10W into four or eight ohms? An option to consider at the lower end of that range is using an op amp and STD03s; same basic idea as the LME49811+STD03 builds on this forum, just with lower rails and reduced output swing. More complicated than a LM1875/LM3886/LM4780, but also higher performance with a good op amp (LME49710 or LME49990 are good starting points). And still much simpler than a discrete amp.
I suggest LME49811 with Class A output stage. Our ears are most sensitive to these freq. With Class A, you don't have high-order crossover and switching distortions. Low power Class A is not difficult/expensive to build.
Why not class A with an op amp? Or AB? The 49811 measurements I've seen hold 0.01% THD in the transition from A to B, so I'd expect the subjective differences between class A and AB 49811 designs to be small to negligible. Assuming appropriate supplies, anyway. The 49710 and 49990 have higher Avol and GBP than the 49811 and are unity gain stable and lower offset. So they offer a better system gain budget, greater excess loop gain (particularly in the highs), and reduced need for a DC servo or blocking caps.
It occurs to me an op amp plus STD03s (or STD01s) without a servo is arguably no more complex than a servoed LM3886 or similar.
For class D I'd look to the ADAU1592 first.
Interesting combination for low power applications.
A test of 49810 with Class A output stage shows THD < 0.005 % for 10 W into 8 Ohm. The harmonic structure is more important than the pure THD number.
Attachments
What is the noise level of the LME49810/Class A combination?
I'm looking for quiet amplifiers for a high efficiency, multiway active loudspeaker project. Class D and LM4780 chipamps will typically generate 30-40 uV of noise when idling, and that is enough to be audible on a high efficency loudspeaker. Perhaps LME49810/ThermalTrak biased for class A up to 1 W, say, does better?
I'm looking for quiet amplifiers for a high efficiency, multiway active loudspeaker project. Class D and LM4780 chipamps will typically generate 30-40 uV of noise when idling, and that is enough to be audible on a high efficency loudspeaker. Perhaps LME49810/ThermalTrak biased for class A up to 1 W, say, does better?
Agree on harmonic structure. Though at O(0.001%) THD I'm unsure how much details matter since everything is really small.
10W is pretty loud, particularly for triamped nearfield listening to computer speakers---assuming 90dB efficient drivers that'd be roughly 109dB at the listening position. 60dB is probably more typical; that's roughly the SPL of an animated conversation. Power wise it's in the vicinity of 500uW per side, with around 225uW into the mid and bass and 50uW into the tweeter. The numbers'll obviously shift around some based on the speaker design and driver choice, but the power range of primary interest is likely the 10uW to 10mW space. That's usually comfortably within the class A range of a class AB.
Does pure class A offer benefit at such powers? I've seen 3886 and 49811 class AB measurements which go that low, but were done on equipment with a THD floor of 0.003% and hence could only report the 49811+STD03 was below the noise floor. In comparison, the 3886 was around 0.6%. Not aware of any data for a 49810.
10W is pretty loud, particularly for triamped nearfield listening to computer speakers---assuming 90dB efficient drivers that'd be roughly 109dB at the listening position. 60dB is probably more typical; that's roughly the SPL of an animated conversation. Power wise it's in the vicinity of 500uW per side, with around 225uW into the mid and bass and 50uW into the tweeter. The numbers'll obviously shift around some based on the speaker design and driver choice, but the power range of primary interest is likely the 10uW to 10mW space. That's usually comfortably within the class A range of a class AB.
Does pure class A offer benefit at such powers? I've seen 3886 and 49811 class AB measurements which go that low, but were done on equipment with a THD floor of 0.003% and hence could only report the 49811+STD03 was below the noise floor. In comparison, the 3886 was around 0.6%. Not aware of any data for a 49810.
You'll find an output noise spec in the 49810 datasheet if you look. Of themselves, the 49810/49811 are about 8dB noisier than the 3886/4780 whereas the 49710/49990 are 30 or 40dB quieter when operated at unity gain. Actually hitting the op amps' noise floor will take some careful board layout.
I'm unsure of the output devices' effects, though I'd expect the usual sum of squares.
