Any warnings about paralleling chips?

[Morse]

Hi guys and thanks in advance for any info.

I've checked the archives, and when you input "parallel", you get a LOT (over 3000 hits), which is rather a bit to search through. Thus the question:

If I parallel a pair of LM1875's or sections of an LM4766, will I encounter any problems? Are there extra decoupling or zoebel type circuits that I need to look into?

I'm tentatively planning on building an HT amp using either LM1875's or LM4766's and I'd like to have low impedence capability on the sub - preferably with parts already on hand. The other option would be to bridge a pair of TDA2030's (which have good low z handling but aren't so hot at total power out).

Suggestions?

Thanks again and all the best,
Morse

 

[Hybrid fourdoor]

Well I had this question in a previous post, maybe it can be answered here.


If you put two chips in parrellel, they see (according to Nationals Application Notes Pg #2 1st Para.) [# of chips*load]. So if I had 2 LM3875's and a 8ohm speaker each would see 16ohms?

Now my real question is, even though seperatly there are putting out less power, together they are putting out the same.

So 2 LM3875's on a 8 ohm speaker (16 as seen by each chip) will put out half the power of 1 chip on a 8 ohm speaker, but because theres two, it adds back up. So basically you get the same output, but less strain on the chips...or rather greater thermal capacity....is this right?

 

[peranders]

If you check the AN-1192 and follow the recommendations you will succeed.

David: Yes. Think that each chip has the some output voltage and each chip has a maximum output current. This means that you can take out 2 x current => possible to half the load (= increase the load). Remember also that you will get higher output impedance if you connect in parallel. I say, connect in parallel if you really must, otherwise avoid.

 

[Morse]

Thanks Peranders!

That's just the sort of information I was looking for!

All the best!
Morse

 

[digi01]

peranders

you will get higher output impedance if you connect in parallel

I think in parallel,the impedance of output of the amplifier will be reduced.

Perhaps I am wrong ?

 

[banana]

I think in parallel,the impedance of output of the amplifier will be reduced.

I suppose output impedance would be higher, if the ballast resistors were taken into account.

Say 2 chip parallel, with 0.2 ohm on each chip, then the output impedance would not be smaller than 0.2/2=0.1 ohm.

My 0.2 cent.

 

[tlmadsen]

way to go

Hi Morse

I am taking the liberty to repeat the anser I gave to leadbelly about a week or two ago:

It is not as difficult as it may seems to get the parallingen going. There are 3 major issues to look out for:

A. Make sure the to chips don't sink/source each others DC-offset. This can be done by eacher lower the DC-offset as much as possible and/or making the output resistor bigger. Both solutions has its ups and downs.

B. Make sure the gain is the same for both chips so they share the load equal. Using 0.1% is fine.

C. Make sure that the output resistor is as eaqual as possible also to make sure that they share the load.


If you use normal 1% resistor for B. & C. it will moste likely work for 98 % of the time. (1% resistors comming form the same batch are 0.3-0.5 %)

However you MUST deal with the DC-offset in a propper manner.

Remember that if you have an DC-offset at one amp of +12 mV and the ohter amp at say -14 mV and your output resistors are 0.1 R you have 130mA going from one amp to the other (26mV/0.2ohm)

Adding more chips in parallel just make it even more "fun" regarding the current flowing.

Yes, the output impedance become sligtly higher due to the load-resistors, but if you keep them under about 0.5 ohm I think you will be fine (the so called "dampingfactor" is in my wiev highly " overestimated". People seems to forget that you offent have a complex passive x-over between the amplifier and the speaker units that totally "ruin" you dampingfactor......).

Thanks to the loadresistor you don't need a R/L network at the output. However you still need a (one) Zobel, just like in an non-parallel amplifier.

Have fun

THomas

(P.S. you also get slightly lower noise by parallingen, not that I think you well ever notice)

 

[Morse]

Thanks Thomas;

Those are all very useful tips, and I'll follow them! Since I'm from a valve background, I'm pretty aware of the increased liklihood of oscillation with each added device (at least with that technology), so I've been a little fearful of trying it with chips (first amp I ever built that went into oscillation was a chipamp - after completing a number of stable valve amps!).

>>>...B. Make sure the gain is the same for both chips so they share the load equal. Using 0.1% is fine...<<<

Are there standard .1% resistors available now? I thought I was doing pretty well when I bought a few hundred 1% resistors, and several thousand assorted 2% and 5% to go along with all my old 10% and 20% rated carbon comps....

Okay, now I see why the output resistance goes up with paralleled chips - you're using a bigger output resistor (not a problem, I've used 1 ohm output resistors without problems). FWIW, I'm not a big believer in super high damping factors either - after all, my best sounding amp is a 2A3 SET, so you can imagine what sorts of damping factors I'm used to.....

>>>...Thanks to the loadresistor you don't need a R/L network at the output. However you still need a (one) Zobel, just like in an non-parallel amplifier...<<<

Interesting; will I need to make any alteration in the output zoebel's value, due to the differences in output resistor and paralleled devices?

Thanks again and all the best!
Morse

 

[tlmadsen]

Yes you can get "standard" 0.1 % resistors, but they are still expensive. Handpick form 1% batch is the best way to go.

Just make your Zobel 100nF+10 ohm and you will be fine

Have fun

THomas

 

[lawbadman]

Someone correct me if I am wrong.
But if I understand what you guys are saying, that means that paralleling two chips gives no increase in power output.
So why bother parallel if the only advantage is sharing the load between the two chips?

 

[leadbelly]

Someone correct me if I am wrong.
But if I understand what you guys are saying, that means that paralleling two chips gives no increase in power output.
So why bother parallel if the only advantage is sharing the load between the two chips?

No increase in power driving an 8 ohm speaker if everything else remains same...but not all speakers are 8 ohms

And if you change other things, like rail voltage, then you will get more power into 8 ohms.

 

[peranders]

Someone correct me if I am wrong.
But if I understand what you guys are saying, that means that paralleling two chips gives no increase in power output.
So why bother parallel if the only advantage is sharing the load between the two chips?

Yes, why?

1 If you have +- 42 volts and have a 4 ohms speaker? Can a single LM3886 handle this?

2 If you have less than +- 42 volts and have a 2 ohms speaker load. Can a single LM3386 handle this?

 

[lawbadman]

peranders,
Just yesterday I made a chip amp paralleling 2 LM3886. I was wondering if just using one chip will give me the same power as two.
My speakers (MMGs) are 4 ohms and my power supply is +-40 Vdc.

In my case two chips should give me more power right?

 

[peranders]

You will have more power only if:

1 The cooling of one LM3886 isn't efficient enough

2 The current is too much for one LM3886

When we talk about power, don't forget that it's about dB's. Loosing 30% power means only -3 dB!