Sound quality of chip amplifiers

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.

angel

Member
2003-10-09 3:42 am
Norway
"Can" being the operative word here.

Few chip manufacturers go to the lengths required to fully realize the potential of ICs. Some do, though, and their prices reflect it.

I have no doubt that chipamps can sound great; I've owned at least one that did a great job.

However, I have yet to audition one that (given a USD/EUR >1000 source and a pair of USD/EUR >2500 speakers) could perform as well as Balanced Audio, Pass Laboratories, etc..

Give John Curl the freedom to design a large 90nm process with a few runs in the measure/trim cycle, though, and I'm fairly certain those of us who could afford the chip (along with some nice Black Gate and/or Jensen caps, plus some neat attenuators) would throw our discretes out the window in no time.
 

tlf9999

Disabled Account
2005-05-25 2:25 am
none
Kees said:
Chip amplifiers are extremely simple, but what is the sound quality in comparison to discrete amplifiers? :cool:

They are just as good as similar discrete amps. To me, discrete amps have the advantage of being extremely beefy and being able to handle large current.

I would say chip amps are extremely simple: while the physical count is low, they have a lot more devices inside of them. For the same reason, you wouldn't call a Pentium 4 cpu extremely simple.

tmblack said:
Anyone compare the LM3875 with the OPA549?

Tom


I have done that (lm3875 vs. opa541). I think the opa541 has more punch at lower frequency but I don't think i can tell them apart consistently in a scientific test.
 
Chipamps are great.

They are designed to be cheap, but because it's an IC and they can build to exact specs, and have multiple CCS, they work good, and have very low offset, and can work great at almost any voltage in their working range.

Before I really figured out HOW a descrete amp worked, I used chipamps. LM386 is one of my little favorites. You can use em for anything, and they are so small. Also I love to play with opamps.

For larger power, the big chip amps are awesome. I'm using a HT Receiver, and the two front 100W Channels are driven by a Pioneer chip and 6800uf caps. The highs are great and sound clean, and it drives good power. The bass isn't too bad, but it doesn't use really large filter caps, but still sounds good.

I have a 2nd Amp for the Bass, and my 12" 4 ohm subwoofer. I'm running the 2 channels in parallel, the amp already had 0.1 resistors on the outputs in series with the speaker, hehe, real easy! It's also a chipamp, it has a hefty transformer, and this badboy has some real power. Amp says 120W/ch x 2 and has a STK4231 in a cheap Sharp amp.. I guess the chip is pushed near it's limits at full volume, but this thing has blown small speakers in the past when being used stereo. :D :eek: However, I've had to replace the chip once. Only thing I don't like other than the small 5600 uf caps, is that I hear a small AC hum if My ear is 6 inch away from speaker. :(

My receiver sounds the best however SQ wise, best highs, no hum. I'm replacing my Subwoofer chipamp with a Single channel Mono Descrete design I just built and still completing that has 5 pairs of MJL4281/4302 output transistors and 5 pairs of 4700uf (23,500uf) for the output, and another pair of 4700uf for the Vas, to drive 4 Ohm load using CFP. With my ear in the speaker, I cannot hear any hum at all, only a very faint hiss :D :). I hear more hiss when I connect the line level connection, than with nothing at all. Hehe, line level hiss. It also sounds much better than the old STK chip amp. I was thinking of using the Transformer from my chip amp, but I'm really wanting to get a Toroid trans that will give me 350W RMS or more into 4 ohms instead.

Chip amps are great, but IMO DIY chipamps vs commercial may be better depending on design, or just a good make that uses them. Now of course it depends on the chip, as some sound better than others, but I still think that chipamps sometimes can be a bit noisy, and have more hiss than a good Discrete design.
 

ThorstenL

Previously: Kuei Yang Wang
Konnichiwa,

Kees said:
Chip amplifiers are extremely simple, but what is the sound quality in comparison to discrete amplifiers? :cool:

This is a troublesome topic.

The problem is that a lot is lumped under "chipamp".

The best chips (power and low level) tend to be as good as at the very least the upper midrange of serious high end discrete designs and then some. Most chips are not as good as that thoug. So, if you know what is good and how to use it well among chips you can easily and relibly get a soundquality that takes a lot of effort and failure to achieve in traditional discrete circuitry.

