Hi to all
I am under a big enervation from day before yesterday.
I had completed my amplifier from Thursday, and i made with success the tests of auxiliary protection, indication, and EMI suppression circuits. These are:
1) Mains input thermal breaker, Tyco 5A/250Vac--> Pass
2) Mains switch push on type DPDT, ALPS 4A/250Vac. Accross each pair of switch contacts, are soldered two 100nF-X2 caps to suppress spikes and thumps during on/off.--> Pass
3) For inrush current limitting of the 500VA toroid xformer, i use a Rhopoint "surge gard" type NTC which is shorted after 0,5 sec from a 30A rated Matsushita SPNO relay--> Pass
4) For output protection (either speaker or amp. output from short circuit) is used a 30A rated Matsushita SPNO relay, one per channel.
a) Turn on delay predicted at 3 sec--> Pass
b) DC level for output disconnection predicted at 8Vdc--> Pass
c) Thermal protection managed by a LM35DT mounted on main heatsink predicted at 80oC--> Pass
d) Current limitters (except the incorporated VI limitter into each channel pcb) of output transistors seperate for PNP and NPN rows, predicted at 2,5A per transistor--> Pass
e) VAS signal versus output signal comparator for tracking differences (i.e. output heavy distortion) predicted at 1,5Vrms--> Pass
Why all of that? Because this amp. supplied with +/-60Vdc (individual per each channel, there is NOT contact between the gnd nodes of the two channels) from a 500VA xformer and 28000ìF smoothing caps per channel, can produce a massive power of 190Wrms/8Ù per channel. It is not a toy... because to its dynamics can be as well a killer for expensive loudspeakers in cases of distortion or malfunction. I respect the speakers of people. Moreover, the protection circuit is fully buffered from the audio path and it is fully transparent.
And now the bad news... The 100Hz noise produced by the bridge rectifiers, it pass through the audio path of amplifier in its output. I have tried everything in the bridges... 10nF Murata caps accross each terminal... only one accross the xformer secondary AC terminals... caps of different value, ceramic, MKT, MKP... but nothing.
It should be pointed, that i use a big 15" PA speaker of Peavey (350Wrms, 103dBSPL-1W) to be sure that i hear everything possible noise. The buzz of 100Hz is heared from this speaker enough clear.
Some indicating values taken with the scope (AC coupled) respect to gnd are:
a) On the supply rails of amp there is presented the 100Hz ripple with a shape of inverted sawtooth and with a value of 168mVpp.
b) After the VAS stage, is reduced significantly and finally presented in output not as sawtooth, but as a line with debounches of small period (at the same frequency of 100Hz) which off the amplitude is 7-8mVpp. The output offset is only 2-3 mV (added the noise). Although the values presented are very small, the annoying noise generated from the bridge rectifiers (i know, i know... it is also the current throw back to windings of xformer etc..etc) it is audible.
In the amp board, i use one 100ìF elec. and one 100nF MKP per rail for decoupling. I tried also to add more or less capacitance for this, but nothing again. I changed the 1000ìF cap between feedback and gnd with a 100ìF... the same.
The toroid that i use it is custom maden for me. It is fully dipped in a metal case with rosin. Its case it is firmly grounded on the mains earth.
I started to think maybe the problem it is the very large OL gain of amp in low frequencies. Maybe the CCS (individual for LTP and VAS) does not operate in 100% of their possibility.
The project it is an all bjt builded AB class amplifier with 50mA Iqs per output device. It used single LTP mirrored in input, VAS a la D.Self, and tripple darlington EF output stage.
Any advice for the poor (according to my hearing) PSR?
Thanks
Fotios
I am under a big enervation from day before yesterday.
