Chipamp for sub, parallel lm3875 or lm3886

[coolbeer]

I have just received word about getting an old Sony TA-VE 610, witch is a "5x100w" surround amplifier. "5x100w" is maby stretching it, therefore the ""'s, as the powerdraw of the unit is 240w. Although this might be good news, as I'm apparently getting a unit that delivers double the wattage out than it uses

The plan for this one is to use the transformator and case(and desoldier the rest and use when needed) to build a sub chipamp for a MTX Thunder 2000.
Now, since the sub is a 4Ohm unit, and the transformator is 240W, I'm thinking of paralleling 4 lm3875's or lm3886's, 4 of em to be shure the chips don't go into freakmode when the omph is on(I'm guessing the sub is dipping lower than 4Ohms over the effective range).

Now, since I don't know the railvoltage of the transformator I'm just collecting information and knowledge about the subject before I get the facts on the table.

I am thinking of running it with a regulated power-rail, as I've heard that it will sound a bit more controlled, but of course, nothing is set in stone just yet.

As for chip choice I'm thinking of the 3886 over the 3875 due to the power, and that the better qualities of the 3875 is missed on a sub(if there are any, this forum seems to be divided on the matter :P). With 4 of the chips in parallel we are looking at 40*4=160W for the 3875 and 38x4=152W for the 3886. Note that theese numbers are minimums, we still don't know what powerrails we have to deal with. But we can see that on minimum power we are well within the reach of the transformator, although double is recommended we are still in the ballpark.

The regulation of the powersupply will probably be done with a lm338 type chip, with some caps tossed on(might find some nice ones in the sony, who knows ).
For rectifying I again might find something inn the sony, and I have a bridge I stole from a PC powersupply aswell.

When it comes to the boards I will probably be making one board for each amp, with the output resistor on the board, so I just have to tie the wires together. This is up to debate though, as I'm not sure how that will work out, been hearing mixed messages about it.

I'm wondering about the quality of lm's bought from e-bay, as they are just about 10$ for a pair. I'm not shure they are original, how much do "real" chips cost?

I think that was everything I have fiddled with so far, I'll update as needed with new information. This will be my first chipamp, maby it's overkill to do 4parallel as a first project, but what's life without a few challenges.

--
Kolbjørn

[coolbeer]

I've been fiddling around with eagle a bit and thrown together something based on things I've read on the forum and on the net.
Please note that the board is not in it's final state as I don't yet know the exact dimensions of the components.
So far the component values are not etched in copper yet, so treat as such.

[pacificblue]

Since the Sony is given with 100 W into 4 Ohm the transformer voltage could be in the correct range for your plan.

Those power dissipation figures are far too pessimistic. The dissipation depends on voltage and load impedance. If you use four ICs in parallel, that dissipation is shared across them. Assuming the rail voltage is just right to achieve 100 W into 4 Ohm, the total power dissipation will probably be around 55-60 W, which means 15 W maximum per IC.

The LM338 may be a bit hard pushed in that application, while at the same time reducing the available voltage for the LM3886 too much.

Here in Germany genuine LM3886s start from 4,30 € upwards.

[coolbeer]

I'm looking at possible values for C4 in the schematics, as this basically is a high pass filter, it is probably advised to have it as big as possible... But I keep reading about peeps using polarized caps at around 10-100uF.
Is that really a good idea, since an audio signal is AC, can we really get away with it due to the low current at that particular point?

There is the possibility of tossing together two electrolytes to create a bipolar cap, what are your thoughts on doing that? And how big should I aim? I got 4700uF's I could wire together

Will post later today when the Sony arrives with info on rails and stuff.

[AndrewT]

c4 & r1 are a high pass filter that sets the bass roll off frequency and blocks DC entering the power amp.
Many suggest F-3dB between 7Hz and 20Hz, but I think it sounds better when set to between 1Hz and 2Hz. That needs R1*C4>=0.08seconds.

Choose R1=R2 and then calculate what value of C4 that gives the F-3dB you decide to set.

C3 >= C4 * 1.4 * R1 / R3
C3 value demands an electrolytic. C4 should be a plastic film type. Notice the symbols are different.

A single chipamp should be your first project. Build and test in stages. Get it working with an 8ohm speaker. Learn about how and why it works BEFORE attempting a parallel chipamp.
Go and read the current threads on parallel chipamp woes as a warning.

BTW,
you can get 60W into 8r0 by choosing an appropriate transformer and smoothing cap values.
You can also get 60W into 4r0 or 6r0 by choosing the PSU

If you go parallel later a dual parallel can generate 120W (60+60W) into 4r0.
But for driving a reactive speaker I suggest a triple parallel for a 4ohm load. That will get you a fairly good 120W into 4ohm amp, but it takes some experience and skill to get it working reliably.

[coolbeer]

Originally Posted by AndrewT

c4 & r1 are a high pass filter that sets the bass roll off frequency and blocks DC entering the power amp.
Many suggest F-3dB between 7Hz and 20Hz, but I think it sounds better when set to between 1Hz and 2Hz.

