Post by Owen Y on May 15, 2020 14:31:03 GMT 12
Around 20 years ago, 47 LABORATORY (Junji Kimura) produced the Model 4706 Gaincard amplifier design: - The world's smallest number of parts in an amplifier (9 per channel). - The world's shortest signal path length (32mm including parts lengths). - The world's shortest negative feedback loop (9mm). - A high-end amplifier designed around one op-amp (per ch.). - Very small PSU filter capacitor (1,000uF). - Output power 25W+25W (also 50W version). STEREOPHILE (2004) said: " ...a sense of transparency: a crystalline clarity that seemed to extend from the lowest bass to the highest treble. The resolution of fine detail was quite extraordinary... The sound had a directness, a feeling that music was being reproduced with a minimum of artifacts getting in the way. " 
|
Post by Owen Y on May 15, 2020 14:52:30 GMT 12
This design has inspired & spawned numerous 'Gainclones', one of the most well-known is by AUDIOSECTOR (Peter Daniel), who offers PCB kits. Jeremy has had one of these in kits for a while & is now wanting to build it. For US$68, you get 2 amp PCBs with components + 1 PSU PCB with rectifier components. To these, you need to add: - Power transformer - Chassis - Volume control - Input, output socket hardware. - Mains power input parts.  (One PSU bd unbuilt)  (Panasonic FC caps)  ('Ultrafast recovery' diodes.)
|
Post by sadface on May 15, 2020 16:22:40 GMT 12
Cool stuff. I do so love the ENIG finish on the pcbs. How far off running are they? If it of interest I have some spare pcbs for Carlos Felippes's snubberised and regulated psu for gainclones. I'd be happy to gift you one or two. The lastest incarnation of my gainclone boombox uses 1 board to power the channels.
The heatsinks dont go above ambient until one is listening medium loud, at least into a 6 or 8 ohm load. I have not tested the unit on a 4 ohm speaker yet. 2 of them in dual mono should prove bullet proof powering a gainclone.
|
Post by Owen Y on May 15, 2020 16:53:22 GMT 12
This design is specified to use a 25k or 50k vol pot on the front end, if you wish. (The outline schematic of the basic amp is below - R1 is omitted.) We have a 100k pot on hand here & I think that there should be no serious implications swapping instead of a 50k pot. Input impedance (of the 100k pot in parallel with the 22k shunt R) would be a little higher (18k) than when using a 50k pot (15k) - probably of little consequence with modern sources. However 'source' impedance seen by the LM3875 would be increased, especially at higher volume settings - would this affect the performance of the op-amp and/or the amp input sensitivity? (Perhaps colinf, you could offer an expert comment?) 
|
Post by colinf on May 15, 2020 19:18:16 GMT 12
Hi Owen, as long as the 22k is there the dc offset from the LM3875 won’t be too bad. Without the 22k resistor the dc offset from the amp might go above 200mV at mid volume control position, not what you want across your speaker! Having a 22k resistor from the wiper of a 100k log pot will distort its log characteristic, so you may find you need to turn it up quite a bit at first then the volume would increase rapidly as you go past say 1 o’clock position. Check the dc output of the LM, it should be less that 50mV. From experience these chip amps do better with a capacitor in series with the NFB ground resistor, or a dc servo opamp, to keep it at less than 5mV. A cap in series with the NFB resistor degrades the sound slightly so a dc servo would be better. A digital dc servo would be even better and I’m designing it as we speak!
AMR-iFi R&D
|
Post by Owen Y on May 15, 2020 20:18:29 GMT 12
Hi colinf - thanks. With say a 50k vol pot, would the volume curve characteristic be somewhat less 'distorted' do you think?
|
Post by Owen Y on May 15, 2020 20:45:39 GMT 12
Hi colinf - DC Offset, from the build manual.... " This particular amp (right channel) measures 32mV offset when volume is completely down (input shunted to ground) and 70mV when volume is at the maximum (15K combined input impedance of 50k pot and 22k input shunt resistor).
This is rather average chip when comes to offset, and for selected ones the offset is usually at 50mV or less at max setting.
The offset will be also lower when low impedance output preamp is connected.
When we switch the amp off and no speakers are connected, we can observe that for a moment or two the offset jumps pretty high to 4V or so. This is normal behaviour, and
when load is connected (you may try it with 10R resistor across binding posts) the offset will stay low when turning the amp off. "
|
Post by Owen Y on May 16, 2020 14:19:18 GMT 12
How far off running are they? If it of interest I have some spare pcbs for Carlos Felippes's snubberised and regulated psu for gainclones. I'd be happy to gift you one or two. Many thanks, sadface. Firstly, J will get them going, that will be a big achievement. Then we'll see how he feels about the sound. We'll take one hurdle at a time 
|
Post by colinf on May 16, 2020 19:45:54 GMT 12
Hi Owen, yes 50k pot would be better but still not perfect. You could always use a 20-25k pot or less and not use a 22k resistor. But then If the pot loses contact briefly as it gets old and worn, the dc offset from the amp with a disconnected input terminal will destroy your speaker! Best to use a resistor, with a 25k pot use between 47k and 100k. A TVC used with a shorting switch would allow the LM to work with very low dc offset and would be best sonically. With a sensitive speaker 70mV is quite a lot, aim to get it under 30mV with a normal pot and no dc servo.