I had completed my amplifier from Thursday, and i made with success the tests of auxiliary protection, indication, and EMI suppression circuits. These are:
1) Mains input thermal breaker, Tyco 5A/250Vac--> Pass
2) Mains switch push on type DPDT, ALPS 4A/250Vac. Accross each pair of switch contacts, are soldered two 100nF-X2 caps to suppress spikes and thumps during on/off.--> Pass
3) For inrush current limitting of the 500VA toroid xformer, i use a Rhopoint "surge gard" type NTC which is shorted after 0,5 sec from a 30A rated Matsushita SPNO relay--> Pass
4) For output protection (either speaker or amp. output from short circuit) is used a 30A rated Matsushita SPNO relay, one per channel.
a) Turn on delay predicted at 3 sec--> Pass
b) DC level for output disconnection predicted at 8Vdc--> Pass
c) Thermal protection managed by a LM35DT mounted on main heatsink predicted at 80oC--> Pass
d) Current limitters (except the incorporated VI limitter into each channel pcb) of output transistors seperate for PNP and NPN rows, predicted at 2,5A per transistor--> Pass
e) VAS signal versus output signal comparator for tracking differences (i.e. output heavy distortion) predicted at 1,5Vrms--> Pass
Why all of that? Because this amp. supplied with +/-60Vdc (individual per each channel, there is NOT contact between the gnd nodes of the two channels) from a 500VA xformer and 28000ìF smoothing caps per channel, can produce a massive power of 190Wrms/8Ù per channel. It is not a toy... because to its dynamics can be as well a killer for expensive loudspeakers in cases of distortion or malfunction. I respect the speakers of people. Moreover, the protection circuit is fully buffered from the audio path and it is fully transparent.
And now the bad news... The 100Hz noise produced by the bridge rectifiers, it pass through the audio path of amplifier in its output. I have tried everything in the bridges... 10nF Murata caps accross each terminal... only one accross the xformer secondary AC terminals... caps of different value, ceramic, MKT, MKP... but nothing.
It should be pointed, that i use a big 15" PA speaker of Peavey (350Wrms, 103dBSPL-1W) to be sure that i hear everything possible noise. The buzz of 100Hz is heared from this speaker enough clear.
Some indicating values taken with the scope (AC coupled) respect to gnd are:
a) On the supply rails of amp there is presented the 100Hz ripple with a shape of inverted sawtooth and with a value of 168mVpp.
b) After the VAS stage, is reduced significantly and finally presented in output not as sawtooth, but as a line with debounches of small period (at the same frequency of 100Hz) which off the amplitude is 7-8mVpp. The output offset is only 2-3 mV (added the noise). Although the values presented are very small, the annoying noise generated from the bridge rectifiers (i know, i know... it is also the current throw back to windings of xformer etc..etc) it is audible.
In the amp board, i use one 100ìF elec. and one 100nF MKP per rail for decoupling. I tried also to add more or less capacitance for this, but nothing again. I changed the 1000ìF cap between feedback and gnd with a 100ìF... the same.
The toroid that i use it is custom maden for me. It is fully dipped in a metal case with rosin. Its case it is firmly grounded on the mains earth.
I started to think maybe the problem it is the very large OL gain of amp in low frequencies. Maybe the CCS (individual for LTP and VAS) does not operate in 100% of their possibility.
The project it is an all bjt builded AB class amplifier with 50mA Iqs per output device. It used single LTP mirrored in input, VAS a la D.Self, and tripple darlington EF output stage.
Any advice for the poor (according to my hearing) PSR?
Thanks
Fotios
keantoken said:
Hi keane
Please, no... not in common view...
I started this thread after enough thought, because usually i resolve such type issues by alone in a couple of days (i am used in 48 hours continuous work without sleep) but now i am some despaired because the time is against me. This project it is a vital item for my economics, because i am in debts from it and the customer is in waiting for delivering the amplifier. I make this job for living... i ask your comprehension but i don't want - for the above mentioned reasons - to turned the thread in an endless discussion about this and that. Please, respect my difficult place. Let me finish the job, and after we have enough time for post and discussion about the circuitry.
I have a hope, that an expertised person here, he had the same problem like me and thus maybe has the ready solution, or the right suggestion.
Anyway, if you or anyone else want to see the schematic, drop me a link to send you the pdf or you can see it in the forum of lineup. And forget SPICE etc. These problems resolved only in workbench with tools and instruments.
Regs
Fotios
can be helpfull?
I find the notepad where i had noted some measurements with my gen and my scope:
1) Open Loop gain: for source used a sinus wave of 20mVpp. These are the values obtained in output under 8Ù dummy.