A single chipamp should be your first project. Build and test in stages. Get it working with an 8ohm speaker. Learn about how and why it works BEFORE attempting a parallel chipamp.
Go and read the current threads on parallel chipamp woes as a warning.

Thanks for the input on the high pass bit of the cap/resistor, really appreciate the help

Yes, I plan to do one first and then extend it, but I really want them modular, so it's just to hook them together(well, perhaps not "just", but in an ideal world blahblahblah...).
My plan was to build one(as in channel) and use it to drive one of my rear speakers(witch is the only 8Ohms speakers I got). Then when I got that good and working build another one for the other channel.
Next step would probably be to ride those in parallel(probably center this time, as it's 6ohm nominal, 5ohm minimum), as I hopefully can get the boards simple enough.
I do realize that paralleling these are going to be a bit more work than a walk in the park, but I'm farely certain it can be done, and as I understand it, getting the chips to work as similar as possible is going to be my ultimate challenge.
You are right though, babysteps at first, then kill the dinosaur.

I did receive the Sony(yay!), I have measured the transformer, didn't make a slick of sense to me, so I did some extensive googeling and came up with the service manual for this beast. Now, the service manual does not hold any VA ratings, voltages or anything like that, but it does show how the transformer is wired, and with that knowledge I can measure it.

We have one primary, 220-240v I guess.
Then we have three secondaries, each with a centertap(yay! I guess), the voltages(AC) are 30-0-30, 20-0-20 and 2.5-0-2.5. My guess is that the 30-0-30 is the "main" winding with the most "oomph" as it has the big rectifier(others does only have single diodes) and the fuses on that rail is 2x6.3A/250v...
Now, 30-0-30 is a tad high... That means we are running 42.42v after rectifying it, as I did measure 80.4 after the bridge rectifier(some loss in the diodebridge I guess).
The question does become, what to do with that, do I wind the voltage down with a few lm338's or do some other unearthly magic, what would be best for designing a power supply that would drive the project from one lm3875/3886 to 4 of them in parallel.

I do apologize for the rather lengthy post, and appreciate any input on how one should go about a project like this. I'll also be reading everything I can on the forum about chipamps, as to broaden my understanding of how stuff works.

-
Kolbjørn

[AndrewT]

30-0-30Vac is too high for a chipamp from National.
Keep in mind that the transformer secondary voltage changes with load and with mains supply voltage.
At low load and high mains supply you may well be over the 84Vdc (+-42Vdc) max that National specify for a cold chipamp.

You can regulate, but this complicates the project.
You could go discrete and solve all the current output problems by selecting output transistors to suit the speaker loads.
Alternatively look at lme49810, 49811 and 49830.
These are designed to work with a much higher range of supply voltages and a wider range of speaker loads and a much wider range of maximum output powers. No paralleling of chips is necessary.

[coolbeer]

Yes, 30-0-30 is just on the edge, and will probably run over it at low loads. I checked out the lme49810/49811/49830, and it's pretty straightforward, although way more complex than lm3875's.

I don't really think of a regulated supply as a huge challenge, as I was planning on one anyway.
I was thinking of dropping the voltage down to +-35, so 7v will be over each lm338. This will not be particularly hard on the 338, although a proper heat sink is in order, atleast not with a single 3875.

With +-35v we are looking at about 60W Output power in 8Ohms according to the datasheet, although I suspect that 56W continuous is the limit, but we can peak at 60 if we need to.

At 60W we can deliver about +-30V(5v dropout max in the chip) which would require about 2amps on each lm338. (This part I'm not exactly 100% sure on, been a while(~10years since I had this in school)).

Another lengthy post, I tried to space this one out a little bit better, so it would be easier to read.

Andrew - If you think paralleling the 3875s are really harder than the added complexity of the 49810 with output stages I will seriously consider it, I've seen some of your posts on the forum and you seem to know your way around a integrated circuit.

Also, this won't be a rush project, I do have the capability to create the pcbs myself, but parts need to be ordered and money needs to be made before this can be a reality, so I figure a few months before I have my first tune out of anything. Therefore I have plenty of time to redesign everything and using this board for information.

Oh, and I redesigned the schematic for a single lm3875 stage, now the input dc/hp filter will be on a separate board due to it being better in my mind to filter it once instead of on every board that goes in parallel eventually. And it cleans up the schematics a bit.



-
Kolbjørn

[coolbeer]

I've gutted the sony some more and theese are my thoughts.



The red areas are for the chipamps, when/if running 4 we stack two and two.
The green areas are for the powersupply, the heat sink for the lm338's would be screwed or glued on.

The sony was really easy to dismantle, everything is made up in sections, with atleast one side of any wires able to disconnect, making it a non-solder job to take it appart.

[Redshift187]

Did the amp have a switch on the back for 8 ohm / 4 ohm? The 20-0-20 secondary might be for 4 ohm speakers, so you could just use that for your LM3886 project.