DIY Amps

[russel]

I realised with all my playing about with amplifiers I had never heard

1. The Hitachi MOSFET design
2. An LM3886, gainclone.

I hacked up a couple of boards and although the Hitachi is well worth building for a hobbyist, its simple as can be seen from the picture it does sound reasonable, not in the top class like my pass class A, home brewed class A, non switching and current dumping designs its perfectly listenable. Not so keen on the gainclone, turned it off after about 15 mins, just did not get on with it. Both had the same power supply 40,000uF per channel, double balanced choke filter, snubber and inrush limiters. If any one wants the design files for either amp plus the power supply you are welcome to them.

IMG_3747 by Russ Hughs, on Flickr


LM3866 by Russ Hughs, on Flickr


IMG_3749 by Russ Hughs, on Flickr


Edit: the mosfet board will also take the larger package double die FETS, so 100Watts should be easily be obtained.

 

[russel]

Here are the PCBs used, as can be seen the MOSFET PCB supports double die devices, all the PCBS and schematics are in KiCad for any one wanting play around with them.



[url=https://flic.kr/p/2kxBrfF]IMG_3839 by Russ Hughs, on Flickr[/URL]

 

[Arkless Electronics]

You designed boards and had a PCB manufacturer build them at considerable cost and all just to try them out? Have you heard of Veroboard? Only if I was needing to build more than 10 of something would I even consider a PCB!

Both designs can sound good IME and especially the gain clone (I detest that name for them!). I have heard LM3886 amps sound really stunning but they are strange beasts to get decent sound out of... try with PSU's "too crap to be hi fi" for best results! Just 2200uF smoothing caps...

I had results I can't explain with them as well... they seemed to change in sound and "go off"! When I first built my LM3886 amp it sounded stunning, as in Krell, Levinson type stunning. I used only tiny smoothing caps and soon confirmed that indeed adding more capacitance seemed to kill the sound. Over a week or so of experimentation it seemed to lose both it's sensitivity to the PSU and its ability to sound really stunning. After this it sounded slightly better with large smoothing caps as you would expect but nothing would take it back to being gob smackingly good as it had been at first. I still have it and it gets used now and then but whilst it sounds perfectly reasonable it is no better than something like an A & R A60. At lest it seems to be stable at this level of performance...

 

[russel]

Yes, I do build boards to try things out and it’s not at considerable cost, currently a 100x100mm board two layer board is less than pound, min qty of five so it’s beer money, besides I hate building on vero board it sucks. A&R A60 is a good description if the LM3886, it’s average but then it’s dirt cheap so it’s good value for money. The hitachi design is a well worthwhile upgrade from the 3886, the only caveat is that although the bass is excellent it has a high damping factor so it may make small loudspeakers sound a bit bass light, if you have a pair of loudspeakers that need a bit of extra grip then they are just the job.

 

[Arkless Electronics]

I quite like the challenge of making up a veroboard layout as I go along...

When I was at my last employers IIRC there was an £80 charge per board for simple single layer just for setting up costs etc from our local PCB maker.... I guess you're using Chinese PCB pool type thing?

As I said I have heard LM3886 sound really good... as in £2K amp good! I wish I knew why it "went off" from that to A60 type performance...

I found the Hitachi design quite good (fast, tight, crisp etc) but could be a bit "transistory" sounding.

 

[S-Man]

I quite like the challenge of making up a veroboard layout as I go along....

I tend to design veroboard layout on a pcb CAD program. If I make it up as I go along I always realise I've missed something obvious.

I found the Hitachi design quite good (fast, tight, crisp etc) but could be a bit "transistory" sounding.

I tend to agree, however there are so many close variants that sound a bit different, that it's worth voicing/tweaking.
I think there was a thread listing the variants: Claymore, Goldmund and Job (cascoded JFET on the front end), Maplin and of course the JLH variant here:
https://pinkfishmedia.net/forum/threads/kicking-this-dirty-habit.82791/page-3#post-4177208

This is still my favourite amp.

It also sounds different with different PSUs! (Although I know you don't agree Jez).

 

[S-Man]

Here are the PCBs used, as can be seen the MOSFET PCB supports double die devices, all the PCBS and schematics are in KiCad for any one wanting play around with them.



IMG_3839 by Russ Hughs, on Flickr

In your 405 thread you said the Hitachi mosfet amp was bright or harsh. Now you seem to have gone back to it.
Did you change your mind? Or is this different?

 

[awkwardbydesign]

I've still got the Maplin mosfets. If anyone wants them for price of postage they can have them. I'll never use them now. I tested them a while back with one of those very cheap Chinese component testers and they seemed OK.

 

[russel]

In your 405 thread you said the Hitachi mosfet amp was bright or harsh. Now you seem to have gone back to it.
Did you change your mind? Or is this different?

