GainClone Design -- PSU, which chip, grounding, clipping queries

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

First of all; great forums. I've been reading here on and off for a couple or three years, but I usually don't have much to add. :) It's suprising how often diyAudio comes up in electronics-related Google searches (audio or otherwise).

So, my 30-year old discrete component 20W mono amp is biting the dust. It makes all kinds of racket, unless you slowly flip the power switch up and down a bunch of times. Then it's good for a while. It also colours the hell out of the sound, muddies everything up, and in general drives me nuts.

Time for a new amp, I guess! But I have to do it on the cheap, with multiple small purchases so I don't get yelled at. ;)

My requirements are fairly simple: I need to build a new practice instrument amplifier, and I'd like it to double as a stage monitor if possible down the road. The quality of the amp must be good, but doesn't need to be insanely good (i.e. if it sounds as good as the 1988 Harmon-Kardon HK440 tuner/amp in my living room, I'll be very happy).

For practice amp mode, I'd want to drive a couple of 8 ohm speakers in stereo. Using some online calculators, I've figured that 20W of power per channel will be *plenty* (peak 103dB SPL where I'm planning to put them).

For stage monitor mode, requirements are a little dicier. I need MORE POWER (but I have no clue how much -- need to compete with guitar amps, PA, etc), but I don't need stereo. So, my thinking was to bridge the amp with a single 8 ohm wedge pointed right at me.

I was originally going to build an old-school discrete transistor amplifier, because my last experience (20 years ago) with amplifiers-on-a-chip wasn't so great. Well, except for the time I accidentally built an AM radio instead. That was pretty cool.

But I digress. I've read the reviews online, and looked at project complexity, and it looks like a Gain Clone amp is right up my alley -- not too difficult, lots of information online, lots of designs available beyond the manufacturer's sheets, and good reviews from the community.

So -- first question -- the LM3886 (or the dual-LM3886-on-a-chip) is probably my only choice here, if I want to run bridged into 8 ohm, right? The data sheet for the LM3875 suggests that it won't drive a 4 ohm load.

After picking the chip, I need to decide whether to design/build a board for it. Having no real amp design experience, and barely-OK PCB etching facilities, it looks like going with one of BrianGT's boards is the smart thing to do. Actually, doing some math against Digikey, it looks like BrianGT's kit sare very economical -- the break-even point for the PCB cost is *negative* if you build the kit exactly as spec'd (either that, or I can't add. Which is possible).

So. LM3886, BrianGT board. Can I put the LM3886 on the LM3875 board, and benefit from the Zobel network? Do I need the Zobel network if I always plug into the amp with a good preamp and the same cables? (I have a Carver rackmount preamp around here somewhere...) The AC enviornment is subject to change without notice, as are the instruments plugged into the preamp... or mixing board...

Next question set -- I'm having a devil of a time trying find a "recommended" transformer for a price I want to pay at the moment. So:

1 - Can I use a matched pair of transformers? If so, should I treat the pair as a double toroid, paralleling the primaries and hook the secondaries up to Brian's snubber PS as usual, or should I also wire the secondaries in series and treat the connection point as a center tap?

2 - If I find a suitable center-tapped transformer (e.g. 24-0-24), can I hook it up like this? (Diagram actually shows a double-transformer set up as a center-tapped secondary, to also illustrate my musing above. Yes, I know, I can't draw, and paintbrush is all I have on this computer).


3 - When running as above, will the 60Hz ripple in the neutral cause a noticeable effect in the sound? Or will the DC just happliy dance above it?

4 - What is the right way to earth-ground the setup above? Primary neutral tied to ground, tied to chassis? What about the signal grounds? (input and output) Do they stay isolated? From each other?

5 - Speaking of earth-ground, what is the right way to earth-ground the dual-primary, dual-secondary transformer setup with BrianGT's snubber PS board?

6 - Do the input and output signal grounds have to be at the same levels on a dual LM3886 setup? If not, can I do a "naive" bridge by inverting the input wiring on one amplifier, or do I still need to put some buffers (e.g. unity TL072) in front of it?