AMR-iFi R&D
|
Post by sadface on May 16, 2020 23:17:26 GMT 12
Well that is an interesting lesson in ac vs dc coupling.
Ac coupled gainclones seem much much safer on the speakers.
I get 0.2mV and 0.5mV at the outputs of my jaycar boards.
|
Post by colinf on May 17, 2020 21:59:03 GMT 12
That’s good, very low as it should be with a cap in the feedback line. You can upgrade the cap for better SQ, or use a whole lot of small value caps in parallel so that the ESR is very low and distortion lowers appreciably. For more interesting reading about this you could check out Lavardin amps and the alternative methods they use for dc stability, if you’re game!
AMR-iFi R&D
|
Post by sadface on May 18, 2020 17:55:40 GMT 12
That’s good, very low as it should be with a cap in the feedback line. You can upgrade the cap for better SQ, or use a whole lot of small value caps in parallel so that the ESR is very low and distortion lowers appreciably. For more interesting reading about this you could check out Lavardin amps and the alternative methods they use for dc stability, if you’re game! This was the thinking behind the massive 22uF Wima MKP feedback caps I used on my LM1875 boards.
Now that the jaycar LM3876 boards have come out of the boombox, I might get around to installing these 2....
My next order from JLCpcb will include some of these: I found this one on DIYaudio. Supposedly, this is about as good as an open source design LM3886 gets. Bettered only buy the commercial offerings like Audiosector and Nuerochrome.
So many amps to build and not enough time or money to make them all 
|
Post by sadface on May 18, 2020 18:03:21 GMT 12
How far off running are they? If it of interest I have some spare pcbs for Carlos Felippes's snubberised and regulated psu for gainclones. I'd be happy to gift you one or two. Many thanks, sadface . Firstly, J will get them going, that will be a big achievement. Then we'll see how he feels about the sound. We'll take one hurdle at a time If they are up and running before I get around to reassembling the boombox, then I will gladly lend you the built PSU to test.
Its set for +/- 24vdc.
At some stage I would like to do something like a head to head with a cap bank to see how much the different PSUs are noticeable.
Also, I have a small sized CRC board in the works that takes 18mm diameter caps because I got 25 of these for $1.5 each on clearance a few weeks back. My first version takes 24 caps in 4 banks of 6. I've decided 40000uF per rail is a bit excessive for a pair of gainclones. Version 2 is in the works for 4 banks of 3 caps for 20000uF per rail (probably still excessive but when the caps are so cheap, why not?)
|
Post by colinf on May 18, 2020 18:16:27 GMT 12
Different brands and types of power supply cap sound different as well. It would be good to have a selection of them, not just regulation vs one brand of cap. The smaller on-board caps make a difference as well. They handle the sharpest of the class B switching distortion on the power supply pins and so generate the most ringing. Power supply rejection on these chips becomes quite low at ultrasonic frequencies so these on-board caps make a significant difference, and especially their grounding.
AMR-iFi R&D
|
Post by sadface on May 18, 2020 20:41:29 GMT 12
I was inspecting the jaycar boards recently and I was quite unimpressed with the layout. The on board decoupling is a mile away from the chip and the feedback path is hilariously long. The interesting part about the boards I just posted is their designed for 3 on board decoupling caps. 2x 470uF, 1x 47uF polymer, 1x 4.7uF X7s cap. neurochrome.com/pages/supply-decouplingThe impedance curve looks much nicer with the arrangement.
|
Post by colinf on May 18, 2020 21:31:13 GMT 12
That link explains it well!
AMR-iFi R&D
|
Post by Owen Y on Aug 23, 2020 22:48:47 GMT 12
Apologies, we've all been time-short around here, but we've just powered this thing up tonight &... wow. This LM3875 'Gainclone' sounds REAL good - hooked up to my Spendor SP1s, so transparent, so clear, clean, nice subjective bandwidth, oodles of harmonic richness & decay, natural & effortless, out-of-the-speakers imaging... with absolutely zero hours on it. Pics to come, but for US$98... I'm shaking my head.
|
Post by colinf on Aug 23, 2020 23:01:42 GMT 12
They can be good can’t they. So is it LM3875 used, or LM3876 (or LM3886) ? Because the LM3875 has been out of production for a while now and only counterfeit ones are available.
AMR-iFi R&D
|
Post by Owen Y on Aug 24, 2020 10:17:19 GMT 12
|
Post by Owen Y on Aug 24, 2020 13:28:19 GMT 12
The Audiosector LM3875 Gainclone cosmetically unfinished but doing some 'burn-in' on top of one of the Spendors... 
|
Post by colinf on Aug 24, 2020 17:41:22 GMT 12
Lovely construction as expected! Nice heatsink as well. I like the pic of DC offset measurements for different chips. The difference between 0 resistance input impedance and 15k is revealing of input stage variability between different chips.
AMR-iFi R&D
|
Post by Owen Y on Aug 24, 2020 19:55:55 GMT 12
We measured DC offset at the output posts (with inputs shorted & at 0/max vol) - 13mV-18mV. We'll check it at listening vol (some way up the 50k ohm log pot).
|
Post by Owen Y on Aug 25, 2020 17:46:23 GMT 12
I've been pondering why this amp sounds so good & why the average audio enthusiast should pay any more for amplification.... Alright, given that it's a DIY project & the above cost does not include Power Transformer, chassis, volume pot, socketry, labour, etc... it's still spectacularly inexpensive for 40-50Wpc of 'hi-fi' quality (dare I say from just an initial impression, 'audiophile' quality) amplification. Firstly, low parts count - the minimalist amp circuit consists of only 4 resistors + a single 'high performance', low noise audio amplifier IC 'chip' - that's an ultra simple signal path & it's got to be 'transparent'. Next, there's no caps in the 'signal cct' - those big caps on the amp bds are the 1500uF PSU smoothing & 'decoupling' caps, one per +/- supply per ch - they are very small capacitance compared to conventional power amps & that is part of the Kimura/47 Labs philosophy - no doubt contributes to the sound heard. Ultra-short signal path - this build is a bit longer (due to input switching) than 47 Labs' miniscule 35mm or so, but still very short & the feedback cct length, for example, is only a few mm. A beefy sized PTX - this one is 300VA. 'Ultra-fast recovery' rectifier diodes - I've heard improvements from using 'ultra fast/soft recovery' & Scottky type diodes before (not uncommon nowadays), but I'm not sure why these particular types (MUR860) were chosen for this kit, but I have no doubt, given the simple design, that these contribute to the resultant 'sound'. Good parts quality - there are so few parts anyway (the 'Premium' kit upgrades a couple of the Rs to Kiwame carbon film). The Volume pot in this case, is a TKD 2CP601 - more on this later. Aside from the small PCB, point-to-point wired with 0.7-0.8mm diam. solid core, un-tinned copper. Still scratching my head....
|
Post by Owen Y on Aug 25, 2020 17:50:05 GMT 12
The PTX, sourced from Tortech in Sydney, 300VA / 2 x 0-22v, 6.82A..  
|
Post by colinf on Aug 25, 2020 18:59:58 GMT 12
The LM3875 and it’s relatives use quite a lot of negative feedback and as such the power supply rejection ratio worsens at very high frequencies. Soft recovery diodes keep excess switching noise out of the power supply, small PSU caps have higher resonance and only need small bypass caps close by the chip power supply pins. No feedback cap allows the chip to operate with its lowest distortion. The TKD pot sounds good, I use them here. Nice and smooth which complements the sound of the chip nicely. Choice of power transformer influences the sound quite a bit as the winding resonates with each pulse of the rectifier diodes’ switching. Soft recovery diodes dampen the resonance. It’s a reason I stick to valve rectifiers in valve preamps, the recovery is very soft and puts out noise that’s easily filtered out by the PSU cap, which can then be kept small-ish. Poweramps have a bit more current demand and soft recovery diodes make the power supply less soggy than a valve rectifier. Also keep RF out of it as much as possible, especially from the speaker leads. A small value inductor in parallel with a resistor, in series with the speaker lead isolates the negative feedback in the IC so it can work with low distortion. The inductor influences the sound markedly.
AMR-iFi R&D
|
Post by Owen Y on Aug 28, 2020 19:02:52 GMT 12
Someone brought to my attention John DARKO's review (~2011) of the AUDIOSECTOR (Peter Daniel) Gainclone integrated amplifier:  (Google image) " Micro-dynamics are taut and underscored: rhythmic running men painted with the most definite of brush strokes. There’s no hint of hesitance... " " ... musical coherence a strong suit.... Articulation is elegant, precise and delicately emphatic. " " It’s a solid state amplifier for tube lovers who want to retain the tonal colour of their 2A3 (or similar) but also yearn for a tighter, more disciplined sound. "
|
Post by Owen Y on Aug 29, 2020 15:32:42 GMT 12
Jeremy's build follows a similar layout:  (Spot the mistake in the heatsink )  (Underside)
|
Post by colinf on Aug 29, 2020 19:12:49 GMT 12
Very nice, good access. Make sure to earth the metal plate holding the power transformer!
AMR-iFi R&D
|
Post by Owen Y on Aug 29, 2020 20:30:50 GMT 12
Noted & passed on, thanks Colin!
|
Post by Owen Y on Aug 30, 2020 11:48:52 GMT 12
Still got some Walnut left over. No dovetails this time (learnt his lesson ), just mitred corners. (An over-reliance on glue?...I'm thinking.)  
|
