LM38xx : Left behind by all those chips ! :D

[youyoung21147]

Hello,

I'm now ready to build my GainClone and already have some parts :

- 10 Panasonic FC 1500mfd/50V caps
- 4.7mfd MKP input caps
- dozens of 1% metal resistors
- 1.3°C/W heatsinks

and plan to order those :

- ALPS 50k log potar
- 225VA toroidal trafo ( +- 25V secondaries)
- 8 ultrasoft 15A 600V rectifiers to build a double bridge
- 47nf and 100nf snubbers
- all the connection stuff etc...

In fact, I need to choose the main part : the chip amp !


I've read much on the forum about the LM3875 and LM3886, and also analyzed the datasheets :


What I could read is contradictory :

- according to the datasheets, the LM3886 has better distorsion figures, twice current driving capabilities compared to the 3875(meaning best damping factor), and a 19v/µs slew rate.


- according to people, the LM3875 sounds better due to the absence of mute.


So I have a little questions to ask you, since National doesn't allow the order of different samples to compare them.

- Which one of these chips fit the best the parts list I have ?
- Which of these has the most detailed sound ?
- Which of these has the best damping factor ?
- Is it possible to shunt the mute function on the LM3886 ?
- Is it possible to put two chips in parallel directly soldered on the same components ?
- Is it good to place two capacitors in parallel on the chips to increase capacitance, once snubberized ?
- What is the preferred operation mode for them ? Inverted ? Non-inverted ?


My speakers will be a pair of fullrange FX-120 TQWTs.



Thanks in advance !

[Nordic]

Ok, I'm a bit of a noob too, just started some 2 months or so ago... but have been spending all this time on the gainclone (I'm disabled - no job to go to = LOTSA TIME....which you need to get parts here)

Will try to answer some of your questions untill the masters step up...

Firstly I have some questions for you... think it can only help to get a firm idea of what you are after... I have about 50% of the total components left I bought to make my gainclones, becasue I couldn't settle for one thing...

Are you planning to do an inverted or non-inverted amp?

With the basic inverted sceme a straight pot ahead of the input will affect the gain which is calculated by deviding the feedback resistor by the value of the input resistor (which the pot would become a part of)...

A more common approach for the inverted setup (which I can recommend to beginners for its simplicity, I can not imagine my amp sounding any better, but I do believe it is possible after reading some posts) is to use a buffer or preamp as the volume control. (I have yet to build one)

OK now for your questions

Sound 86 vs 75... from what I understand the concensus is that you probably won't be able to tell them appart in blind listening tests...

Diffirences between 86 and 75
86 can drive load down to 4Ohms 75 can drive minimum of 8 ohms only.
86 have mute function
86 rated to produce 18Watts more power
75 probably needs less components for circuit so lends itself better to point to point wiring.

If the 86 sounds better than the 75, it must be like angels singing... these babies are crisp and mellow at the same time, a bassdrum produces a percusive sound and a bassguitar or contrabase will produce a warm real tone.

I have constructed mine from mostly pirate parts, ongoing process to replace less worthy components with better ones... they all pretty much sound equaly good... almost everytime though with these chips it seems less parts sounds better (maybe why 75 has many punters)

Funny thing is I find on sites I read many people wire in a component do a quick listen then swap it out for an alternative, then decide between the 2 without allowing for burn in (I used to), I find that my amp's sound is realy changing the longer it plays... it keeps getting better, well first it got a bit worse....

I have another theory that says alot of the chip's burn in actualy has to do with the thermal compound, which "changes phase" the first couple of times it heats up and cools down to form better contact and thermal bonds (this is from experimenting alot with computer cooling and reading quite a bit of thermal compound leaflets...)

Right where were we... lol I am actualy enjoying this, hope you are too...

Parts I used weer the following;

2x 1000ufd electrolytics on pin 1 and 4 other 2 pins connected together formed my powerground star (dont think 1500ufd going to make a diff either way between 75 and 86 choice) If it sounds too boomy mybe get some 1000s. You could solder an 0.1ufd cap over the pins, I tried with and without, and for the current combination I have, it made no AUDIABLE diffirence to me and my wife, so I removed mine.

10k input resistor, try to match 2 using a multimeter (some people apply another between pin 7 and the signal ground star).

4.7ufd input coupling capacitor ((BIPOLAR/NONPOLAR, will be marked NP or BP)(you can try 2.2, 3.3 4.7, but give the amp time to burn in for a week or so before you rip em out.)

220k feedback resistor, once again makeing sure you have to equal ones will ensure equal gain in your channels and even soundstage.

10ohm 2watt resistor to connect between power ground star and case
100nfd (.1ufd) capacitor (not electrolytic) in paralel with 10R resistor...

Also what are you planing for a PSU, straight rectified, Snubberised, or swithced mode PSU....?

