lm3875tf strange distortions

dbdell

Member
2005-08-03 2:42 pm
have recently finished lm3875 gainclones along with 4780!!!!!
and as expected they sound bloody good however i have started to notice a strange distortion on the 3875s which are the TFs(insulated) at about 150hz (havent measured yet but will post)
it seems to only become apparent when the output actually produces that specific frequency from the source then you get a lower amplitude buzz/ring almost as if the sound engineer put it in but HE DIDNT. this happens at any volume but is less noticable the louder you play them probably due to db and spike protection.
It is particularly annoying when you are testing or demonstarting as it is not immidiately obvious. They are the Brian Bell premium kits mur860 .1%res etc 18-0-0-18 ps with 6.3a per rail. Initially I thought this was some sort of ground fault but doesnt seem so !!!!!! any ideas!!!!
Also my 4780s are of paralell design so 2*4780s giving me about a 100w a pair if i do not tie the sig g together on the seperate chips i get terrible hum. these are monoblocs with seperate mur rectifiers and same tranny.!!!!
Its always the little things that do my head in.............................:confused:
 

dbdell

Member
2005-08-03 2:42 pm
Netlist said:
Did you rule out the speaker? Also, a scope would be handy here.
Is the amp silent with no source connected or input shorted?

/Hugo :)

I have tried a range of good quality speakers from 4 to 16 ohms & the noise on the 3875 only occurs with/shortly after a bass note on all speakers !!! a kind of decay afterwards at a specific hz im guessing about 150-200hz maybe lower i will do a frequency test and post results i dont yet have a scope although it is my next bye!!
thanks for your reply Hugo..
:)
 
Can you tell us a little about the dimensions of the heatsink you are using?

the TF's thermal impedance is 1 C/W higher than the T series -- thus your heat sink requirement is going to be (very roughly) 33% greater surface area given |25 VDC| rails. You may be experiencing a kind of thermal oscillation -- I demonstrated this with my LM4780 bridged design -- with a variac I was able to get the chip to cycle on and off by raising the rail voltage and PDMax.

If you are using Brian's boards there shouldn't be an electronic problem --

one other thing -- if you don't have insulated cables to the input pin use twisted pair, and keep them away (or at right angles to) the power supply leads, etc.
 

dbdell

Member
2005-08-03 2:42 pm
jackinnj said:
Can you tell us a little about the dimensions of the heatsink you are using?

the TF's thermal impedance is 1 C/W higher than the T series -- thus your heat sink requirement is going to be (very roughly) 33% greater surface area given |25 VDC| rails. You may be experiencing a kind of thermal oscillation -- I demonstrated this with my LM4780 bridged design -- with a variac I was able to get the chip to cycle on and off by raising the rail voltage and PDMax.

If you are using Brian's boards there shouldn't be an electronic problem --

one other thing -- if you don't have insulated cables to the input pin use twisted pair, and keep them away (or at right angles to) the power supply leads, etc.
If it was current surely noise will be less prominent at low levels!! I also use exactly the same supply to drive tWo pairs of 4780s
with no issues and definetly more current drive than the 3875s it is immediately noticably when comparing 4780 have much fuller low end but maybe not the finese at the top so i am aiming to use the 3875s to run my tweeters (kef T33)and the 4780s to run my woofers(jmlabs 6.5 poykevlar) actively.
Also my 3875s currently share a sink approx 75 *200mm with 30m fins salvaged from old car amp the 4780s have a massive heatsink (they need it to stop spike) another old car amp 250*300mm ally extrusion (ideal for testing)
These are not final design for heatsinks but are practical for testing I never really notice the 3875 getting hot so i dont feel is an issue i use expensive transfer paste as well . I have also used double and triple shielded cable with care taken in the way grounds terminated on the individual shields have had much experience in car audio with grounding issues so is probably one of my stronger points. caps seem to be polarised correctlu but surely it wouldnt work if they werent!! the noises described in my original post are very difficult to hear but vvvvery annoying when you do so i do not belive it is a major issue just somethig minor.
Once more thanks for any input or replys....
dbDell.......................
 

dbdell

Member
2005-08-03 2:42 pm
peranders said:
Do you use any output filter, zobel? Bad pcb layout, currents in the wrong places?


no output filters and no zobel
i have used various cables and speakers and have same results !
If i use an active filter on the3875 i can lose the noise but this is not ideal eventhough i will be using one i wish to resolve first and the noise is above ideal hz on xover frequencies
Thanks
dbDell:cannotbe:
 

dbdell

Member
2005-08-03 2:42 pm
jackinnj said:
Try decoupling the output with 2,7R, 100nF from the outpin to ground -- an oscillating amplifier will manifest many problems.
im using brians boards and i beleive they have the function built in! 2.7 and 100nf in series?? allready have places on gold boards what wattage should 2.7 be??
thanks
dbDell............