Welcome to Solid State Guitar Amp Forum | DIY Guitar Amplifiers. Please login or sign up.

October 03, 2023, 07:16:08 PM

Login with username, password and session length

Recent Posts


Thinking about building a SS amp

Started by edvard, June 09, 2018, 04:27:10 PM

Previous topic - Next topic

J M Fahey


    Paralleling *voltage*  sources is a disaster waiting to happen.
    The additional condition that now gain must track each other within 1% or less, any volume, any frequency, plus NEEDING cheesy series resistors to somehow help speaks volumes about what a poor Engineering choice it is.
Well... maybe. Pretty much no really high power amplifiers in existence get by without emitter resistors. The emitter resistors are there to wash out the microscopic differences in Vbe.
Please don´t compare pears to oranges.

1) Transistors in parallel are emitter followers, either real or quasi complementary, with a gain=1 so a small value resistor is *ample*  to equalize current.

And if you have a Vbe difference, say 50mV between two of them (which is GROSS), at full power driving some load, if one puts out 5A through 0.33 ohms so dropping 1.65V, the other, due to the 50mV difference (say it´s lower, but analysis is the same if it´s higher) it will apply 1.6V across its own 0.33V , so current now will be 4.86A , a 3% current difference, which is NOTHING.

* And in any case current from both is "useful" since it flows into the load.

* And emitter resistors "disappear" because they are inside the NFB loop.

* And since there are not "2 brains trying to control the same muscle and opposing each other", even 5% or 10% resistors everywhere are perfectly acceptable.

2) Now to parallel voltage source, high NFB amps.
To begin with, *each*  will try to keep output voltage tightly controlled, and obey its own offset and NFB; will NOT cooperate with the other amp but fight it, feeding as much current as needed into the output node as necessary to achieve that.
Fine if driving a passive load , such as a speaker or resistor, a disaster if feeding a node which is also fed by another voltage source, worse if the other one has same NFB and same muscle because current will flow from one into the other trying to keep shared output node at what HE thinks it should be, and even mV difference can mean AMPERES of current .

Which are WASTED Amperes flowing from one to the other, instead of into/through a load.

That´s why I accept *bridged* amps (load is connected *separating*  them and current has only one path: through it) buit hate parallel ones like the kludgy mess they are.

TYrimming NFB within 1% (or else) and adding cheesy 0.1 ohm resistirs *outside*  NFB only confirm that is a poor design.

Why do they exist then?
2 main reasons:

a) not too qualified "weekend designers" find they can "design high power amplifiers::) without all the nasty Math, just pile up LEGO building blocks and pray.

b) you will need to buy MANY chipamps to build them, and also allow typically low to mid powered chipamps occupy a market sector they are not really qualified to= $$$$$$ for Chipamp makers  ;)

Now as I said before, if you use them *separate* , each one feeding its own speaker , fine with me.


I have nothing but respect for your technical expertise and experience.

However, I did a look at the numbers, and given:
1) well behaved chipamps, like the LM3886 (perhaps excluding others; I did not do the math there)
2) good heatsinking
3) precision feedback resistors and biasing resistors
paralleled LM3886's can be made to work reliably.

My professional and personal experience up to very recently would agree with your position. However, if TI went to trouble to write up an app note on bridged and paralleled amplifiers with the LM3886, and go so far as to publish their results. I know, I know, lies, d@mned lies, and app notes.

I have read several sendups on the web of people successfully getting the paralled version (and bridged+parallel version) to work.

I calculated the cross currents and the watts they'd add to the chips for the offsets and for the cases of worst case 1% and 0.1% resistors. The math says it's untenable with 1% resistors, but pretty well acceptable for the 0.1% case.

When I was starting, I was told to figure out how to design things with 10% resistors, saving 5% for closer tolerance work. I had to get higher engineering approval for 1%. A few years later they encouraged us to use 1% whenever needed. 0.1% resistors were not in any catalog I knew of (these were on paper back then). But today I can get literally thousands of values and types of 0.1% resistors from my local distributor overnight. So I've had to re-think what it too demanding to even try.

You are correct - this is not a situation for use by one-week-wonder geniuses. But It's not all that far off in the tall weeds, either.


Oh, and I forgot to add - I'm funny about just blankly trying this, too. That was one other thing that voted for "why worry - just do one amplifier per speaker".