100Hz > 120Vpp (clip, supply rails exceeding) > 75,5dB
1KHz > 120Vpp (clip, supply rails exceeding) > 75,5dB
10KHz = 17,6Vpp = 58,9dB
20KHz = 10,4Vpp = 54,32dB
2)Closed Loop gain: source used a sinus wave of 2,6Vpp. Output under 8Ù dummy:
20Hz = 115Vpp = 32,9dB
100Hz = 115Vpp = 32,9dB
1KHz = 115Vpp = 32,9dB
10KHz = 115Vpp = 32,9dB
20KHz = 115Vpp = 32,9dB
50KHz = 111Vpp = 32,6dB
100KHz = 103Vpp = 31,9dB
150KHz = 90Vpp = 30,8dB (this is about the -3dB point)
Fotios
I find the notepad where i had noted some measurements with my gen and my scope:
1) Open Loop gain: for source used a sinus wave of 20mVpp. These are the values obtained in output under 8Ù dummy.
100Hz > 120Vpp (clip, supply rails exceeding) > 75,5dB
1KHz > 120Vpp (clip, supply rails exceeding) > 75,5dB
10KHz = 17,6Vpp = 58,9dB
20KHz = 10,4Vpp = 54,32dB
2)Closed Loop gain: source used a sinus wave of 2,6Vpp. Output under 8Ù dummy:
20Hz = 115Vpp = 32,9dB
100Hz = 115Vpp = 32,9dB
1KHz = 115Vpp = 32,9dB
10KHz = 115Vpp = 32,9dB
20KHz = 115Vpp = 32,9dB
50KHz = 111Vpp = 32,6dB
100KHz = 103Vpp = 31,9dB
150KHz = 90Vpp = 30,8dB (this is about the -3dB point)
Fotios
dudaindc said:That is weird... it would be interesting putting the caps back (one by one) just to check when noise returns.
I know you are pressed for time - it was just an idea.
Well done Fotios!
Thanks dudaindc
To be sincere, at this moment i have shut down everything in my workbench, and i celebrate the disappearence of this devil that i had over my head for 2 days, by drinking vodka.
As i said in each amp board there was one 100ìF elec and one 100nF MKP per rail for decoupling. From the successive trial and error efforts, the only that i remember it is that when i removed only the elec. caps, the problem has remained. I have not tried to keep removed only the MKP and to place again the elec. in their place. I promise to will try tomorrow this test.
Also me i am surprised from this unexpected incident!
Regs
Fotios
Hey, look at this.
Would this have something to do with it?
http://www.diyaudio.com/forums/attachment.php?s=&postid=1677292&stamp=1228549975
On a PCB it is important to make star ground so that no noise is injected into the ground paths as a result of the huge sinking currents from the rectifier. The ground point should be exactly in the middle between the two caps.
- keantoken
Would this have something to do with it?
http://www.diyaudio.com/forums/attachment.php?s=&postid=1677292&stamp=1228549975
On a PCB it is important to make star ground so that no noise is injected into the ground paths as a result of the huge sinking currents from the rectifier. The ground point should be exactly in the middle between the two caps.
- keantoken
Re: I resolved the problem!
In a monoblock this can be done on the PCB.
In a multichannel amp it requires separate signal ground and power ground returns to a central audio ground ( or two audio grounds if you choose).
fotios said:I removed all the decoupling caps from the amplifier boards - electrolytics and MKP - and the 100Hz noise from the bridges, has been removed completelly from the audio signal!
Regs
Fotios
you must keep the dirty ground from the decoupling caps separate from the clean signal ground.roender said:
In a monoblock this can be done on the PCB.
In a multichannel amp it requires separate signal ground and power ground returns to a central audio ground ( or two audio grounds if you choose).
Problems... problems...
Yes, Andrew comment is right. To not destroy my pcbs, simply i removed the dirty parts from the gnd tracks. Moreover their existence was unnecessary by some way, because the cables from power supply board to amp boards are very short. I left only the small 100nF MKP for decoupling.
The noise had reduced signifficantly after this, but a little ammount still remained. In such cases, where the scope is placed in 10mV scale because the signal has very small amplitude, you can't understand nothing with your eyes from the screen indication. In this case, i use my ears and a big power and very sensitive PA speaker (103dB/1W). This reveals every noise, be it so very little. I suggest you this method to hear if your amplifier have except the hiss caused from parts, also buzz or crackling noise.