7 - Star Ground -- Anybody see a problem using a 12" bus bar as the center of the star?

8 - How can I calculate how big a transformer I need for a given output wattage? I went through the National XLS file, added Pd and output power for 35V rails, popped in some head room, and came up with something like 250W of power required PER CHANNEL. Using a power factor of .85 (which, BTW, I just pulled out of my *** -- anybody know the right number?), that yields 294VA * 2 required for this amp. That seems kind of excessive!

Note -- my primary goal for power supply selection is NOT to maximize SPL, but rather to avoid unheard clipping, to help preserve the voice coils in my tweeters. Arbitrary advice welcome in this regard.

9 - When you guys are talking about using ~180VA transformers, is that 180VA per secondary, or 180VA total? If total, how did you arrive at your sizing? Are you assuming that the big caps will supply extra power when bass notes sound?

10 - I've got my eye on a couple of 24V 2A transformers for the first incarnation of this project. These should yield around 56VA each if my numbers are correct. Am I correct in thinking that this would be roughly equivalent to a 100VA toroid, dual-primary, dual-secondary setup?

11 - What happens when your power supply can't deliver enough current? Clipping on high-energy notes? (e.g. bass beats)

12 - What happens when your power supply is running at a lower voltage? Say, 18VDC? Lower volume, lower power output? If I chose lower voltage transformers, how does this affect the power requirements of the circuit?

13 - Is there a good way to reduce the probability of clipping in hardware, when running a power supply which is too small? For example, could I easily lower the gain or attenuate the signal to make sure I never go too loud?

14 - Is there a good way to find out exactly when I'm going to clip? My thoughts at this point are to drive a 4ohm sandbar resistor with the outputs, watch the potential across the sandbar with a 'scope, keep an eye on rail-to-rail voltage with a DMM, and feed a sine wave through the input with an audio signal generator going through the preamp I plan to use in full-blast mode. Then, either turn down the gain or attenuate the input signal in the amplifier until I see no sine clipping and no rail-to-rail voltage sag. Does that plan make sense? Any ideas what frequencies I should be watching with this setup? I was going to test 1KHz, 100Hz, 40Hz and call it a day.

15 - Speaking of volume (power) control, I see you guys seem to prefer to using input attenuation instead of changing the gain. Why is this? Is it just because you can't turn the gain any lower than 10dB over unity? Wouldn't you get a lower noise floor by driving a hotter signal? (I suppose an amp with two knobs, marked "gain" and "input attenuation" would confuse folks).

16 - Anybody have a clue how to calculate the heat-sinking capabilities of an enclosure? Or should I just shoot it with an IR gun during testing as in question 14 and make sure it never gets above 50C? (The enclosure I'm planning to use is steel of some sort, about 3/16" thick, ~5x15x3.5")

17 - Anybody ever thought of building a power supply which voltage-divides the line input with capacitors, then rectifies it and uses it directly? Okay, of course, the answer is yes, so, why isn't it done in the DIY community? It would sure be cheaper than dropping a hundred bucks on a transformer. Noise? Non-isolation of speakers from AC mains? Shock danger due to caps on the line cord?

Thanks for any and all responses. I really see a finished GainClone in my possession in the next couple of months!
> As a newbie here also, I can GUARANTEE you that all of your
> questions can be answered by doing some searching.

Do you really think I could construct a post with such specific questions *without* doing any searching?

I currently have about 35 hours invested in this project between reading this forum and associated collateral (National data sheets, user's guides, etc.)

Were I to have posted this query are the start of my search it would have read: "Need an amp. It has to be loud enough. And Cheap. And sound good. What do you suggest?"

And you're right, there are something 625,000 posts on this site. I haven't read them all. But I most certainly have done a tonne of homework.

I'm a newbie at this too. :)

I'd recommend grabbing one of BrianGT's kits, building it, trying it out and deciding if that suits and move on from there.

You can use single or matched toroids with those kits out of the box, so you don't need to worry about PS design. You can swap to the snubberized PS later, or opt for it from the start. People seem to generally opt for 120-160VA per channel, but some just go with what they have.

