Two interesting documents. Current Drive Amplifiers and How Amplifiers Treat Big Signals

Moderator
Joined 2007
Paid Member
I came across these two papers from John Woodgate and surprisingly can not find any reference to them on the forum... so here they are. The current drive paper will be of interest to many I'm sure.
 

Attachments

Interesting…. Figure 5 on the second paper shows what happens clipping “with a good low impedance power supply”. Unfortunately, it doesn’t show what happens with a crappy one and capacitors way too small for direct comparison. It should.

I’ll give you all a clue - it looks like a ski slope across what’s supposed to be a “flat” top, full of upper order harmonics (and harmonics/IM products with the power supply ripple). - and sounds buzzy and staticky. If you get the waveforms from Figure 5 (with typical music) you don’t even hear the amplifier clipping until it’s far worse.
 
  • Like
Reactions: kgrlee
LM386 Just Isn’t Big Enough. Designed to be able to be “heard” off 4 AA batteries, nothing more nothing less.

Most any differential input feedback amp can be made to work in current mode. Many early solid state guitar amps just used a basic Lin circuit with this same feedback modification, and it basically works. Chip amps in general dont work as well because of their 3 to 4 amp maximum current. And nobody’s is any better than anybody else’s - it’s an IC process limitation. Discretes will usually give you more, but you can’t make ICs on those processes. The TDA2003 appears to hit above its weight class because it can put out that same 4 amps off a very low voltage supply, since it was designed for car radios. It’s also a singe ended CFA, not a regular VFA like all the rest. The - input is very low impedance, which is ideal for a transconductance amp. Hence the “abnormally low” feedback resistor values in normal voltage mode. In current mode they are always low. Chip amps are “easy” to make this feedback modification with, but there’s no reason on Earth the same thing couldn’t be done with a Honeybadger. The only question is why would anyone want to?

LM1875 is the most like a standard op amp of any of them. It just puts out 3 amps instead of 30 milliamps. TDA2030/2050 was designed to give a bit more kick as an audio-specific amp. They are the SAME part, just different voltage grades. The die on the 2050 is a bit bigger, the equivalent circuit is the same. That’s why a fake 2050 is usually just a re-badged 2030. You won’t know the difference till you crank it up.
 
Unfortunately, it doesn’t show what happens with a crappy one and capacitors way too small for direct comparison. It should.

- it looks like a ski slope across what’s supposed to be a “flat” top, full of upper order harmonics (and harmonics/IM products with the power supply ripple). - and sounds buzzy and staticky. If you get the waveforms from Figure 5 (with typical music) you don’t even hear the amplifier clipping until it’s far worse.
Yes. This is a little known factor but very important factor in 'amplifier sound'. Despite Mooly's excellent survey how-much-power-do-you-really-need-for-domestic-listening, I tend to clip a 50W/channel amp & 90dB/W@1m speakers fairly often (I've got a lot of my own uncompressed live recordings).

When an amp clips, the signal is modulated by the sawtooth on the power rails .. a really nasty noise. The worse affected are Class A amps cos they are usually smaller than usual and the large idle current generates giarnomous sawtooths on the power rails.

Something like a JLH 10W Class A, probably my favourite Class A amp, benefits ALOT from regulated SMPS I've confirmed this in DBLTs. With regulated supplies, you can go quite a lot into clipping with music before you notice and even further before you object.

Of course there is modern 'music' (??) where if you are clipping 50% of the time, you won't notice .. let alone object 😱

Today, you can use several laptop PSUs in parallel instead of a truckload of electrolytics hand carved from solid Unobtainium or BS 🙂
 
Last edited:
The only problem with SMPS’s is that ones that can deliver +/-80 volts at 20-30 amps peak current repetitively, safely, an with a long life span still cost THOUSANDS of dollars. Nothing will change that. Anything priced in the low hundreds either will not stand that test of time or go into a hiccup mode when trying to deal with low bass. Far worse than the sawtooth.The amps can be built not to conduct 120 Hz ripple at clipping, but it needs to be under a couple volts to make that happen. Wild 10-20V supply swings eat up too much and the amps just have to pass that on to the speaker in the interest of cost. Which is what happens and why this is a problem. What it takes is a conventional supply with a low impedance - priced well between that of a cheap Aliexpress China supply and a 4800 watt solar inverter. Roughly half way on the log scale. $200 for the toroid, and another $100 for the caps. The rectifier is almost free in the grand scheme. It is what it takes to get supply impedance low, but a bridge too far for penny pinching manufacturers. So the amp “buzzes” when it clips. Youre not supposed to clip it anyway, right?

For the 50 ish watt class amplifier you can not beat a SMPS. Power demand is small enough that a reasonable cost one is reliable, will handle full peak current NOT just the long term average, and regulated don’t cost much more. They can be DIYed. And it will clip clean. It’s just when you scale it up, it hockey sticks in price and the old toroid again becomes attractive.
 
