New Build... the GEM amp

[chipwelder]

Hopefully this doesn't become a hall of shamer... it is quite costly...

I am planning to build a six channel amp for my active set-up, it will be this design: http://gmweb2.net/the_gem.htm
Here is an incredibly long thread about it...
http://www.diyaudio.com/forums/solid-state/60546-incredible-quality-amplifier-graham-prepare-your-ears-5.html

I will build two three channel chassis (so that I could put an amp next to a speaker as advised in the website) each with a 600 VA tfx with 30 V rails, I'm not sure, but around 20000uF per rail should be fine...
Below is a pic of the PCB's with blueing, so they don't tarnish before I get round to it...

I need to figure out where to spend on the amps - it is a low parts count amp, so i guess if we follow my philosophy that you hear parts more as they become less, you tend to hear the quality of the parts more...and where to get the freaking 2.5H 80 Ohm coils
Also how to make a three channel amp with massive heatsinks look nice on the floor or on the shelf next to another one...

So if there are any suggestions fire away...

[Viagara]

Did you not plan to(or did build) a six-channel chip amp some time ago?

[chipwelder]

I have a six channel chip-amp... I planned to upgrade it to Rudi's LM3886 design, just to make sure it is not all I need, however the reviews and sims made me decide on this one instead... I would actually in the end like to get rid of as many IC's as possible in my system, as the the systems with less tend to sound truer to the gestalt of the music to me... May be a wild goose chase, a long one at that, but hey I'm still young...

[GearSlave]

Stick to chips, 'Discrete'welder does not have the same ring to it.

[ludo]

For PSU caps on 30V Rails, Electrocomp might still have some 4700uF 35V Hitano EXR, 18x41mm. Stacking them a couple deep should do the trick. If they're all gone now, ask them to check the Jhb branch as well.

20mF is nine times more than a NAD3020 has on the rails per channel. But the idling current is high, so I'd guess a bit of overkill is in order.

Could be cheaper to have the chokes wound? Then you're not paying for the unused secondaries on 12 x 50VA transformers. Could also be more economical to add some/most of the required resistance external to the choke? Then it's easier to keep the choke cool.

[Ampdog]

Why not the CCS version? Iron is often an unknown.

I will have to treat Gert to a lunch/dinner. There are reasons why I have shied away from chip power amps. But that is another story.

[chipwelder]

How's about MJLwelder has a ring to it...

Ja, the 20mF PSU rails, are for the three channels together, But if I can get away with less - that'd be a very good thing... these tend to be able to drive quite difficult things... A locally built stereo 100W CCS version is apparently driving some rather large Apogees.

Those EXR's should be schweet - at least to get the testmule running, but I have them in the chips so I actually find them OK... I was not looking forward to sourcing those blerrie caps... it is also in the feedback to ground which "seems" to indicate importance... will be bypassed a bit, but starting of with something decent is nice... Friend said Nothing but Black Gates below 1000uF   and preferably for this one too... but I'm not so sure I can foot the bill... I was planning to have them wound, was hoping 2.5H and 80 Ohm make for not such a huge blerrie choke...

The CCS version... wouldn't it be noisier? I asked a buddy and either he was not up to checking my soldering / biasing for a more complex things, or he really thinks it'll be more noisy... But I suspect the testmule would have a ccs and a choke just to hear the diffs...

You know my chip amp is probably 8 yrs old... it had a box change inbetween, and some bypass caps added i'm hoping this one lasts longer, as amps and I are really not such easy build buddies... give me something simple like a speaker box that is OK... but this amp thingy ne... dronk weeluis ek se julle - anyway all this means I want to really sort it out on the test-mule before it happens to the final amp... I know what my soldering and especially desoldering looks like.

[GearSlave]

Heh, a lot of what's in that pic looks pretty familiar

[chipwelder]

You would, wouldn't you... well I'm off to get some caps from Electrocomp, luckily have another errand around there.

Hitanos are OK and at R8-ish cheap compared to the Nichicon Muse at 2.5 USD and Mundorff at 12 - 17 USD, besides for the testmule we can make it all Hitano EXR at little extra cost as I have quite a few... Just to sort out any welding issues with cheaper parts...

[chipwelder]

OK, bought enough Hitanos so I have 80% of the caps needed, have crappy 10 000uf Alcons already, will buy a sprinkling of cerafine / starget / mundorfs in the smaller values to play with and to bypass, to hell with 50$+ caps. I want to get these built! No it won't be done in a month...

[GearSlave]

You should've told me you want some fancy caps. I just placed a beeeeg order on Mouser. I'm quite the Muse nut...

[chipwelder]

  Told you two weeks ago!!!  Muchdiffs between hitano and muse?

[GearSlave]

Told you two weeks ago!!!   Muchdiffs between hitano and muse?

F**K!!! You did!!!! Man, I'm a prat! Geesh, thousand apologies there mate! It completely slipped my mind!

I must be honest I've never tried the Hitano's. Also, the current range of KZ Muse's only go up to 1000uF and they get expensive really quickly when you start parallelling them up. Landed you looking at just over R20 each for 1000uF at 50V. I need to go check what the Fine Gold's go for.

[ludo]

There's just one thing about Hitano EXR. It's so weird that many might not believe it at all, but it happened many times. Don't use them in the feedback loop.

I built a number of Mosfet amps some years back for friends. These are pretty standard Self-like amps with PNP diff pair, loaded by a mirror, the usual VAS etc but with CFP Hexfet outputs. There was a trimmer parallel to the 100 Ohm mirror degeneration resistors so the offset could be set very low to something like 5mV. Being a belt and suspenders kind of bloke, I had a 1000uF 63V cap in the feedback loop too. After about one year, the amps in which I used a Hitano cap came back. The cap had vented. In some, the rubber seal at the bottom had been pushed out. The amps with a Rubycon YXF in the same position are working to this day (10 years) Now why in blazes does a high temp cap with no voltage on it vent? The same caps used on the power rails were just fine, and are to this day.

After replacing the vented cap, the amps worked fine. No offset. If the offset adjustment trimpot had made intermittent contact, the change in offset might have been 50mV or so. Not enough to upset a polarised electrolytic either way.

So I don't use the Hitano caps where there is no significant voltage on them. I guess something is strange in their chemistry and they just want voltage. No other cap I've used has done anything like this, not even cheapies. They work a treat as power supply and bootstrap capacitors though. Rather pretty too.

[Ampdog]

Ouch.

Hitano is the only capacitor I have full specs on, so felt rather comfortable with them. There then ....

Not to be derogatory about any brand, but folks would know by now that I take test results over promotional waffle any time. Tests by Cyril Bateman in his well-known series of definitive tests on capacitors several years ago indicated such marginal advantages for Black Gate that it was not worth the expenditure.

Then perhaps to remember: Several hi-fi heavies have indicated that a single large electrolytic capacitor has disadvantages compared to several smaller ones. At the time it was suggested that above generally 2000µF one should begin to think about more than one in parallel. It is also easier to place space-wise on a pc board than one huge round cylinder. I have never been in a position to verify this, but accept the credentials of the thus advisors.