On the other hand, with discrete circuitry you can choose uncommon circuit topologies which may be ultimatly a better choice subjectively, but be prepared for a lot of hard work.

Sayonara
 

funberry

Member
2005-08-17 3:23 pm
I find that the best Chip Amps ( like LM3875, LM3886, LM4780 )are generally equal in sound quality (noise, distortion) to consummer-grade commercial amps, like those you find in Pioneer home theater systems and the like.

However, they are not up to par with true HI-FI standards.

96 dB SN ratio is not very impressive for a power amp and 0.06% THD is just OK, no better. And most chip amp mfrs specify sound quality parameters in arcane and non-standard ways (like THD specified at 10W for a 50W amp !??), which does not inspire trust.
It's possible to build discrete amps that truly outperform chips by far, and if you really want to know more, you can check out the book "High Power Audio Amplifier Construction Manual" by Randy Slone (Amazon).

There's nothing like discrete Class B to drive really clean 0.0006% THD sound at powers over 400W RMS.
And there is really nothing like the brute force of Class D to drive frequencies between 10 Hz and 100Hz into massive earth-shaking sub-woofers.

National still sells the LM4651/LM4652 pair of Class D 170W Amp chips if you want to build yourself a chip subwoofer system, but hurry, since these parts are on the discontinuation list.

Adrian
 

ThorstenL

Previously: Kuei Yang Wang
Konnichiwa,

funberry said:
96 dB SN ratio is not very impressive for a power amp and 0.06% THD is just OK, no better.

It is true that 96db SNR are not impressive, however the LM3875 specified 98db SNR @ 1W out (unweighted) with the Amplifier rated as 50W RMS this translates into -115db noise below full power. This is near the limits that can be achieved.

As to 0.06% THD being "okay", note that this is specified for all frequencies from 20Hz to 20KHz and as performance limit, carefull observation of the datasheet illustrates 0.003% THD &N for the LM3875 delivering 50W RMS into 8 Ohm. Again, this is well past any issue of good or bad.

As usual, IC's datasheets require carefull reading to be sure that one understands what they tell.

If we where to make a bridged/parallel amplifier with a sufficient number of LM3875 (4) we should be able to produce 200W RMS into 8 Ohm with 0.003% THD and nearly -120db S/N for rated power.

Finally, in the last 50 years or so, ever since Leak made THD a "competetion spec" with their "point one" series there has been a particulary poigniant failure to demonstrate any relationship between "good sound" and low THD, once the THD at full power is in single numbers and monotonic.

funberry said:
There's nothing like discrete Class B to drive really clean 0.0006% THD sound at powers over 400W RMS.

I fail to see any relevance of 400W, to start with most conventional speakers do not like to have such high powers applied at all, ALL MUST be severely distorted and combressed when such powers are applied. Plus, even idealised class B fails to deliver low AUDIBLE distortion at 1W or below, which is arguably where 99.9% of the music happen.

So, I believe you must reconsider your arguments "pro discrete" and "con chip", those you presented do not hold water.

Sayonara
 

funberry

Member
2005-08-17 3:23 pm
The question referred to "chip amps" in general, not to the LM3875, and my comments averaged parts from other manufacturers as well.

There is no doubt that the National chips are an excellent value in the work/reward ratio.

But I don't think you're reading the data sheet right. I sense a bias to read more into the charts than there really is. If you're thinking of the "THD+N vs output power" graph, it's not clear whether the first line after the 10W vertical is a 20 or a 15 on the log scale. The numbers in the rectangles at the top, are the lowest/highest data point sampled. If the rightmost datapoint on the graph is 55.24W, then 50W would fall pretty high up on that vertical part of the curve. The curve low would be at 40W not 50W. And have you noticed that this graph is for 1KHz only?

It's childish to nitpick at a logarithmic curve in this way since it's not an accurate exercise. That's why the manufacturer has to provide tabular specs. And the most authoritative tabular spec is line 4 under Electrical Characteristics: THD+N: 20Hz-20KHz, 40W : .06%. That's what the mfr is HELD to. And it includes a voltage gain figure (26dB), which lends more credibility to the data. When tabular and graphic data conflict, tabular wins. (but they don't conflict, because the graph is 1KHz only). When I compare with discrete, I compare 20Hz-20KHz, apples to apples.