I know that it was not difficult to clip a classic AB amp at 150 watts while listening to a 96db efficiency Fostex FE208 sigma (year 2000) in a Jericho Horn playing a record of a drum kit close to original loudness.

Since then I use at least amps like Behringer Inuke Nu3000dsp which can do the double in watts without clipping
 
The power supplies in those Inukes are the ones your mother warned you about. They won’t just go into a limit mode on bass and cut back, misbehaving horribly. But run them too far over their design average for any length of time and the IGBTs in them (and the mosfets in the class D amp too) will literally wear out from internal thermal cycling. In a home environment you may never see it. But in the field you’ll be throwing it away in 3-5 years. It’s like needing to upgrade your iPhone “anyway”.
 
Can you expand on what you mean?
Pre-regulate and/or clean up the supply voltage to the entire front end. Negligible ripple. Then ensure VAS clipping does not quite saturate the output transistors, leaving only a few volts on them. The problem is the max peak output voltage you can now get cleanly is the BOTTOM of the ripple trough. Great if you’re not losing a lot of voltage to that ripple and can LIVE with the lower power rating. When the supply is losing 10 or more volts a combination of that ripple and load related sagging the manufacturer is not willing to give that output swing up. So we let the transistors conduct all that crap and yes you hear it.
 
LM1875 is the most like a standard op amp of any of them. It just puts out 3 amps instead of 30 milliamps. TDA2030/2050 was designed to give a bit more kick as an audio-specific amp. They are the SAME part, just different voltage grades. The die on the 2050 is a bit bigger, the equivalent circuit is the same. That’s why a fake 2050 is usually just a re-badged 2030. You won’t know the difference till you crank it up.
Would paralleling help getting more current out of them?

I would like to see a schematic for cfa for lm1875 with two or three of them in this mode
 
If you need a bigger amplifier you need a bigger amplifier. Paralleling chip amps is like a band aid being used on a 6” incision.

As I mentioned earlier, you dont need a single chip amplifier to make the concept work. Just take your basic diff pair/VAS/EF2 amplifier and apply the same type of feedback arrangement. As long as the amp is stable with the effective voltage gain you end up with. You can scale it up to several hundred watts per channel if you want to. Or make 50-100 very easy watts with one or two pairs of output transistors.
 
  • Like
Reactions: billshurv
Interesting. In the first paper Woodgate mentions the TDA2003 as usable for making a current drive amplifier.

LM1875 in his eyes has no advantages. Others like TDA2030 or LM386 are inferior for use with current drive mode.

TDA2040 is a 4A/25W version of the 3.5A/20W TDA2030, should be adequate, otherwise there s the yet higher current upgrade TDA2050 wich is good for 5A.
 
I know that it was not difficult to clip a classic AB amp at 150 watts while listening to a 96db efficiency Fostex FE208 sigma (year 2000) in a Jericho Horn playing a record of a drum kit close to original loudness.
How do you know you were not overloading that Fostex thingie? Is "knowing" the same as "assuming" here, or did you actually measure the amp clipping?
 
@wg_ski - A TIP41/42 pair is still only TO220 package so has limited power handling even with an 'ideal' heatsink of c 20W; and be careful SOA doesn't bite you in the hindmost. I can't find a graph for that in the datasheets I have..

But yeah: I find them useful too, for many medium -power uses; great pretty robust 'jelly bean parts' to have to hand : )
 
SOA isn’t any better on those IC power stages. They put TWO of them in a single TO-220 chip amp - putting the chip amp at an even further disadvantage compared to separate TIP41/2.

Oh yeah… the fT is higher on the IC process, I’ll give them that. But only on the NPN side. It’s even worse for the PNP. The only reason crossover distortion is ANY better is because of the inherent thermal tracking. That same thermal tracking provides unwanted thermal FEEDBACK to the input stages, giving rise to the uptick in harmonic distortion at low frequency.

Chip Amps are fine for 10 or 20 watts. Above that they just struggle. Why everyone still wants to use them at 40,50,60 or more watts is just beyond me. Just a recipe for trouble.
 
5 amps is NOTHING. A TIP41/42 pair can do better than that.
We re talking of continous output current, TIP41/42 are 6A continous, so not that much more even if two TO220 cases instead of one is an advantage for thermal dissipation.

TIP33/34 ar a better fit, besides there s numerous designs that have been described using a TDA2030 driving a pair of such transistors, shoud be the best solution overall at least for limited voltages.

That being said TDA7294 is what should have been used if he wanted to keep on totaly integrated devices, there was once a power op amp that would have been perfect, namely the LM12C/CL but it s no more manufactured by NS.