A chip-amp to rival Hi-End - design advice

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

I'm currently in the process of designing a chip amp that I want to sound pretty much as good as it gets. A piece that will rival Hi-End. It is not exactly a "money no object" exercise, but I'm willing to go quite far where expenses can be justified.
My requirements (aside from sonics) include:
- ability to drive big floorstanding speakers well (I don't believe in small speakers)
- power anywhere in the region of 25-150W pc
- it has to be simple to build - I admit I don't have much working knowledge of audio electronics.

So I thought it would be a good idea to ask for advice from some knowledgeable source before I get too far with the project ;)

What have I come up with so far:
- a bi-amped stereo design, based around Peter Daniel's LM3875 kit (DIY Chip Amplifier Kits, PCB's, Components and Information.)
I have built a simple stereo amp based on it before and I was quite pleased with it, its attack, bass control and especially how direct and immediate, "heart-grabbing" the sound is - this is my main criterion when judging the sound. I'm hoping that bi-amping, and perhaps a slight bias into class A will improve things further; perhaps also will enable it to resolve fine details with greater clarity (this is not its strongest side at the minute).
- power supply and pre-amp will be in separate cases
- I intend to use the best resistors / caps I can get hold of (a great shame Black Gates seem to no longer be available)
- four silver-plated copper heatsinks (don't laugh - I have them anyway)
- bare solid core silver wiring throughout pre- and power amp sections. Insulation provided by sandwiching wires between layers of wood that will form the case.
- passive preamp, consisting mainly of a single Noble pot (I can't really justify spending the money on a transformer from Stevens & Billington - and I have the Noble anyway)
- a 500VA +-25V toroid from antrimtransformers.com, with antistatic screen, encapsulated.
- the power amp layout at current stage has all input wires of exactly the same length, output wires are also all of equal length. I'm working on power supplying cables, might come up with a perfectly symmetrical layout for them too.

I have put quite a lot of thought into all this, but I realise I may be going in an entirely wrong direction here or there... So if you have any suggestions, any advice will be greatly appreciated.
 
As far as speakers go, bigger speakers are supposed to be easier to drive than small ones. Given that, it depends entirely on exactly *what* "big floorstanding speakers" you are planning to use. I've used several chipamps from a 110Wpc bridged LM4780 to a 30Wpc single LM3886 to drive my floorstanders (3-way, single amp per channel), and they did it with ease.
 
Thanks for the answer Redshift. At the minute, I'm using Wharfedale Middle Field Monitors (3-speaker, 2.5-way floorstanders, 89dB). I'm actually looking into buying new speakers anyway, and I'll make sure they have good sensitivity.

My main concern is with the actual "core" of the design - that is, the choice of chip and topology, as the array of available options is truly bewildering. I'm sure I won't go far wrong with LM3875s (after all, they started the whole Gaincloning movement) - but this time I'm looking for something capable of putting out a truly exquisite sound, not just "good for the money". I hope this makes sense...
 
You don't need to design anything--it's already been done, and is now being offerred in a group buy at an amazing price. See all threads involving the Mauro Penasa "My Ref" design, but especially the current Group Buy thread if you're really ready to build an amplifier that easily equals or surpasses MANY expensive commercial designs, both chip-based and discrete.

Peace,
Tom E
 
Wahab - I respectfully disagree. Gaincard from 47Labs was based on LM3875, and, upon release, outperformed pretty much every amp in its price range (which was 1500$ for the basic version, if I remember correctly). In light of that fact, your claim is somewhat unfounded.

Madisonears - I didn't mean actually designing the entire circuit, sorry if I didn't make myself clear. I'm looking for a "ready made" design, preferably a kit, which I can then refine. So far, Peter Daniel's LM3875 kit has come out on top, because it's compact, elegantly designed, and very straightforward. Complex designs don't appeal to me at all, they look messy and lack subtlety.
 
Hi,

LM3875/3886 are old amps designs ,dating from the mid 70s at best, that were integrated in a single chip.