Read the Randy Slone book if you have any doubts. There are clear examples of discrete power amps with THD and slew rate levels one order of magnitude or more better than chips.

But here's the twist: You and I may not HEAR this difference because our ears aren't good enough. That is way I stated that Chip amps are equal in sound quality to home audio and theater systems, which I think is realistic and unbiased.

Stating that chips are the topmost sound quality that can be achieved would be biased.

And I endorsed the LM4651/LM4652 pair of Class D 170W amps, which should do even better as a subwoofer amp, since distortion wouldn't matter so much there.

You should not scoff at powers like "400W" since individual requirements for power vary. I am using 500W currently in my house for parties, distributed among 8x 12 inch speakers, 4 per channel, connected in a series/parallel combination to give 8 ohms net. Who said you should put 400W through 1 driver? that's crazy, I'd rather increase the available motive surface area.

Factors other than sound quality are important too, like protection and reliability, and the National chips excel in that area. A discrete amp will never be as well protected as a 3875, and if it was, it would be very complex and expensive. And a novice would not want to build it. Or may hurt himself doing so.

Cheers
 
I think alot of the disagreement about the general performance of chip amps comes from the implementation. If you build a lm3875 amp using run of the mill parts just thrown together in a circuit with little care you will get an amp that sounds very good but on average just very good midfi. If you futz around with it using the appropriate parts and carefully laying out the circuit with a little trial and error(especially with respect to the input cap value and brand) you will get an amp that is just at the begginning of where amps start to sound "magical". I believe they can be be bettered but only by the truly knowledgeable among us.
 
I agree. I spend about 120 hours on 1 amp and
am buildng a set of chip amp for about $1500 in
parts. The sound is absolutely phenomenal.

One chassis has 3 transformers (with 2 toroids), thick wiring,
MUR1520 diodes, 8 voltage regulators, 300,000uF of
capacitance, 8 LM3875 chips. Holco resistors. Nichicon
Low ESR caps. Blackgate DC coupling caps. Quality
output terminals. Quality RCA jacks. Input buffer.

I've also build some that sounded ok but much more
moderate in price, effort and sound.

Harry
 
Very impressive. Mine is on a much smaller scale. It's 2 channel p2p non-inverting and uses Panasonic fc for input and V+, BC components V-, Shinkoh tantalum input resistor, Pheonix metalfilm feedback resistor, Dale brown body feedback to ground, Avel lindberg 250v @18v, really good industial barrel diodes. All of the parts where chosen through trail and error using Rikens, Caddocks, Rubycon zl/za, Maryland toroid transformer, and other brands. Overall, this is the best amp that I've heard in my audio career. I've owned Manley 300b amps, Atmasphere otls, small diy set amps and other good ss gear. It doesn't nail the absolute realism that the best tube amps I've owned but has an honesty and directness that surpasses them, for me, while sounding almost as good.
 
It depends what speakers your driving. I think for
small 8ohm speakers a single chip should be sufficient.

If you use large speakers or lower impedance you
need to think about parallelling.

My amplifier is configurable. I can run it as 4 x 30 W
or 2 x 30 W or 2 x 120 W or 1 x 120 W. or a mixture
of those. Right now I'm running it as
4 channels 30 W (60W 4ohm). My speakers are
bi amped.

It's a great sound.

The amount of detail is much higher then with the tube.
This results in a sound that sound perceive as more analytical.
I find that complex music sounds much more detailed and
soundstage has widened.

These amps replace my SET AE SE811. These amps are
good but I'm going to put them up for sale. I prefered
them over a Classe CAP151. I also like them better
then Bryston monoblocks and a Krell 300. On the other
hand they are no match for the new chip amp.

This is also true for the pre-amp based on op-amps.
 

ThorstenL

Previously: Kuei Yang Wang
Konnichiwa,

funberry said:
The question referred to "chip amps" in general, not to the LM3875, and my comments averaged parts from other manufacturers as well.

Your quotes on performance are derived from the LM3875 Datasheet (without attribution too).

funberry said:
There is no doubt that the National chips are an excellent value in the work/reward ratio.

I agree.

funberry said:
But I don't think you're reading the data sheet right.

I merely read what is contained in the datasheet. I will insist that my interpretation is correct.

funberry said:
I sense a bias to read more into the charts than there really is.

