You can find the last version of my BOM for My_Ref RevC (TP version) on my Google Folder (click the link)
I don't think it's a suitable solution for a complex circuit like a power amplifier for space reasons.
Also I'm not sure the PS doesn't inject noise on circuit...
IMHO a bad idea, sorry.
Finally, I am beginning to collect parts for at least one FE build. Despite the 20% reduction currently at PC, I cannot afford the Mundorf MLGO power supply caps (unless you all can persuade me to give up food for a few days!). I have a pair of 12000uF TSHA's on hand that I plan to use. They have slightly lower ESR and higher ripple than the 10000 typically used, so there might be minor sonic benefits. Is there any cause for concern regarding the higher inrush current, or will the transformer limit it enough that it will not affect the rectifiers or any other components?
I am cannibalizing an existing MyRef C for BG Std at C9, and I plan to fit a naked Z-foil Vishay in place of Susumu at R12. There are several other passive parts that will also transfer, as they all concur with the premium BOM. All active components will be new, along with all other SMD parts and a few polystyrene caps instead of Wima's.
I'm planning 33pF at C10 and 22pF at C34 for compensation because that's what I had in the Rev C and it sounded good to me.
Any reason to not use all MP Caddocks? Is the MK necessary for good sound? MP's use the same material and have a higher power rating at the same or slightly lower cost.
One thing that concerns me on the new board is the position of R3 blocking access to the LM3886 screw. It was already tight quarters on the Rev C boards--now it's worse. Any tips on getting a tool in there? If there is ever a revision, I'd prefer to see R3 moved slightly away from the chip and toward C1/C2.
Also, the output ground is buried in the corner near the power supply. Could a trace be run under or around the relay to put the pad on the other end of the relay? I know this is dreaming, but it doesn't hurt to mention it.
The boards look great and I'm eager to begin a new phase of experimentation with surface mount and experiencing the FE.
Peace,
Tom E
I am cannibalizing an existing MyRef C for BG Std at C9, and I plan to fit a naked Z-foil Vishay in place of Susumu at R12. There are several other passive parts that will also transfer, as they all concur with the premium BOM. All active components will be new, along with all other SMD parts and a few polystyrene caps instead of Wima's.
I'm planning 33pF at C10 and 22pF at C34 for compensation because that's what I had in the Rev C and it sounded good to me.
Any reason to not use all MP Caddocks? Is the MK necessary for good sound? MP's use the same material and have a higher power rating at the same or slightly lower cost.
One thing that concerns me on the new board is the position of R3 blocking access to the LM3886 screw. It was already tight quarters on the Rev C boards--now it's worse. Any tips on getting a tool in there? If there is ever a revision, I'd prefer to see R3 moved slightly away from the chip and toward C1/C2.
Also, the output ground is buried in the corner near the power supply. Could a trace be run under or around the relay to put the pad on the other end of the relay? I know this is dreaming, but it doesn't hurt to mention it.
The boards look great and I'm eager to begin a new phase of experimentation with surface mount and experiencing the FE.
Peace,
Tom E
Tom, the mounting screw for the lm3886 is tricky. I originally used an cap screw with an allen head. It was hard to start the threads and once it had finally started to to tighten it was slow going with an allen wrench that I cut down on the short side. My case design slides out of the heat sinks so I needed a better way that was less fiddly. I am now using a long set screw and a small brass nut. I can tighten the set screw from the outside onto the nut and then back it up to the chip. Much easier and nice and tight.
They should be fine and you shouldn't have problems with the value.
It's not necessary but it will not hurt either
If you have to buy it, the SMD version (VSMP, from Texas Components) costs less and will mount perfectly.
In my build I'm using VSMP in R12 because I've bought them for my BOM tests and they sound a bit better than Susumus.
The difference is not so big to justify the expense, IMHO, and they're a PITA to determine orientation (they come without any marking on them)
The FE compensation is a not small part of the improvement.
But Rev C compensation will allow you a more 'direct' comparison with Rev C builds.
You could use all MPs if you want but they sound different from MK132.
In R13 a MK132 is mandatory, IMHO, while in R104/R204 you can use 1/2W PRPs (use same orientation of the MK132).
The only real solution is not in the tool but in the screw.
You can use a screw with hexagonal head and use a key to tighten or thread the heasink, use a threaded insert and tighen a bolt.
It has been debated on the build thread.
The position you're suggesting (similar to TP one) would have led to a longer feedback trace.
I don't know if would have been any difference soundwise, but the fact you can use bolts led me to the actual choice.
This can change, though.
A trace, hardly, but it could be possible to rearrange protection circuitry.
Maybe for a new realease.
I've used 12000 uF Elnas in some Rev C builds, and there were no issues with inrush current.
C7, C30, C32 and maybe C12 are good candidates for polystyrene.
R37, R39, R42 and R43 are good candidates for non-magnetic resistors (though I've only tried them at R37 and R43 in a through-hole v1.4 PCB). SMD non-magnetics may be hard to find, even in thin-films like Susumu, Vishay, Koa, etc. - YMMV.