For alot more please visit http://myweb.tiscali.co.uk/nuukspot/...gainclone.html
Remember to click the Gainclone Index item on the menu... to see other ways/options...

[youyoung21147]

I'm planning to use the non-inverted circuit because it seems more "rational" and stable. I'm sure to get good results this way and I furthermore respect absolute phase of the source. I assume there is very little difference anyway. The higher imput impedance dispends the use of a buffer IMHO.

I will put a resistor before the pot to avoid having a 0ohm load for the source when the pot is at the lowest level.

I will use a 4.7 MKP cap for the DC decoupling, the best available type. Maybe nothing, as far as I will use my computer's soundcard as source : it has 0 DC offset !

I'll maybe add a buffer later but I first want to use it "as is".

Concerning cooling, I'm a cooling-maniac too ( lol ) and will use some generously sized 1.3K/W heatsinks with CPU compound I have in the shop. I will also order little fans for hot days but I doubt I will need them.

For the feedback I assume you mean a 22k resistor ? (not 220k)

I'll put a 22k input resistor.

I will use a PSU based on a 2*25V 225VA toroidal trafo, followed by a double discrete diode bridge and directly driving the chips (with the 1500mfd on their legs)

I will snubberize it according to Carlosfm with 100nf and 2ohm resistors on the chip

I'll also put 100nf + 2ohm on the output.


Concerning the chip choice, I'm still divided but I have some reasons to use rather the LM3875 :

It has a better signal-to noise ratio (98dB vs 93dB) , meaning a lower stand-by noise which I particularly hate.

It hasn't got this ********** mute function that is just here to pollute the sound. In the datasheet it isn't clear if there is signal going thru this pin.

As I mainly listen at moderate volumes, I assume it will be a better choice.

Well, they could allow the order of different samples at the same time instead of allowing 5 sample orders a week !


Thanks for advice

[Nordic]

For inverted configuration the resistor between output and input in the one I did was 220k and the resistor between input capacitor and input is 10k giving a gain of 22.

I'm also using the 85, it sounds lovely! If I want mute, I turn the volume down.

Suppose what it comes down to is, inverted = easier to build, but requires buffer/pre... so overall it probably balances out...

[scottw]

according to the datasheets, the LM3886 has better distorsion figures

Regarding the distortion, it may make a difference that the 3875 was measured at +/- 35v and the 3886 at +/- 28v. Never heard this mentioned here but it is in the fine print in the data sheets.

I've only played with the 3875's but have been thinking of trying the 3886's. If your speakers are hard to drive, or less than 8 ohm, or you need the current, the '86 may be the wise choice. As far as cooling, I have two chips on a 10 inch piece of 4 inch Al channel and near the chips the temp is usually only about 5 degrees F over ambient at normal listening levels(but my speakers are ~ 94dB/).

scottw

[G4ME]

3875 can drive 4 ohm loads, just as long as your primaries are low enough to not exceeded the max current i think 22V torriod is fine

[youyoung21147]

Originally posted by G4ME
3875 can drive 4 ohm loads, just as long as your primaries are low enough to not exceeded the max current i think 22V torriod is fine

Mmmm... You mean in my case, +-18V toroidal transformer would be better ?

I'm shared between the clipping risk (lower secondaries) vs. the sound quality issues (higher secondaries).

I assume increasing the secondaries voltage increase the rush of current through the speakers at high levels, but doesn't it also limit the intensity going through the PSU of the chip at normal levels ??

I see that most gainclone use +-25V trafos but maybe a more "audiophile" approach to gainclones would be a "quieter" PSU with a +-18V toroidal ??

[Nordic]

From the datasheets it looks like it was intended to run at 40Watt.

What I'm still waiting to hear is if one can stack 2 of these chips....

[joensd]

The FX-120 are 8ohm loads and if the amp is primarily intended for them I wouldn´t worry too much.
"Higher" secondaries and the LM3875 should be absolutely fine.
I even think 22VAC secondaries is not high but a good middle-of-the-road value.
I believe Peter Daniel used 22VAC in his first GC´s and he certainly must have given the choice of transformer more than just one thought to create a universal amplifier (for his clients).

You could even go higher for a few more watts but that´s up to you.(IMO 22-24VAC are just right for 8ohms)
With ~+-30VDC you´ll have a good ~40W into your fostex which should be plenty.
Check the calculator to see what happens:
http://www.national.com/appinfo/audi...gn_Guide13.xls

greets
Jens

PS.: FYI:
I don´t recommend it but I´m driving my sub (4ohm) with a LM3875 at the moment (+-25VDC) and it´s fine for domestic use and my usual listening levels.) So yes it DOES work.

[joensd]

What I'm still waiting to hear is if one can stack 2 of these chips....

The forum actually is full of answers to that question.
I believe there actually must be dozens of PCB layouts for bridged and parallel duty of such chips as well.
Google and download for "AN-1192"; that´d be a good start/read.