Last edited:
LME49811 solutions seems like overkill for a 10W PC system. Especially since the spec sheets recommends +-20 or more for supply voltages. Simplest solutions are lm1875 or tda2050, but both work best with split supplies. There is a thread in here http://www.diyaudio.com/forums/chip-amps/158644-12v-dc-gainclone.html for 12 volt solutions, mostly bridge amps designed for the car market. The most notable is tda152Q class h (most power without separate supply). The questions raised seem to be: How good? How simple?
output noise from a 49811 amp
The output noise from a 49811 amp (Compact 49811 with ThermalTrak) is measured. Note that the amp is not placed in an enclosure. The input of the amp is shorted to its input ground. The measured noise level is 32 uV (BW=22 k) or 40 uV (BW=30 k). The FFT is shown below. The gain set resistors are 5.6 k and 240 Ohm. Their thermal noise affect output noise level. Low resistance should be use.
The noise level shown in 49811 data sheet was measured from the Fig 1 circuit. The gain set resistors are not optimized for low noise.
The output noise from a 49811 amp (Compact 49811 with ThermalTrak) is measured. Note that the amp is not placed in an enclosure. The input of the amp is shorted to its input ground. The measured noise level is 32 uV (BW=22 k) or 40 uV (BW=30 k). The FFT is shown below. The gain set resistors are 5.6 k and 240 Ohm. Their thermal noise affect output noise level. Low resistance should be use.
The noise level shown in 49811 data sheet was measured from the Fig 1 circuit. The gain set resistors are not optimized for low noise.
Attachments
Why do you say it's thermal noise and not voltage or current noise from the 49811? The 1.8k input impedance National uses is a pretty reasonable tradeoff between resistor size and preamp drive current requirements, but in my experience that means the dominant noise term's current noise through the 56k. With a 5.6k gain set I'd guess the dominant term is probably still current noise from the 49811, but a 240 ohm input impedance is not a terribly practical configuration due to the input current requirements.
Not to sound like an op amp fanboi, but this is another reason why combining a unity gain stable op amp with output transistors looks attractive for low power amps. Current and voltage noise from the chip is lower and, since all gain set resistors are the same size, it's possible to choose resistors in the 1-2k range where the voltage and current noise terms are comparable and the drive current requirements on the preamp are reasonable.
Unfortunately, the LME498xx data sheets do not provide parameters commonly found in opamp. Without those data, how can we minimize noise? One approach is to use as low resistance in critical parts as possible.
You have been suggesting using unity gain opamp with transistor buffer stage. Where do we get the voltage gain?
Hi,
Well, 10W can be handled using +/-15V rails. The dissipation to provide this as Class A (single ended) into 6 Ohm is also not that great, as we only need 2A peak current.
So I might use whatever op-Amp's sound I like with a simple Darlington Emitter follower on a big heatsink with a 2A CCS and regulated +/-15V Rails, total dissipation only 60W, so not too bad...
Or just use a bunch of Op-Amp's that can provide highish current in parallel, the LM6172 can provide 100mA peak current per channel, so 10pcs in parallel can do 2A peaks at 12V peak voltage, giving around 12W into 6 Ohm.
Using 40pcs bridge/parallel can even give as much as 50W/6Ohm...
Ciao T
Well, 10W can be handled using +/-15V rails. The dissipation to provide this as Class A (single ended) into 6 Ohm is also not that great, as we only need 2A peak current.
So I might use whatever op-Amp's sound I like with a simple Darlington Emitter follower on a big heatsink with a 2A CCS and regulated +/-15V Rails, total dissipation only 60W, so not too bad...
Or just use a bunch of Op-Amp's that can provide highish current in parallel, the LM6172 can provide 100mA peak current per channel, so 10pcs in parallel can do 2A peaks at 12V peak voltage, giving around 12W into 6 Ohm.
Using 40pcs bridge/parallel can even give as much as 50W/6Ohm...
Ciao T
We're generally in agreement on resistor sizing. But one can also model things. And when I've done that thermal noise is usually one of the smaller terms in the noise figure.
As Thorsten just reiterated, for a 10W amplifier per the OP's request no voltage gain's needed; parts like the LME49710 and 49990 maintain good linearity to 10V. Reduce that swing by the STD03s' Vbe and you get about 10W RMS into 4 ohms (if you look at my first post in this thread you'll see I ask the OP what impedance they wanted 10W into for precisely this reason). Similarly, an LME49724 based H bridge---or fully differential amplifier, if you prefer to think of it that way---goes to 40W into 4 ohms, also at unity gain from a differential input. For power levels above that the 49811 is the best part I know of. But, as I've observed already, 50+W is not required to make a pretty good racket for home audio purposes.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.