For example, although the noise measured with DSO is only 3,8 mVrms and 7,3 mVrms respectivelly for left and right channel, in silence it is heared from this speaker. I have two integrated amplifiers of 2 X 50Wrms each for comparisson with my project. From these is heared only the hiss. But they have little gain compared to Av=39 (1,1Vrms in--->42,4Vrms out) of mine project. Thus the comparisson is unreliable. I don't know how much acceptable is the buzz or crackling noise for such bigger power projects. From PA amps, it is heared but it does no matters.
Tell me please your experience from your projects.
I quote a picture to get a better idea.
To be continued.
Fotios
Yes, Andrew comment is right. To not destroy my pcbs, simply i removed the dirty parts from the gnd tracks. Moreover their existence was unnecessary by some way, because the cables from power supply board to amp boards are very short. I left only the small 100nF MKP for decoupling.
The noise had reduced signifficantly after this, but a little ammount still remained. In such cases, where the scope is placed in 10mV scale because the signal has very small amplitude, you can't understand nothing with your eyes from the screen indication. In this case, i use my ears and a big power and very sensitive PA speaker (103dB/1W). This reveals every noise, be it so very little. I suggest you this method to hear if your amplifier have except the hiss caused from parts, also buzz or crackling noise.
For example, although the noise measured with DSO is only 3,8 mVrms and 7,3 mVrms respectivelly for left and right channel, in silence it is heared from this speaker. I have two integrated amplifiers of 2 X 50Wrms each for comparisson with my project. From these is heared only the hiss. But they have little gain compared to Av=39 (1,1Vrms in--->42,4Vrms out) of mine project. Thus the comparisson is unreliable. I don't know how much acceptable is the buzz or crackling noise for such bigger power projects. From PA amps, it is heared but it does no matters.
Tell me please your experience from your projects.
I quote a picture to get a better idea.
To be continued.
Fotios
An externally hosted image should be here but it was not working when we last tested it.
dudaindc said:I always enjoy looking at pictures of your projects Fotios - they are really neat!
Thanks dudaindc for your kind words.
I have builded this unit in my home with my tools at whole. I use a mitre shaw which includes also a table for parallel cutting of large alu pieces. I have also builded a hand bending machine. For cuttings inside the plates (like square holes) i use a hand operated milling machine. And the finish brussing, with... hand.
Álways i try to do the best, for those things depend on my hands. For the rest, the responsible is the supplier. Look the toroid for example. It is custom maden for me. It is rosin dipped and enclosed in metal case. I paid 75 euros for this 500VA xformer with individual supplies per each channel. I have arranged the rest boards based on this - supposed - noiseless xformer. But the crackling still exists, as well a small electromagnetic field.
I build each amp board and i make the appropriate tests and callibrations by using my big lab power supply. Unfortunatelly, when the boards placed iside the unit and connected to the power supply, then starts the problems. I was forced to connect the gnd nodes of the 2 channels with 10R resistors in chassis for earthing. The left channel is most noisy, but when i place the bottom cover (from 3mm thick alu) the noise moved to right channel! Amazing! This caused from the electromagnetic field of toroid. To resolve this, after enough search i have soldered a small piece of pcb vertically in the gnd track for screening. You can see it on the right channel.
I had also some oscillation problem with the screened signal cables that i used. What to hell! I don't know why, but after replacing them, the problem resolved. The screened cables!
It is a big story my friend - and friends - which i will continue tomorow.
Regs
Fotios
PCB layout problems?
Forgot to use independent signal and power grounds for each channel? (joined by 10 ohms or so to force coupling at RF despite wire inductance).
In the chipamp forum there are usually some threads dealing with this.
BTW: If you leave the decoupling capacitors out, you may run into instability problems instead.
Transformer magnetic humm is picked up by large signal loops (poor layout again).
Construction is nice but very bulky for today's standards. Consider that 3000W@4 ohms class D mono may be packed into 30x20x7cm.
Forgot to use independent signal and power grounds for each channel?
(joined by 10 ohms or so to force coupling at RF despite wire inductance).In the chipamp forum there are usually some threads dealing with this.
