Help! Parallel 2x 3875 per channel design - please advise.

[loong]

I have most of stuff on hand, and want to have a small intergearted amp (for my 4x5m room). Being a newbie with very little electronic knowledge, it' better to seek for comments from experts here.

Note:
- LM3875TF x 2 per channel ( parallel ), no servo
- Power +/- 35V dc 250 VA per channel
- Cap 2x 1,000UF /per chip
- All R 1/4W 1% except Rout 0.1Ohm 3W 5%

1). Will this design work, particularly for 88dB speakers?
2). Is the input impedence is too low (output from CD go directly to the Amp via 50K pot).

3) More ideas for improvement are always welcome.

Many thankssss in advance!

[leadbelly]

No.

Read this link to understand theory/issues:

Bridgeclone

[macboy]

1) Yes it will. You should be able to get around 110 W into 4 ohms according to the Overture Design Guide spreadsheet. As always, YMMV.
2) The input impedance should be OK.

The only suggestion that I would make is to use a zobel network per chip, before the 0.1 ohm output resistor, rather than just one zobel after the resistors.

When constructing, do not connect the outputs (do not install the 0.1 resistors) until you have done some basic but important testing. You are basically connecting the outputs of two amps together with nothing but 0.2 ohms between them. If something goes wrong, it will go wrong in a most spectacular way.

Each amp should function perfectly on it's own, with no oscillations, no excessive heat, and no other signs of trouble. You should measure the offset voltage of each amp, and ensure that it is not too different. Remember that the difference in voltage here will create a current flowing from one amp to the other through just 0.2 ohms of resisitance! Also measure the AC output of each amp with a steady test tone. It should be very close to the same. If it is too different, then the gain of one amp is higher than the other. This also creates a current from one amp to the other which will cause extra wasted heat and unnecessary drain on your power supply. Find matched pairs of gain setting resistors using an ohmeter to ensure that the gains are equal.

[leadbelly]

I disagree with macboy. At the very least, you need 0.1% tolerance resistors setting the gain to make sure each chip is sharing the load. This is discussed in National's App Note AN-1192 and in the thread I mentioned.

[GregGC]

Originally posted by loong
I have most of stuff on hand, and want to have a small intergearted amp (for my 4x5m room). Being a newbie with very little electronic knowledge, it' better to seek for comments from experts here.

Note:
- LM3875TF x 2 per channel ( parallel ), no servo
- Power +/- 35V dc 250 VA per channel
- Cap 2x 1,000UF /per chip
- All R 1/4W 1% except Rout 0.1Ohm 3W 5%

1). Will this design work, particularly for 88dB speakers?
2). Is the input impedence is too low (output from CD go directly to the Amp via 50K pot).

3) More ideas for improvement are always welcome.

Many thankssss in advance!

I'd say if this is your first project why use 2xLM3875? For 88dBm speakers and the room you specified single IC LM3875 should be just fine for 6-8 Ohm speakers.

I'd keep it simple as a first attempt. You may find you don't need more any way.

/Greg

[macboy]

I disagree with macboy. At the very least, you need 0.1% tolerance resistors setting the gain to make sure each chip is sharing the load.

Did you not read my comment or do you just enjoy disagreeing with others? I said...

Find matched pairs of gain setting resistors using an ohmeter to ensure that the gains are equal.

We do not need to use high precision resistors because we are not mass-manufacturing this thing. Hand matching will suffice quite well enough, and in fact can give better than 0.1% results.

[macboy]

For 88dBm speakers and the room you specified single IC LM3875 should be just fine for 6-8 Ohm speakers.

Judging by the load in the schematic loong posted, I would assume that the speakers are 4 ohm. In that case, two parallel 3875's will perform much better than one.

[leadbelly]

Originally posted by macboy
We do not need to use high precision resistors because we are not mass-manufacturing this thing. Hand matching will suffice quite well enough, and in fact can give better than 0.1% results.

Well then you had better describe the procedure to loong who described himself as a newbie on how to get better than 0.1% match with what probably is his 3-1/2 digit multimeter...

[bigparsnip]

One thing that I would sujest which may be of some use here, is to use seperate input caps for each chip (rather than the single, common one currently being used). This will allow slightly diferent input bias curents to flow into each chip (which won't be identical) without creating a potential difference between the two. But, you would still need to use very close tolerance resistors to set teh feedback for the circuit.

[macboy]

Well then you had better describe the procedure to loong who described himself as a newbie on how to get better than 0.1% match with what probably is his 3-1/2 digit multimeter

Point taken. I made the assumption that he had access to a decent multimeter, but heh, you made the assumption that he had 0.1% resistors in his parts drawer... or at the local supply house. I think that more people have access to a 4.5 digit or better DMM than to .1% resistors. It doesn't matter though. We agree that the gain setting resistors must be very close in value so that the gains are equal and don't cause problems. That's really the point. Whether they are expensive high precision devices or just hand-matches cheapies is not.