can do this caculate:
the thd+n= 0.015% at 10mW 4ohm
10mW to 4ohm is 0.2V rms
0.2vX0.015%=0.00003V=30uV
is meaning :THD+N is 0.00003V. also is 30uV
------------------
30uV is THD+N, 22uV is N, so , you can verify them yourself.
rg
fumac |
|
|
| soongsc |
| quote: | Originally posted by Pafi
fumac!
soongsc's calculation is not wrong, just he calculated a different thing from you (didn't). 10*lg(400/0.005)=49 indeed. (The question is what did he want to tell with this. Maybe that THD should be specified at low power too.)
Class-D don't have PSRR. It's only a Class! Amplifiers have. Eg. "MHz amplifiers" have low PSRR, some others (eg. Hypex UcD) have higher PSRR. |
That is exactly what I'm trying to understand. I my guess is that most amplifiers on the market will show THD+n for higher levels, and mostly in the linear range. My guess is, showing THD+N in the lower level range such as 0.1W or even 5mW, if the amplifier dynamic range is good, noise level would still be maybe 40db lower than the signal, and thus have little effect on measurements. If this is true, then the THD + N measurement at 5mW would represent how linear the amplifier is at low level signals, which should be a good indication how good the amp will reproduce detail information.
Lower PSRR in class D amps would mean that the efficiency is high. Would this be a correct assessment? Thus the more effecient amps rely on better power supply to obtain best performance. |
|
|
| soongsc |
What I am trying to say is that at 5mW, the noise would not be a dominating factor for the THD + N figure. Rather linearity is still the most dominating factor. Other specs being equal, the better THD + N figure at low levels would indicate better performance in revealing detail.
I think the data you have posted is much thorough than what other manufacturers care to provide, this is a very good start. Lets hop other companies can post more challenging data.| quote: | Originally posted by fumac
i don't know what are you saying
do you caculate a dynamic range as mr. song?
that's right? so funny
-----------------
Audio
The dynamic range is defined as the difference between the minimum and maximum amplitude a given device can record. For example, if the ceiling of a device is 10 dB and the noise floor is 3 dB then the dynamic range is 4.85 dB, since 10 dB3 dB = 4.85 dB (recall that care must be taken when adding numbers in the decibel scale).
------------------
from wikipedia
http://en.wikipedia.org/wiki/Dynamic_range
Mr song:
you need to caculate the max output voltage and the noise floor level,
then you can get the dynamic range.
5mW, not the noise floor,
5mw also not the minimum amplitude at MCD.
you can find the noise floor picture or No. at our datasheet.
we have posted more test pictures than others.
mhzpower-2 we got a noise floor at 22uV (tested by ap syetem. )
22uV to 4ohm is 121pW, ----this is the noise power output.
you can do this caculate:
the thd+n= 0.015% at 10mW 4ohm
10mW to 4ohm is 0.2V rms
0.2vX0.015%=0.00003V=30uV
is meaning :THD+N is 0.00003V. also is 30uV
------------------
30uV is THD+N, 22uV is N, so , you can verify them yourself.
rg
fumac |
|
|
|
| fumac |
| quote: | Originally posted by Pafi
fumac!
Class-D don't have PSRR. It's only a Class! Amplifiers have. Eg. "MHz amplifiers" have low PSRR, some others (eg. Hypex UcD) have higher PSRR. |
Pafi
do you have tested other class-AB and class-A amp?
how about the PSRR of class-A amp?
when you know the PSSR of class-A , you will know all the class-d to now have a poor PSRR .
your point out that the class-d amp , they have pssr at 65db (from the datasheet of them)
our mcd is 71db, which one is better?
65db and 71db are poor also.
many many class-A power amp have a >120db PSSR
please take a look at the datasheet of 3886
http://www.jaycar.com.au/images_uploaded/LM3886.PDF
take a look at this 1$ chip, it have a 120db PSSR. not 65db also not 71db.
and this link will told you about :
Closed-loop Class-D power amp delivers improved sound, PSRR
http://www.automotivedesignline.com/products/206900612
if you do know class-d , and comparator class-d with class-ab and class-A,
you will never told others, very good PSSR at class-d.
any class-d have a poor PSSR now. including Closed-loop Class-D.
PSSR of Closed-loop Class-D is better than the openloop class-d,
but also not very good
please comparater with class-A or AB,not just stare at the class-d
any class-d will be better when they use seperate powersupply.
this also happened at UCD and MCD
rg
fumac |
|
|
|