BTW: If you leave the decoupling capacitors out, you may run into instability problems instead.
Transformer magnetic humm is picked up by large signal loops (poor layout again).
Construction is nice but very bulky for today's standards. Consider that 3000W@4 ohms class D mono may be packed into 30x20x7cm.
As i said, indeed it is mistake the common gnd lines. I take my lesson for today Eva
To be sincere, i have a big failing. From the begining i try to built a complette unit. You can see the two back boards with the discrete balanced to single ended converters, as well the input attenuation pots. Also those does not work well.
Next time, i will build the same unit only in two pcbs.
To not throw so much work in the trash bin, i have spent one week to resolve problems with tricks. Untill now i have dropped the output noise - mainly crackling - to 5 mVrms. In domestic speakers with less sensitivity, you may stick your ear in the cone of woofer under absolute silence, to hear a very low noise from xformer.
As for instability, the two remained 100nF MKP are enough by some way. In bigger power PA amplifiers like Peavey CS1200 used only four 220nF MKT per rail for decoupling. Also, in the back of central board there is an Error Analog Computer which deactivate instantly the output relays in the case of a spike which exceeds the 1,2Vrms like the unplugging of an input plug during operation.
Finally, we speak here for a class AB amplifier for domestic use and not for a class D PA amplifier for live stages. If i turn the amp pcbs vertically in relation with xformer there is not any interference and the amp it is verry quite. But i placed horizontally the pcbs for easiness in case of repairing.
Regs
Fotios
To be sincere, i have a big failing. From the begining i try to built a complette unit. You can see the two back boards with the discrete balanced to single ended converters, as well the input attenuation pots. Also those does not work well.
Next time, i will build the same unit only in two pcbs.
To not throw so much work in the trash bin, i have spent one week to resolve problems with tricks. Untill now i have dropped the output noise - mainly crackling - to 5 mVrms. In domestic speakers with less sensitivity, you may stick your ear in the cone of woofer under absolute silence, to hear a very low noise from xformer.
As for instability, the two remained 100nF MKP are enough by some way. In bigger power PA amplifiers like Peavey CS1200 used only four 220nF MKT per rail for decoupling. Also, in the back of central board there is an Error Analog Computer which deactivate instantly the output relays in the case of a spike which exceeds the 1,2Vrms like the unplugging of an input plug during operation.
Finally, we speak here for a class AB amplifier for domestic use and not for a class D PA amplifier for live stages. If i turn the amp pcbs vertically in relation with xformer there is not any interference and the amp it is verry quite. But i placed horizontally the pcbs for easiness in case of repairing.
Regs
Fotios
Eva said:
Some people can do chipams, others can do class A or AB transistor amps or tube amps and others again have the ability and the park of measuring instruments to design state of the art, tightly packed high-end class D amps.
Today's possibilities are not necessarily today's standards. Class D will probably be the industry standard in the not too far future, partly thanks to your work, too. Today it is still the class AB transistor amp by the numbers.
Given the demands on proper layout and advanced knowledge of EMI effects plus the necessary measuring instruments for successful class D design, class A or AB will probably remain strong in the DIY community for a long time still. And when class D finally takes over, class A and AB transistor amps will find their aficionados to maintain the technology alive, just like tube amps did.
Saludos
David
Eva said:PCB layout problems?
Forgot to use independent signal and power grounds for each channel?
In the chipamp forum there are usually some threads dealing with this.
BTW: If you leave the decoupling capacitors out, you may run into instability problems instead.
Transformer magnetic humm is picked up by large signal loops (poor layout again).
As proved finally, this issue can’t be resolved with conjectures and such things. After one week of detailed examination on the workbench, I have resulted in the following:
1. Big electrolytic caps (100-220ìF) are not necessary for decoupling the ripple of supply rails. Instead, they can multiply the ripple as I have seen in the screen of my DSO. One small MKP of 100 to 220nF per rail can do the job just fine – at least in my circuit - without instability problems.
2. PCB layout is just fine and very carefully designed without gnd loops. The problem caused from the PCB place against xformer. Unfortunatelly i entrusted the performance of this hypothetical super low noise toroid, and i placed the PCBs as shown in the picture. If the PCBs turned vertically, then the buzz noise dissapears complettelly.
