High Power Chip Amp

[MJL21193]

Originally Posted by AndrewT

...An extra X dollars in how many hundred thousand dollars? Are we talking about 1% extra or 5% extra?

The extra cable, the extra cable runs by labour, extra mains outlets to plug in all these amp, extra harnesses to hook everything up, more extra labour extra, extra.... It's never as simple as it seems.

[a.wayne]

Originally Posted by AndrewT

I put chipamps at the bottom of my ranking. They are designed for cheap mass production assembly to eventually power speakers in TVs and radios etc.
The Krell problem that I found was a stability problem, possibly brought about by my selection oc components. I awaited all the other builders testing their Klones, but to date, some two years later, very few have actually got any Klones running.
My Sugden ClassA is at the top of my ranking, all the ClassAB are between these extremes.

OOPS, I did buy a Cmark PA amp, but I daren't connect it to any of my speakers considering the quality of design and assembly inside that chassis. I suppose that is the lowest of my ranking. Doesn't say much for the chipamps though.

ps,
which analogy? Was there one?

Fix has his Krell up and running , all seem well there , Hmmm it even made 700 watts into 1 ohm ...

[danielwritesbac]

Originally Posted by AndrewT

I put chipamps at the bottom of my ranking. They are designed for cheap mass production assembly to eventually power speakers in TVs and radios etc.

Methinks he finally heard the screech of the amps that are special made for loud clear voiceband in tv use. Yes, the majority of chipamps are either severely underpowered and/or should be bolted to their own dedicated speaker because of other performance issues (such as sounding like a tv amp).

Instead of having a big chip amp run in parallel mode to turn its own screech into heat (LM3886's don't make that noise at exactly the same time, so that triple-parallel removes the noise on PA150), thus eventually resulting in clean audio. . .
Even though that works just fine. . .

I'd prefer to do it a different way.

Since the majority prefers the sounds from LM1875, why not bridge it so that it becomes a pre-drive for a reduced parts count solid state amp? Is it impossible to do a nice job with this? I've never seen it fully implemented. It seems that this would be a useful introduction (springboard) into projects that aren't fully sealed shut inside a chip.

P.S. ABS Max volts for long life LM1875 is 26.5+26.5vdc regardless of load. However, that's still a pretty healthy shove as a pre-drive. It can scale to a massive amp. Howabout it?

[Mark Allen]

I've had both of my Krells up and running for at least a couple of years now but hardly ever use them. I prototyped the first batch of PCBs for both of those amplifiers. No stability problems at this end on either of them and the Krells are well known for their excellent stability into any load. Not sure what you did especially considering your easy loads there.

Just because the chip amp chips are cheap doesn't automatically deny them from being the high quality performers they really are. You can't discount the reviews that Peter Daniels gets with his Patek 3886 based amplifiers!! Parallel as many as you like just as Nelson Pass parallels gobs of tiny fets... You can do anyhthing with any device if you put your mind behind it! The BPA 300 running here in stereo mode is certainly far superior to all the amps I have here and by a long shot. You can do all the math or what ever you want, spend alot or a little but the only thing that matters in the end is how it sounds.

Mark

[a.wayne]

True Dat !!! .... errr ahh, what he said .......................

[star882]

Originally Posted by Mark Allen

They would probably drive anything if you parallel enough of them. Say 8 or ten should easily drive a 1 ohm load. Each chip is capable of about 7 amps at operating temperature and a bit higher when cold. You'd also need to increase the power supply capacity as well. Andrew could work the math on this for us because he is fairly familiar with the 3886. Stop and think though that paralleling 8 or 10 or 12 of them would still be very cheap! They cost about 3 bucks each in quantity... I believe the three paralleled that I am using might actually do a 2 ohm load, for sure 4 ohms with ease... might need to go to four chips though for 2 ohms. I'm not the math expert here by any means...

Mark

The TI hybrid can do 2 ohms in single channel BTL and it only costs about $7 each in quantity. Then there's the very significant savings from the reduced power supply and heatsinking requirements. (Unfortunately, if you buy a kit, the hybrid premium cuts into the savings. But the lower operating costs mean you'll be saving anyways.)

Any more power than that and a discrete hybrid or pure digital makes the most sense. After all, the main motor inverter in an EV is basically a very big digital amplifier that outputs up to hundreds of kilowatts.

[Mark Allen]

Well, we're not burning rubber here.... just chip amps! I have no doubt an all digital system might be superior but I don't have the time to devote to even attempting that.

[star882]

If you're looking at high power, a hybrid or pure digital using discrete power electronics is the most cost effective. FDPF52N20T MOSFETs (200v, 52A) only cost $1.37 each in small quantities! FDPF52N20T Fairchild Semiconductor MOSFETs

[danielwritesbac]

Originally Posted by Mark Allen

Well, we're not burning rubber here.... just chip amps! I have no doubt an all digital system might be superior but I don't have the time to devote to even attempting that.

Not burning rubber? Oh, you haven't bridged it yet.

[danielwritesbac]

Originally Posted by Mark Allen

They would probably drive anything if you parallel enough of them. Say 8 or ten should easily drive a 1 ohm load. Each chip is capable of about 7 amps at operating temperature and a bit higher when cold. You'd also need to increase the power supply capacity as well. Andrew could work the math on this for us because he is fairly familiar with the 3886. Stop and think though that paralleling 8 or 10 or 12 of them would still be very cheap! They cost about 3 bucks each in quantity... I believe the three paralleled that I am using might actually do a 2 ohm load, for sure 4 ohms with ease... might need to go to four chips though for 2 ohms. I'm not the math expert here by any means...

Mark

I was going to try to argue with this and say that output transistors cost less for big amps. That idea failed somewhat.

However, parallel does make for cancellations, in this case used to quite good effect in making LM3886's fight off their own noises (that don't occur precisely at the same time). Unfortunately, if you were to cause enough cancellations then you also erase audio that is similar to the cancellations.

That is the point (beyond a carefully made pa150/bpa300) where output transistors start to get obviously more competitive on making clean audio.

The cancellations from paralleling can be reduced as well as the heat when the output devices are matched. Matched Lm3886's scale to pa200/bpa400. UNmatched output transistors scale to pa200/bpa400. Matched output transistors scale upwards of 1kw. Heatsink expense is decreased in proportion to cleaner audio from output transistors. And, the thermal interface area is considerably larger.

I'm not talking about a complex and dizzying time analyzing all of the possible options of a discrete preamp/predrive. What I'd like to see is a ready solution involving a chip-based bridged predrive running 8 output transistors per channel (bridge+parallel). Preferably, it can be "bugsplat" wired without a circuit board, in 20 minutes or less to result in a very powerful high fidelity amplifier. Design? Anyone?