The new "My Ref" Rev C thread

Even at +/- 35V rails (higher than the recommended +/- 32V), the dissipation in each resistor is only 23*23/1k = 0.529 W. The 1W resistors originally specified are perfectly adequate, and 2W gives a huge margin of safety. The combined dissipation in R1 and R4 of ~1W is not a concern - consider that most modern PC processors dissipate 65W to 130W, and the VRM MOSFETs for these often don't have heat-sinks and are in close proximity to small electrolytics.
Maybe one of the causes for computers to last no more than a handful of years. Then they start to randomly fail, capacitors to leak, etc...
But you cannot fight against programmed obsolescence :rolleyes:

I would try to separate both resistors a couple of milimmeters from each other to give them a bit of air. But in practice, because they are 1 or 2 cm raised, the can be bent to their sides so kept separated and stay above adjacent components.

My vote goes to Andrew's nomenclature solution. Better naming 2.x if we are making changes in the schematic or the topology.

Has anybody here tried something as extreme as to regulate the LM318 with a shunt psu? What about using a shunt regulator too for the LM3886?
 
I would try to separate both resistors a couple of milimmeters from each other to give them a bit of air.
...
Has anybody here tried something as extreme as to regulate the LM318 with a shunt psu? What about using a shunt regulator too for the LM3886?

R1 and R4 can be separated a bit. The present zener regulator is a shunt regulator, albeit with source impedance in the 10s of ohms - which is not a problem, since the LM318 is driving a load in the 10s of kilo-ohms.

You don't want to shunt regulate the power rails of the LM3886 - it will destroy the Class-AB efficiency of the chipamp, and the LM3886 doesn't really need regulated rails. It has fairly high PSRR as it is, and the PSRR of the LM318 makes the combination even better.
 
If zener regulated supplies are shunts and so well regulated, why it is more wide spread using three terminal regulators? Is there a really big difference?

Zener shunt regulators are ok for low currents, maybe up to a few 10s of mA, as in the RevC. For higher currents, three-terminal series regulators are better - and they have a bunch of other advantages, including lower source impedance, better thermals (TO-220 package), more accurate output voltage, etc.

However, for audio op-amp power regulation, I tend to prefer the zener shunt regulator - its higher dynamic impedance is actually better for sonics in the form of higher 2nd harmonic distortion.
 
OK, here's the current version of the 1.3 layout, which is the release candidate #1 (rc1). If there's no show-stopping problem that's found over the next 48 hours, I will post the schematic and board files here, and also extract the Gerber and Excellon files for sending over to the local PCB fabrication outfit. I haven't yet decided on the number of PCBs for the first run, but it may be a small batch (< 50 boards).

Edit: The ground net is not fully routed near the PGND PTH - 2 small unrouted airwires are still visible. I'm leaving it as it is, and don't anticipate any problems.
 

Attachments

  • revc_1_3rc1.png
    revc_1_3rc1.png
    108.4 KB · Views: 720
Last edited:
I am down with a unknown flu/cough, which has shot the schedule off the wall. Anyway, I was able to use some of the time to do some fine-tuning of the layout as well as drill optimization (now down to 5 drill diameters only). Here's the current version 1.3 layout - which is Release Candidate 2. Some changes suggested by Quadtech haven't made it into this layout - I'll merge them in later.
 

Attachments

  • revc_1_3rc2.png
    revc_1_3rc2.png
    111.8 KB · Views: 603
OK, I'm done with the optimizations - this is the final Version 1.3 release. The changes from 1.3rc2 are as follows:

1. C1, C2 and C9 now have dual 5mm/7.5mm pitch - which is more useful than 3.5mm/5mm.
2. Orientation of C3 is now correct (it was accidentally flipped earlier).
3. The 1N4004s now have a 10mm pitch vs. 7.5mm earlier (should be easier to fit some brands with the longer plastic package).
4. The CL60 has now been brought outboard, while the safety_gnd connector has been moved inboard - should help reduce stress on the PCB.
5. The number of distinct drill diameters is now reduced to 4 (from 5 in 1.3rc2).
6. Some minor placement and track-routing fixes to increase clearances at a few places.

The attached .zip file contains the .brd and .sch files in Eagle V4.16 format. The terms of use are the same as those previously offered by Mauro and Russ - generally unrestricted for further use and modification as long as all the design credits are retained in some form. If anybody becomes a millionaire from this, they are to remember Mauro.
 