3. It is unacceptable for a so much armoured toroid to produce EMI. And the guilty which pick-ups the xformer noise finally, it is the large 4,7ìF polypropylene foil (SCR) caps and not any dirty gnd. I remember one time when Chris “Anatech” he said that a big MKP foil capacitor can be as well an inductor.
4. According these all, I removed the gnd node of each channel connected via the 10R resistor to chassis earth. Now the channels are again individual between them and there is not any increase in buzz noise. Also I decided for easiness to build again the two small back boards where I will place also the input DC blocking MKP capacitor, far away from xformer. Fortunately I have many data sheets and service manuals (mainly these of Bryston are very helpful) to get ideas or to make comparisons with mine project.
5. The problem obviously caused from the EMI of toroid. When the amplifier is without his top and bottom covers (which are made from 3mm thick alu plate) the noise measured at 6 mVrms is presented only in the left channel, while the right is quite. When mounted the bottom cover, the noise moved to right channel, while the left is absolutely quite. When mounted also the top cover, the noise shared in both channels… a little more in left channel. The level of noise it is between 5 to 7 mVrms. With domestic speakers you must stick your ear in the woofer cone to hear it.
In the past, when I constructed big power PA amplifiers, I used a small size 22ìF electrolytic cap in input and a 1KVA monster EI type xformer with individual secondary windings per each channel. I never had buzz noise problems. I have 2 such type xformers yet today. Those are screened with two copper foils, one inside between primary and secondary windings, and one outside around the whole core. Of course the foils are grounded to mains earth. Except this the transformers have dipped in varnish bath. Their core is unbelievable compact. I tried one today besides the PCB and no one EMI noise exists (either buzz or crackling). From the other hand, I must to admit that these EI xformers are more expensive from toroids by 20%. For example, this one included in my amp, it is rated at 500VA and costs 75 euro. For a similar EI (but with 600VA rated core due to the standard dimensions of bobbin) I must pay 90 euro. From the other hand, the extra cost of EI xformer I think is reasonable because the bigger number of parts used for its construction. The issue finally it is that, either toroid or EI xformer if it is constructed with big care, then there is not any issue with EMI or Crackling noise.
Best regards
Fotios
pacificblue said:
I think this story with class D amps, will be the same like this of music storage media. From vinyl records we had the going to Compact Disk. From Compact Disk we had the going to Flash Drives and MP3. And each time, the sound quality is reduced in favour of easiness.
From the lot of MP3 music that i have heard, when i listen music from my CD player i think that the tweeters of my speakers are destroyed!
Fotios
Eva said:
In PA applications indeed... because their weight and not their sound quality. I know some sound contractors that are using AB class monsters to drive their band-pass subwoofers.
Under 120Hz, class D amps they are to an inferior level - it is a law of physics, not mine - compared to class AB. The lot of 3000W and maybe in one year of 100000W in reality it serves - be it so by 100% - above 100Hz.
This is my own opinion. I don't like to generalize things.
Greetings
Fotios
I have forgot to reffer that, when a channel of my amplifier - as is with only one common gnd track and without decoupling caps - is supplied from my external big lab power supply the noise measured in output it is ONLY 1,2 mVrms! And no one noise is heared from the PA speaker. My lab supply, it has not any decoupling component inside. It is composed from: A 6A variac, which feeds a big 1KVA EI type transformer... then follows a bridge rectifier KBPC3502 and two smoothing capacitors 10000ìF each. How to hell, the amp module when connected WITH 1 METER CABLES to this supply it is so quite, at the moment that the ripple meassured on the rails IT IS ALMOST 1 Vrms!
Oh my!
What can i do?
It is obvious that the amplifier pcbs must be placed far away from this :bs: supposed armoured and noiseless toroid!
What an idiot i was to entrust Greek constructors of toroids. The amp case has 10cm height only, but it is enough for two EI xformers of 250VA each.
Fotios
Oh my!
What can i do?
It is obvious that the amplifier pcbs must be placed far away from this :bs: supposed armoured and noiseless toroid!
What an idiot i was to entrust Greek constructors of toroids. The amp case has 10cm height only, but it is enough for two EI xformers of 250VA each.
Fotios
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.