Funny, there is at least ONE feature in these chips that was first documented in the 1990's by D. Self in his Wireless World article series.

There s no way for them to compete with a properly designed discrete power amp..

This I may give, if you define "properly designed discrete power amp" as one that outperforms (within the power limits of the NS Chips) the LM3875/3886/4780. I personally would actually agree this, as there is little point for discrete designs that are worse, yet objectively and subjectively a well implemented LM38XX is hard to beat in the 30 - 50W power range.

Ciao T
 
Audiosector!

Wahab - I respectfully disagree. Gaincard from 47Labs was based on LM3875, and, upon release, outperformed pretty much every amp in its price range (which was 1500$ for the basic version, if I remember correctly). In light of that fact, your claim is somewhat unfounded.

Madisonears - I didn't mean actually designing the entire circuit, sorry if I didn't make myself clear. I'm looking for a "ready made" design, preferably a kit, which I can then refine. So far, Peter Daniel's LM3875 kit has come out on top, because it's compact, elegantly designed, and very straightforward. Complex designs don't appeal to me at all, they look messy and lack subtlety.

Hello uncle_leon, I build the LM3875 kits from Audiosector, they sound really good and the best of all is it is so easy to build them.....that a caveman can do it!. But I would "strongly" recommend to use BIG power supply at least 10,000uf per rail. I know it comes with just 10uf cap, but I would not use them.I say this because I try many kind of caps in the power supply and they sound much better with big caps, you will get a punchy bass. There are many people that say the "small" cap gets faster sound..Not true!. just use alligator clips to test them and you decide if you like with big or small caps.
good luck.;)
 
Those NS circuits are LIN topologies with a transistor connected
as emitter follower to drive the voltage amplifier stage...
A current mirror is used as load for the single differential input stage..

In this respect, the design is the one that was famed as doug self s
blameless amp, with the big difference that the LMs use the
dreadful quasi complementary arrangement for the output power stage.
Anyway, the current capability is very poor compared to a discrete design,
and therefore, it cant drive difficult loads in a satisfactory way...
 
Administrator
Joined 2007
Paid Member
An interesting thread... and quite timely as I had been looking along similar lines, out of curiosity really, about designing a high performance amp based around monolithic output IC's.

There are pros and cons regarding IC versus discrete. My initial investigations have been based on the TDA7293 device which allows for parallel "master and slave" operation giving the option for driving very low impedance loads. At this point in time I'm not sure it can rival a top flight discrete design sonically. Technically it can not.

uncle_leon mentions using the best materials etc... which is fine although I'm not a fan of "boutique" parts myself.
Of much more practical importance is getting the wiring correct, particularly the grounds and points of reference for the audio signal. Don't underestimate this... it's where the majority of DIY builds fail.

Good luck with it :)
 
Those NS circuits are LIN topologies with a transistor connected
as emitter follower to drive the voltage amplifier stage...
A current mirror is used as load for the single differential input stage..

In this respect, the design is the one that was famed as doug self s
blameless amp, with the big difference that the LMs use the
dreadful quasi complementary arrangement for the output power stage.
Anyway, the current capability is very poor compared to a discrete design,
and therefore, it cant drive difficult loads in a satisfactory way...

But that all changes if you parallel them :D
 
Hi,

Those NS circuits are LIN topologies

As are around 90% of the discretes, the rest are 9% dual opposite polarity (I call it "mirrored") Lin. Only 1% or so are any interesting (CFB structures by and large).

with a transistor connected as emitter follower to drive the voltage amplifier stage...

And the input differential also has followers (which D Self does not have).

In this respect, the design is the one that was famed as doug self s blameless amp,

Yes. Also note the miller compensation cap connection (more Self).

the big difference that the LMs use the
dreadful quasi complementary arrangement for the output power stage.

Why "dreadful". Even in this day and age the PNP Transistors are still not as good as the NPN. While not LOOKING symmetrical on paper, this output stage actually often shows better electrical symmetry at high frequencies (where it matters more) than the seemingly more symmetric complementary stage.