Thanks for all the tips and encouragement, guys.
Has anyone found a non-magnetic resistor to use at R1? Not many one ohm parts out there. I was considering paralleling two 2.2 PRP's.
I will forgo the Caddock's at R104/204 for cheaper PRP's. I'll keep the MK132 at R13.
Keep in mind there is no R42 in the FE. Seems like your value is getting smaller there. Per Dario, I removed it from my RevC quite some time ago. I could not really detect much difference.
Regarding the rest of compensation network, I am interested in your experiments. I don't need to worry much about deep bass because I use an active xover at 320Hz before the amp, so I'm concerned mostly about mids and highs.
Peace,
Tom E
Has anyone found a non-magnetic resistor to use at R1? Not many one ohm parts out there. I was considering paralleling two 2.2 PRP's.
I will forgo the Caddock's at R104/204 for cheaper PRP's. I'll keep the MK132 at R13.
Keep in mind there is no R42 in the FE. Seems like your value is getting smaller there. Per Dario, I removed it from my RevC quite some time ago. I could not really detect much difference.
Regarding the rest of compensation network, I am interested in your experiments. I don't need to worry much about deep bass because I use an active xover at 320Hz before the amp, so I'm concerned mostly about mids and highs.
Peace,
Tom E
If you're OK with Carbon-Comps, Allen-Bradley is available in 1 and 1.5 ohm (the exact value doesn't seem to be critical - R11 is just a ground isolation/lift resistor). You can get a bunch of 5% ABs and find a pair within 2%, maybe even 1%, without much difficulty.
I have also tried a 2R 5% Futaba (pink) and a 1.5R 5% Panasonic ERX1 MFR (green) and there was no noticeable difference, though the Panasonic has magnetic end-caps or something magnetic in the body.
Untried alternatives:
1) Dale RS-1A wirewound - not sure if it is completely non-magnetic.
2) Shallcross RWR80N mil-spec non-inductive wirewound, which is also non-magnetic, AFAIK.
Yup, same here - no detectable difference in Rev C. Keeping it may help a bit in damping out the Q for some high-Q film/foil caps at C32 - since I was using a KP1834 at C32, I decided to retain R42 around 50 ohms (actually, anything above ~10 ohms should be fine for damping out Q).
There's no much difference, simply one part less on the signal path.
The FE compensation have an impact on bass perfromance but the bigger effect is in hf sweetness and sounstage.
I've looked at TI's LM318 app notes and the chip schematic. The function of C10 and C34 remains unclear to me, but I'm accustomed to that.
Is it possible that, if these two caps function together, their values should be adjusted proportionately? Original MyRef C design was 22/10. It seems logical to me that changing one value by 50% to 33 would entail changing the other's value by 50% as well to 15. Does that make sense to anyone else?
Dario, I'm not doubting your results. I think that small adjustments in component values can sometimes result in significant sonic changes, and I want to do some research in my own, with your and Siva's guidance. Can you tell us how many different values and combinations of values you tried before settling on 33/27? I went back and read the threads that dealt with comp changes, but I didn't learn much other than that some drastic value changes were not stable or didn't sound good.
One other general question: can SMD caps be paralleled by stacking one on top of another? If I tried smaller values to begin with, could I simply solder another cap on top of it to get a larger value? It could be fun to keep stacking until you hit the sweet spot.
Peace,
Tom E
Is it possible that, if these two caps function together, their values should be adjusted proportionately? Original MyRef C design was 22/10. It seems logical to me that changing one value by 50% to 33 would entail changing the other's value by 50% as well to 15. Does that make sense to anyone else?
Dario, I'm not doubting your results. I think that small adjustments in component values can sometimes result in significant sonic changes, and I want to do some research in my own, with your and Siva's guidance. Can you tell us how many different values and combinations of values you tried before settling on 33/27? I went back and read the threads that dealt with comp changes, but I didn't learn much other than that some drastic value changes were not stable or didn't sound good.
One other general question: can SMD caps be paralleled by stacking one on top of another? If I tried smaller values to begin with, could I simply solder another cap on top of it to get a larger value? It could be fun to keep stacking until you hit the sweet spot.
Peace,
Tom E
Initially I though the same and I wrote to Mauro about the first alternate compesation.
Mauro was really upset by the approach and explained me some more about those caps.
C10 is the feedforward compensation and has the biggest part of its effect on the LM3886 transconductance bridge while C34 compensate the LM318 and has no direct and/or linear relationship with C10.
I've tried every step in every value I've been able to buy evaluating by ear and Siva simulated my findings to confirm stability and results.
Siva, after that, experimented a bit more with those values obtaing his own version of the new compensation that best suits his taste.
Yes but it's a bit fragile.
Well, that's the end of my guessing!
Thanks, Dario, for a typically exhaustive, authoritative answer. That explanation from Mauro makes perfect sense when looking at the schematic. Now I feel confident enough to build my FE according to your carefully chosen component values, with only minor variations in brands.