Great reading and helpful info :dbtu:
I don't know if it helps but a while back I posted a blab on a Behringer monitor (Truth 2031) which had a blown tweeter. The unit runs 2x LM3886 in parallel driving a 4 Ohm woofer and a single LM3886 for tweeter.

I was surprised that the rails were +/-40 and seems to work without issue,, for the low freq at least. I have no idea why the hi freq blew but that chip was too hot to touch and measured rail voltage on the output which obviously burnt out the tweeter.

No real heat sink except for a 6mm flange which is also bolted to outer alloy case.
These were obvioulsly a pair and last I spoke to the owner the other unit is still working 3 years later, being used as a local monitor.

I'm not keen on building things that rely on super critical parts to run. Better to over build it. 8|

Adding schematic of the truth speaker and gut shot.
It just bolts onto the back of the speaker box.


Quote[...] The unit runs 2x LM3886 in parallel driving a 4 Ohm woofer and a single LM3886 for tweeter.
[...] but that chip was too hot to touch and measured rail voltage on the output which obviously burnt out the tweeter.
[...]No real heat sink except for a 6mm flange which is also bolted to outer alloy case.
You already know the answer then.
Read this:
It contains a lot of my advice to RDV about heatsinking LM3886s. If you simply attach a chip amp to a flange and hope the chassis will take care of it, you;'re almost certain to be disappointed.

Even if an adequate heatsink is used, the mounting is critical. It simply MUST be in good THERMAL contact, so the heat can get to the adequate heat sink. Loosened mounting bolts can undo whatever good a heat sink might do you.

In this case, it is entirely possible that poor operating conditions made the LM3886 oscillate. That's a suspicion based on (1) it was a Behringer, (2) it was attached to a tweeter (3) the heatsinking wsa inadequate. Oscillation would make it dissipate a huge amount of heat, and any inadequate heatsinking could let it fry before the thermal protection could cut in.

There is nothing inherently wrong with the parallel use of LM3886s if you do it right, use the right precision parts, and pay proper attention to thermal issues. But just like the use of fire, you can do a bunch of things wrong before you get smelting metals and internal combustion engines to work.

There is no substitute for knowing what you're doing. Like the song says, "you gotta know when to hold 'em, know when to fold 'em".


gain thanks *RG* for the explanations,, all very helpful. :tu:

Yes I agree the Behringer problem is most likely a design flaw,,, but then there must be hundreds of them still running fine.
So was it user abuse,, design flaw or a mixture of both.
The owner never gave me the impression that he abused his gear.

Anyway I advised the owner that to the best of my readings about LM3886 and the potential problems at such high voltage rails,, even if I replaced the parts it could just happen again,, so he put his money to something new. So for a while it sat in the junk pile.

1% tolerance rings alarm bells for my hobby brain with limited experience,, :lmao:
One wonders what happens when those 1% components go out of spec?
Meanwhile a bog simple discrete power stage can run emitter resistors 20% out of spec for years and no smoke.

Re the Behringer story;
As the low freq still worked I linked the biamp back up to full bandwidth and fitted it on the back of another speaker box with 8 Ohm driver and passive Xover for horn.
I figured that an 8 Ohm load would be less stress on the chips and I wanted to see how well it would cope.
It's been flogged a few times and gets hot but keeps going,, only time will tell. :-X

Re TDA7294 chips
A mate picked up a Skytek biamp rig with intermittent crackles (see pics, this one has 3 of those TDA7294 chips). This time 2 running in bridge mode for the sub, the other running a pair of small satellite drivers. I fixed the cracked solder pads on the supply and all seemed ok for about 10 min,,,, then we cranked up the volume and Boof.  :o :o 
Oh dear,, it melted pin7 on one of the bridge chips (positive supply rail). :'(

Sub was 4 Ohm and even with the massive heat sink it broke.
Claimed 300 Watts,, Ha what a joke. :lmao: :lmao: :lmao:
Even if the heat could get across to the heat sink those 15 pins are very thin and I doubt they could handle the current as they are only 1mm x .86mm,, while the 11 pin LM3886 are .38mm x .81mm  BUT it does not claim to handle 100Watts.
Anyway more spare parts for my junk draws.  :tu:


You can start with combining a booster(as the input stage), a distortion pedal, a speakersim(Visit www.tonepad.com for the PCB/schematic) and an amplifier module which is sold on Taobao or Amazon or other Internet shops