Anyway, the current capability is very poor compared to a discrete design, and therefore, it cant drive difficult loads in a satisfactory way...

Really? With 10pcs LM3886 I can drive peak currents of around 120A. And with 20pcs bridge/parallel I can generate 50V RMS at > 80A which translates into 4KW into 0.625 Ohm for short duration.

I am not suggesting to actually build such a thing, I am merely pointing out that these chip's are quite easily scalable to higher power, if needed.

Ciao T
 
Dear All,

Like mentioned earlier. Both has pro's and con's. A Chip amplifier doesn't have to be limited to it's application sheet design with a single chip. You can do very creative things with it. Bridge/parallel, Nested feedback enz. enz.

The benefit in my opinion for chip amps is, that they are generally more reliable then discrete designs. You can also benefit from shorter signal lines, which made the original gain-gard so popular. The same time, less matching issues with discrete parts.

It isn't really the discrete or IC thingy that matters, but it is the creativity and imagination (and skills) of the designer that matters. A creative good designer can make a good sound out both. A Bad designer can screw up both. With discrete you have way more options, but also more ways to screw it up.

I have seen some very creative opamp/chip designs with wonderful sound results.

Often we are also influenced by the fact we know which topology get used. Years ago when I wasn't into designing myself but a reviewer for a hifi magazin an a pure listener, I was totally blown away by a certain Linn Amplifier. Years later I found out that this amplifier is based on TDA chips. I am sure if I knew that by that time, I never liked it as much anymore as I did :D :D We are all prejudged in a way.

With kind regards,
Bas
 
Really? With 10pcs LM3886 I can drive peak currents of around 120A. And with 20pcs bridge/parallel I can generate 50V RMS at > 80A which translates into 4KW into 0.625 Ohm for short duration.

I am not suggesting to actually build such a thing, I am merely pointing out that these chip's are quite easily scalable to higher power, if needed.

Ciao T

Currently I am designing such a thing with 12 OPA549's :D Also with a new I think innovative feedback technique (a little bit stolen from Quad Current dumping but then different :D) to reduce the class B distortion and other non linearities. Like I said before it is the imagination and implementation of the designer that is the limit. Not those chips.

With kind regards,
Bas
 
Administrator
Joined 2007
Paid Member
Really? With 10pcs LM3886 I can drive peak currents of around 120A. And with 20pcs bridge/parallel I can generate 50V RMS at > 80A which translates into 4KW into 0.625 Ohm for short duration.
Ciao T

Just an observation :)
Single phase domestic UK mains is nominally rated at 13 amps per socket or just over 3kw in old money.
Whenever claims for huge outpuits are made I always wonder where the power come from... a dedicated spur from the consumer unit maybe.
4 kw into a load ? and at what efficiency from the amp ? The primary draw would be more like 6+ kw.
 
Administrator
Joined 2007
Paid Member
Often we are also influenced by the fact we know which topology get used. Years ago when I wasn't into designing myself but a reviewer for a hifi magazin an a pure listener, I was totally blown away by a certain Linn Amplifier. Years later I found out that this amplifier is based on TDA chips. I am sure if I knew that by that time, I never liked it as much anymore as I did :D :D We are all prejudged in a way.

With kind regards,
Bas

Very true indeed.
 
Another funny story. And some people will flame me for this... so be it :D

In the past my company was specialized in modifications on current equipment, we developed a output amplifier PCB based on an OPA627 with BUF637 with DC servo's which replaced the whole HDAM discrete JFET circuit in the high-end marantz cd players. The SA-1, the CD1 enz. My customers and myself got blown away by the sonic improvement in sound over the original discrete JFET design. People like Charles Hansen would never be agree on this, and I understand their point of views, but I truly believe today with the more advanced silicon techniques, it is possible to produce IC's superior to a lot discrete designs. Doesn't say a IC is always better, of course not! but for sure also not the other way around.

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