I don't think you can put a 3886 in a 3875 board. The 4780 boards have zobel circuits in the current revision, maybe they all do. Ask Brian.

You can experiment with the zobel and your preamp - decide which works best for you.

Too many questions! :D

From the sound of it, anything you build is going to outperform your old amp considerably, so make some choices and warm up your soldering iron..

Oh, and good luck!

Hi, Michael!

> You can use single or matched toroids with those
> kits out of the box, so you don't need to worry about PS
> design.

Perhaps I worry too much. I once blew a VERY set of expensive tweeters through excessive clipping. And then, not knowing why, I blew them again. I had assumed it was a line surge or something (I knew the speakers were way beefier than the amp), so I added a surge suppressor the second time. Fortunately, the second time they blew, I noticed that the amp was smoking like hell. Opened it up and found the transformer was so hot it ignited the schematic diagram, transformer paper, and scorched the PCB. A more knowlegeable person than I (the owner of the speakers) then explained WHY the tweeters blew. I felt like a real idiot. And got out some more cash.

> I don't think you can put a 3886 in a 3875 board. The 4780 boards
> have zobel circuits in the current revision, maybe they all do.

Yeah, I'll probably build the 3886 board as-is and then decide on the zobel network as you suggest. Looks like the hard part (design) has already been done, I should be able to lift and graft from the 4780 boards. I think. :) I'll have to figure out if the impedance of the parallel inputs on the 4780 board will affect the network Brian used there. Wish I had more AC theory under my belt. :(

> Too many questions!

Pick and choose which you want to answer, any info is greatly appreciated. :)

FWIW -- I have a habit of wanting to get my ducks in a row *before* I start building. I'm (by nature and circumstances) a very spurty builder. I pull out/assemble my work area on a free weekend, and work my backside off 'till I'm done. If it's not done in a weekend, it could be *weeks* before I can get back to it -- and with parts requiring long lead times (or a 300 mile drive) around here, that means I want to source everything I think I'll reasonably need before I even pick up my iron. Like my stupid Toaster Oven, it's been missing a 6V Zener for almost a year now, because I trouble-shot it and then packed up my tools. :)

> People seem to generally opt for 120-160VA per channel

Okay, that's helpful. I wasn't sure if the quoted numbers I've seen are per channel or per pair of mono amplifiers. Any idea how those numbers were achieved? Just trial and error?

The reason I'm curious is that the Overture Design Guide -- -- says [madly typing] that an LM3886 running 35V rails can put out 61W and dissipates 31W of heat when running an 8 ohm load. This makes for 92W required just for the chip, or 108VA(?) per channel.

But then, if we change to a 4 ohm load at 28V rails (bigger and the spreadsheet throws thermal warnings), we're making 73W of output, and dissipating 40W of heat. For a total of 113W or 135VA(?) of power required (just for the chip) per channel.

Hey, these numbers work better than the numbers I had last night. By over 100VA per channel! I wonder what I did wrong? Thanks for making me double check that!

Anyhow, so that means 120VA per channel is about as low as I'd like to go in 8 ohm mode, and 150VA for 4 ohm -- and another 75W for bridge mode. [What happens if I go too low? Can I prevent speaker-damage-problems by reducing gain or attenuating input?]

So. If I want to bridge these puppies, I'm looking for a minimum of ~250VA per channel. So these guys running 180VA dual toroids (90VA per secondary, right?) must not be running their amps full-tilt 100% of the time, and are definately not running in bridge mode. Or maybe their transformers are putting out more power than the sticker says they can...

> From the sound of it, anything you build is going to outperform
> your old amp considerably,

Yeah, practicing tonight was painful, when all I could think about was getting rid of the mud so I could hear my boogie-woogie bass lines a little more clearly. :)

> Oh, and good luck!

Thanks! Your post was very helpful, at least I feel confident in what the *requirements* for power for these things are now. (As opposed to ... I used X and it worked for me -- makes shopping for anything other than X difficult).

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