You correctly sense a bias, it is however one that merely is insistent on getting ones facts right.

funberry said:
If you're thinking of the "THD+N vs output power" graph, it's not clear whether the first line after the 10W vertical is a 20 or a 15 on the log scale.

It is perfectly clear, actually, by simply counting the remaining lines up to 100....

Of corse, if one is prejudiced to disregard the subject under discussion as third rate such minor blunders may be excused....

funberry said:
The numbers in the rectangles at the top, are the lowest/highest data point sampled.

Not quiet, they are the values at a given point along the traced graph, in other words, this graphs numbers in the top illustrates 55.24W @ 0.00344% THD. This is easily ascertained by looking at the other graphs, if one can be bothered to make that minimal 3 second effort.

It is a trivial job to read the datasheet correctly, if one wants to, trust me.

funberry said:
And have you noticed that this graph is for 1KHz only?

I certainly have. Moreover, I noticed that your casually "thrown in" figure of 0.0006% THD was devoid of any qualification of conditions whatsoever, making it completely irelevant for any kind of comparison.

funberry said:
It's childish to nitpick at a logarithmic curve in this way since it's not an accurate exercise.

I believe you do the nitpicking here and I agree, it is childish.

funberry said:
That's why the manufacturer has to provide tabular specs. And the most authoritative tabular spec is line 4 under Electrical Characteristics: THD+N: 20Hz-20KHz, 40W : .06%.

Yes, and the implications is that Amplifier must illustrate less than 0.06% THD & N for all frequencies between 20Hz-20KHz, it is a guranteed limit which at most frequencies will be (considerably) exceeded towrads lower numbers.

funberry said:
When I compare with discrete, I compare 20Hz-20KHz, apples to apples.

Actually, in the only "comaprison" you quoted you did not compare apples to apples, you compared apples to science fiction, eg. to a dim ensionless, unqualified number, without stating test conditions.

funberry said:
Read the Randy Slone book if you have any doubts.

Reading a book generally does not help to relieve my doubts, I find paper very patient to abuse.

funberry said:
There are clear examples of discrete power amps with THD and slew rate levels one order of magnitude or more better than chips.

Again, a completely unqualified statement.

BTW, I agree it is possible to do all sorts of things, both with discrete circuits and with integrated circuits. As low level Op-Amp's illustrate to us it is possible to make IC's with a slewrate materially greater than ANY discrete circuit, if this is done or not is another issue, if it is audible or not, a third.

funberry said:
Stating that chips are the topmost sound quality that can be achieved would be biased.

To say so would be arrogant and biased indeed, just as the opposite claim would be. Luckily I made no such claim, I merely pointed out that to exceed the performance of the best chip's for low and high level requires a considerable effort and just saying "any discrete is better" misses the point.

funberry said:
You should not scoff at powers like "400W" since individual requirements for power vary.

I do not "scoff".

I merely pointed out that in the context of a domestic sound reporduction system any power level requirement significantly above 10W indicates the use of speakers where anyt question of quality is resolved by gross levels of distortion and compression produced by the speakers.

funberry said:
Factors other than sound quality are important too, like protection and reliability, and the National chips excel in that area. A discrete amp will never be as well protected as a 3875, and if it was, it would be very complex and expensive. And a novice would not want to build it. Or may hurt himself doing so.

Absolutely correct, another reason for recommending them.

I agree that it is possible to exceed the sonic and objective performance of the LM3875, however to do so must be considered non-trivial and fraught with a risk of large amounts of smoke being released from output transistors and anyone knows that outout transistors work based on the smoke they contain, realease it and they stop working.... ;-)

Sayonara
 

tmblack

Member
2005-01-13 6:33 am
UK
After 25 yrs of chip amp building I've yet to find the idea chip to replace costly and complex discrete amps.

I started with the LM386 and found it had harsh cross-over distortion.

Now I found the LM3886 to be too analytical and loose in bass compared to premium discretes.

Such analytical performance is something I have never experienced with live music and is quite disturbing to me.
When testing the chip with full range ribbon speakers the sound was only mid-fi.

Using the inverted mode is much superior than the noninverted configuration.

If there is a power chip that sounds like the OPA627 then I could live with it.

Tom
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.