Bob, I won't think much about C13 until someone posts the results of their own experimentation that such-and-such a cap unequivocally surpasses TC. I will happily mount it directly on the FE board, but I don't expect much improvement from removing a couple inches of lead length.
Peace,
Tom E
Thanks, Dario, for a typically exhaustive, authoritative answer. That explanation from Mauro makes perfect sense when looking at the schematic. Now I feel confident enough to build my FE according to your carefully chosen component values, with only minor variations in brands.
Bob, I won't think much about C13 until someone posts the results of their own experimentation that such-and-such a cap unequivocally surpasses TC. I will happily mount it directly on the FE board, but I don't expect much improvement from removing a couple inches of lead length.
Peace,
Tom E
Off Topic - Amprobe AM-270 DVOM
Sometimes, you have to be careful about the tools you use. I know that a few of us on this thread have the Amprobe AM-270 digital meter. I have discovered a potential problem in my specific meter that may show up in others. Overall, I still like this meter very much, but I'm writing this so that you can be informed and careful.
What I have discovered is that, when using the millivolt scale and measuring AC voltage, a negative DC voltage seems to offset the answer by roughly 20 mV. I discovered this when measuring a power supply with plus and minus DC voltage. I flipped over to mV AC to check for any ripple and found the 20 mV on the minus DC voltage. When I flipped the probes so that the same cap was being measured as a positive voltage, the mV AC went to zero.
I brought in my old Fluke 75 for comparison. Although my old Fluke 75 is not as precise as the Amprobe, the Fluke and the Amprobe with positive DC measure the same mV AC within accuracy of the Fluke.
I hope this helps those of you with Amprobe's avoid confusing yourselves as I have.
I now return you to your regularly scheduled programming.
Sometimes, you have to be careful about the tools you use. I know that a few of us on this thread have the Amprobe AM-270 digital meter. I have discovered a potential problem in my specific meter that may show up in others. Overall, I still like this meter very much, but I'm writing this so that you can be informed and careful.
What I have discovered is that, when using the millivolt scale and measuring AC voltage, a negative DC voltage seems to offset the answer by roughly 20 mV. I discovered this when measuring a power supply with plus and minus DC voltage. I flipped over to mV AC to check for any ripple and found the 20 mV on the minus DC voltage. When I flipped the probes so that the same cap was being measured as a positive voltage, the mV AC went to zero.
I brought in my old Fluke 75 for comparison. Although my old Fluke 75 is not as precise as the Amprobe, the Fluke and the Amprobe with positive DC measure the same mV AC within accuracy of the Fluke.
I hope this helps those of you with Amprobe's avoid confusing yourselves as I have.
I now return you to your regularly scheduled programming.
Bing,
There is no easy answer to that. In the past, Dario has suggested that resistors in the signal path should be metal film and that resistors that go to ground should be carbon, at least most of the time. That is a recommendation when there is no additional information.
Dario and Tom and others spend a lot of time trying different designs and brands in each location to discover what sounds best. That is why building the MyRef FE and following Dario's BOM is a great thing. Dario has spent a lot of time and effort tuning and optimizing the amplifier for excellent sound. Since he has very good ears and very good taste, it is easy for us to follow his suggestion and get a good sounding amplifier.
From the sims, when C32 was changed from 150pF upwards to 330pF (in Rev E), C34 had to be changed *disproportionately* from 10pF to 47..100 pF to retain stability. Increasing C34 limits the local closed-loop gain of the LM318 at higher frequencies, improving stability but it also contributes to higher THD20 of the LM3886 + LM318 combo. That's what is the weakness in Rev E - the upper mids and highs become edgy (but it gives an explosive and percussive mid-bass, as well as an immersive soundstage, which is beneficial to certain kinds of music). I was able to tame the edginess in Rev E, using the LF01 or similar Class-A modules to replace the LM318, but Rev E will nevertheless have higher THD20 (than Rev C).
In Rev FE, with the improved regulator, C32 was retained at 150 pF (as in Rev C) to be able to maintain stability, and to tweak the sonics with C10 and C34 alone. C10 has smaller impact on stability, so the change from 22pF to 33pF does not drastically weaken stability, compared to Rev C. Likewise, the increase of C34 from 10pF to 22pF..27pF is fine - it contributes to stability, while increasing THD20 by a few dB.
My hunch is that we can get some of the same effect by increasing the product of R43 x C34 - so 22k and 27pF should be similar (though not necessarily identical) to 27k and 22pF. Likewise, 22k and 22pF should be approximately equivalent to 27k and 15pF. However, higher R43 may contribute to a flatter phase response of the LM318, which might (in theory) improve sonics.
That's what I'm experimenting with, and it doesn't seem to hurt sonics so far, but I need to wait for the boards to run in a little (some electrolytics were also changed) and listen a bit more.
No major problem with ceramics, though desoldering just one of them becomes tricky. But plastic films (like Panasonic ECHU, which are available in the range of ~220pF upwards) are easily damaged by excess heat, so stacking is a problem.