That was the all FET design, there are two designs in the Hitachi data book one is a bipolar/FET design commonly called “the Hitachi MOSFET amplifier” and the other one is an all FET design which is the one I referenced.

https://paul-kemble.tripod.com/sound7g.html

 

[S-Man]

I've still got the Maplin mosfets. If anyone wants them for price of postage they can have them. I'll never use them now. I tested them a while back with one of those very cheap Chinese component testers and they seemed OK.

Yes please, I would like to try them.
I'll send you a PM if I've been quick enough.

 

[Arkless Electronics]

/\ Damn beaten to it I demonstrated some of these I'd built on a Telequipment scope screen as a science project type thing for a parents evening at 6th form college.... Nostalgia's not what it used to be though

 

[Arkless Electronics]

That was the all FET design, there are two designs in the Hitachi data book one is a bipolar/FET design commonly called “the Hitachi MOSFET amplifier” and the other one is an all FET design which is the one I referenced.

https://paul-kemble.tripod.com/sound7g.html

Yep. Hitachi's own power amps were based on this... at least one of them anyway. The well known version was actually designed by Hitachi USA applications dept IIRC

 

[awkwardbydesign]

OK.

 

[awkwardbydesign]

Arrgh! I had a look for them and realised I gave them away last May! Sorry, but my memory isn't what it was.

 

[S-Man]

Never mind. Thanks for looking.

 

[gingermrkettle]

You designed boards and had a PCB manufacturer build them at considerable cost and all just to try them out? Have you heard of Veroboard? Only if I was needing to build more than 10 of something would I even consider a PCB!

Both designs can sound good IME and especially the gain clone (I detest that name for them!). I have heard LM3886 amps sound really stunning but they are strange beasts to get decent sound out of... try with PSU's "too crap to be hi fi" for best results! Just 2200uF smoothing caps...

I had results I can't explain with them as well... they seemed to change in sound and "go off"! When I first built my LM3886 amp it sounded stunning, as in Krell, Levinson type stunning. I used only tiny smoothing caps and soon confirmed that indeed adding more capacitance seemed to kill the sound. Over a week or so of experimentation it seemed to lose both it's sensitivity to the PSU and its ability to sound really stunning. After this it sounded slightly better with large smoothing caps as you would expect but nothing would take it back to being gob smackingly good as it had been at first. I still have it and it gets used now and then but whilst it sounds perfectly reasonable it is no better than something like an A & R A60. At lest it seems to be stable at this level of performance...

Interesting, my mancave system has used KMTech LM3886 boards for a few years now off a power supply with 44,000uF per side and it has always sounded consistently pretty good to me. Different tastes though, I suppose. Were there any complexities in the crossovers you were using them with?

 

[Arkless Electronics]

Interesting, my mancave system has used KMTech LM3886 boards for a few years now off a power supply with 44,000uF per side and it has always sounded consistently pretty good to me. Different tastes though, I suppose. Were there any complexities in the crossovers you were using them with?

What have crossovers got to do with it? Certainly nothing that I'm aware of.... Spendor BC2's FWIW.

 

[gingermrkettle]

What have crossovers got to do with it? Certainly nothing that I'm aware of.... Spendor BC2's FWIW.

It was more that one of the main identified weaknessss of the LM3886 chip is low impedance performance so a highly reactive impedance with high phase angles could possibly show it up.

 

[Arkless Electronics]

It was more that one of the main identified weaknessss of the LM3886 chip is low impedance performance so a highly reactive impedance with high phase angles could possibly show it up.

I used mine very successfully with my 4 Ohm KEF R105.3's also.... As I said for some unexplained reason it only sounded really special for a short time and then "went off" so it sounded about on a par with numerous generic "not bad" amplifiers. With the very small caps it did "lose it" at higher volumes and get a bit harsh and uncontrolled but at normal levels it sounded very special indeed... for a short while anyway.... Adding big smoothing caps makes it more controlled and allows high volumes without the sound changing but it killed the way the hi fi would "disappear" leaving the musicians in front of you.. during its brief "magic" phase anyway! Now when you go back to small caps it still loses it at higher levels as before but the magic doesn't reappear at lower levels. I haven't a clue way... it still measures just the same.

 

[a.palfreyman]

^^
Oh dear, this will please the subjectivists and annoy the 'measurists' completely...
Been reading up on the LM3886 as a possible interesting project for TV speakers. There seem to be a lot of variations / prices for bare PCBs. According to Neurochrome (LM3886 Done Right) the pcb layout also seems to be be a big part of the sound quality. $34 each for a pair of pcbs plus postage is a bit rich for me though TBH. I see kevinmount (ebay) also has pcbs at about £4 each but there doesn't appear to be any local decoupling right at the chip. Looking at the 'Done Right' version, this has large electros, small electros and then small film right by the chip. Any thoughts?