Attachments

  • revc_1_3_final.png
    revc_1_3_final.png
    112.9 KB · Views: 537
  • revc_1_3_final.zip
    44.1 KB · Views: 209
Just to continue on this thread what was not welcome on the other... ;)

The first thing I want to clarify is which components I consider definitive (and so recommended) and which experimental (and so only if you want to experiment).

Definitive and recommended, with Mouser codes:

C12 Wima FKP2 220pF 100V (505-FKP2220/100/2.5)
C10 CDE CD10 22pF (598-CD10ED220JO3F)
C34 CDE CD10 10pF 300V (598-CD10CD100JO3F)

What to expect:

Even less harshness, wider and deeper soundstage, tighter bass.
The most rewarding is C12, I consider that cap mandatory.
The SMs, which are a fine tuning, must be CD10 because CD15 are too big, particularly in C34 (see photo)

Experimental:

All carbon/MK132 shunts (R10,R11,R13)
MKS2XL film Caps in place of elcos (C1,C2 = 10 uF - C6,C11 = 4.7 uF)
C7 Wima FKP2 10nF 100V
C5 Wima MKP10 LS7.5 10nF 630V
LM318 Metal Can

How experiments are going:

The MKS2XLs seems a BIG improvement and so far they don't seems to hurt stability, any suggestion on tests to make?
The FKP2 in C7 will probably be confirmed, it opens-up sound and bass is even deeper. The temperature of the LM318 with 10nF or 100nF remains identical, could this be a good indication of stability?
The LM318 Metal probably can be avoided not such an improvement and it seems it lacks some detail.
The MKP10 in C5 seems a nice improvement but in someway it lacks bass...
All resistors test are very hard, differences, if any, are tiny...

From time to time I'll inform you, like in the preceding paragraph, but please don't take such reports as a recommendation unless I mark clearly a component as definitive and recommended, as these reports simply describes the path to the final choice and thus can change from report to report.

Obviously any feedback from people that actually tries the alternate components is useful and welcome :cool:

In this post I'm recomending:

  1. Buy/make a good power cable with quality plugs (all chinese fakes are good too) and big section cable, a chinese clone is cheap and fast. 35-50$
  2. The Furutech base inlet, it makes a difference. 9$
  3. If you want quality RCA sockets without spending too much CMC 805 are a good choice. 7-8$/pair
 
I just tried the NatSemi plastic DIP LM318N in place of the TI CERDIP LM318JG, and I can confirm that they're sonically nearly indistinguishable. The TI seems to be slightly more musical (aka brighter) in the mids, while the NatSemi is more neutral.

If you can get hold of the TI CERDIPs, I'd suggest giving them a listen. They're also probably better from a thermal dissipation standpoint. Mine were NOS with a 7810 (1978?) date-code, so it's probably hard to get hold of them now.

I'll also roll in the Linear Tech LT318 later, just to see if it actually works.

10nF FKP2 for C7 seems to be rather low-valued - anyway, I'll give it a try on the next build if I can get hold of a few.

Has anybody tried RIFA-Evox PHE840 Metallized Polypropylene X2 caps for C4, C5? I could only get 47 nF in 7.5mm, but I'll try it out on the next build.
 
Semi Final Recommendations

Definitive and recommended, with Mouser codes:

C12 Wima FKP2 220pF 100V (505-FKP2220/100/2.5)
C10 CDE CD10 22pF (598-CD10ED220JO3F)
C34 CDE CD5 10pF (598-CD5CC100JO3F)
C7 Wima FKP2 10nF 100V (505-FKP20.01/63/5)
R10 Takman REY 1/4W 390R
R11 KOA SPR2 2W 1R (660-SPRX2C1R0J)
R1, R4 KOA SPR3 3W 1K (660-SPR3CT631R102J)


What to expect:

Even less harshness, wider and deeper soundstage, tighter bass, more flesh and impact.

The most rewarding is C12, I consider that cap mandatory.


The SMs, which are a fine tuning, must be CD5 and CD10 because CD15 are too big, particularly in C34 (see photo)

R1 and R4 have been a big surprise...

I've used everytime KOA MOS3 so I was curious to compare them to the ones in kits, no game, MOS were way better!

Much more detail and soundstage!

So I've compared also with SPR3, the carbon version of MOS3 also known as Kiwames...;)

Wow, even better!

Same as MOS3 but better and more flesh and impact!

If you can't stand with carbons:

R1, R4 (Alternate Metal Film) KOA MOS3 3W 1K (660-MOS3CT631R102J)

Regarding resistors it was so difficult to discern between them...

Generally PRPs give a more live and punchy sound that I prefer to the one of Takmans, sweeter and relaxed.

The only position where I've preferred a Takman over a PRP is R10, where it gives a sweeter sound and wider and deeper soundstage but I don't consider it mandatory.

R11 same as R10 but mandatory since the one in kit is a standard metal film.

The Caddock MK132 in R13 was interesting but a bit harsh and confusing, I've preferred the PRP.

Still Experimental:

MKS2XL film Caps in place of elcos (C6,C11 = 10 uF)

In C1, C2 they were not suitable, big electrolythics here give more bass and impact, no match...

In C6, C11 the 4.7uF were too small, with 10uF is much better but I can't still decide what is better...

If someone brave would try them and give feedback it would be appreciated...;)

10nF FKP2 for C7 seems to be rather low-valued - anyway, I'll give it a try on the next build if I can get hold of a few.

You'll be surprised... :D
 

Attachments

  • MyRef_Ultimate.zip
    709.1 KB · Views: 234
Does using a CL 60 used as a safety thermistor have any impact on the sound quality of the My_Ref C circuit? That is to say, does placing a 10 ohm +/- 25% resistor between circuit and earth ground impact sound? Will this effect the all important 'virtual' ground?

TIA,

rick

Hopefully both options have NO affect on the sound!!

The difference is if you are using the "additional" connection to Chassis for safety or sound.

If sound they "should" be indistinguishable. If safety you would need a very large resistor for the same result.

EDIT: And neither will affect "virtual ground" if the amp is constructed and wired properly.
 
Definitive and recommended, with Mouser codes:

C12 Wima FKP2 220pF 100V (505-FKP2220/100/2.5)
C10 CDE CD10 22pF (598-CD10ED220JO3F)
C34 CDE CD5 10pF (598-CD5CC100JO3F)
C7 Wima FKP2 10nF 100V (505-FKP20.01/63/5)
R10 Takman REY 1/4W 390R
R11 KOA SPR2 2W 1R (660-SPRX2C1R0J)
R1, R4 KOA SPR3 3W 1K (660-SPR3CT631R102J)


R1 and R4 have been a big surprise...
If you can't stand with carbons:
R1, R4 (Alternate Metal Film) KOA MOS3 3W 1K (660-MOS3CT631R102J)

OK, the first Version 1.3 boards have been populated and powered up. No smoke, everything cool or mildly warm to the touch, output offset is 1.6 mV, and it is sonically similar to the Version 1.2 Twisted Pear board with slightly different components.

I went with (Version1.3 prototype vs. Version 1.2 Twisted Pear):

C12: 220pF/630V Wima FKP2 (red) vs. 220pF/100V Wima MKI2 (black)
C10, C34: Silver Mica on both
C7: 10nF/275V X2 Rifa PHE840 (blue) vs. 100nF/63V Wima MKS2XL (red)
R10: 390/0.25W Panasonic ERD2FCG 2% CFR vs. generic 1% metal film
R11: 1/0.5W 5% generic CFR (non-magnetic leads) vs. 1/0.25W 5% generic CFR
C9: 220uF/4V Rubycon Black Gate PK on both.
IC1: Natsemi LM318N PDIP vs. TI LM318J CERDIP
C1, C2: 330uF/50V Panasonic FC vs. 100uF/50V Nichicon MUSE KZ

First audible impressions: Both are very nearly identical - however, there is a slight metallic tinge to the upper mids in the Version 1.3 prototype. I can't blame C7 at this point before giving C9 at least 10 hours to burn-in. After that, I might have to revert to 100nF on C7 - we'll see. I'll also try rolling between the Natsemi and TI LM318.

(Picture below shows a partially-populated Version 1.3 board, minus main filter caps and a few other components).
 

Attachments

  • revc13a4_12x9.jpg
    revc13a4_12x9.jpg
    462 KB · Views: 718
Last edited: