I'd go for 38 year Royal Salute.😎Try the values I've PMed you, I bet a virtual beer that it measures better after.... 😉
I'll get the parts today, and work on it next week.🙂
C34, C10 and C32 are crucial for stability and sonics. There's a huge difference in the sonics of the MyRef when these are changed; Rev C to Rev E/FE is a major change in audible sonics, and it involves changes only to these three caps. Without including these, the simulation is largely a pointless kindergarten exercise of some hypothetical nested topology, and nothing to do with the real MyRef.
NatSemi never released an official SPICE model for the LM3886, and the official LM318 model is a generic model without the COMP (compensation) pin. However, there is an LM318 model which can be modified by bringing out an internal node as the COMP pin - you'll have to search a bit on these forums as well as on the web. You can substitute a generic transconductance (voltage-controlled current source) for the Howland, but that's another approximation.
I found some (somewhat obvious) stuff in my simulations:
1) Increasing C10 decreases stability (phase margin).
2) Increasing C34 increases stability as well as THD20.
There's obviously a tradeoff here that needs to be optimized for sonics. C32 is a bit of a mystery - but it is critical, along with C10, for sound-stage and mid-bass definition.
There's other stuff that won't show up in the simulation - when C10 is a silver mica, with high dissipation factor (DF) and non-linearity, the real audible sonics are better than when it is a film/foil MKP capacitor with more ideal linear capacitor properties (low DF).
BTW, THD20 is another kindergarten p*****g contest. Throw away the distortion analyzer and just listen carefully - this is mostly about soundstage, definition and imaging, none of which can be measured by a distortion analyzer. There are plenty of amps out there (e.g. JLH 1969) with high distortion that sound just great.
To cut a long story short, the simulations are mostly worthless for evaluating sonics. Use it just as a first-pass approximation to determine phase-margin/stability. The rest of the exploration is with real boards, speakers, tracks and multiple sets of ears.
This is still an ongoing exploration, but to my ears, Dario's Rev FE compensation, along with my discrete/hybrid LF01 Class-A opamp module in place of the LM318 offers the best sonics to date. They're independent, and complementary.
Last edited:
In fact they're two distinct projects.
The My_Ref Fremen Edition (My_Ref FE) is a My_Ref variant consisting in a new smd PCB with ground planes, double diode bridge and a new LM318 power supply.
The official My_Ref FE compensation is the same of the original Rev C but, in the build manual, an alternate one will be offered in addition.
This new compensation have been empirically determined by me and both tested and simulated by Siva and refined thanks to Mauro informations about the original one and on how technically badly behaved the first version of the new one.
Mauro don't approve, support nor endorse it in any way since it's been empirically determined and in fact goes against some of his project goals (as Siva wrote distortion increases and dumping factor probably is no more as good as the original).
The Rev E (temporary name) is Siva's project of a through hole My_Ref variant with double current pump, but it's better if Siva himself describe it.
Last edited:
Well, high distortion measured with a pure resistive load does not mean high distortion from the speakers, we really need to know what is causing the distortion.
If you can show data that explains why something should sound better, and in reality it does, then it takes some uncertainty out of the way. If you have done work on this, you should certainly have measured the frequency and phase response. Why not show us? It seems simple with a computer and software nowadays.
Listening tests are important, but the performance of the speakers is the biggest issue which varies impressions. It is even more hard to conclude anything when not listening in the same room. I was with some people evaluating interconnects in their listening room. The PMC speakers that they used just was not up to par in terms of detail reproduction, etc. So I could not bear the sound, and let them listen to their heart's desire.
High DF is good for places where you want to lose some energy, so I am not sure what the purpose of C10 is, if according to the original explanation, it is to compensate for output inductance, then it should not be an high DF capacitor. But additionally sometimes making a certain part more ideal will reveal problems in other parts of the system, so I have some reservations purely concluded from listening tests.
C32 was described as part of the mechanism to prevent feedback from the speakers from effecting the circuit operation (or something like that), if it's caused by the speakers, then it's a speaker design problem.
Spice models are still very limited. I build a circuit some 20 years ago that works fine, but always turned up unstable in SPICE simulations even with the exact component models.
Actually, I'm using Rev E as the unofficial name of the Rev FE alternate compensation, transplanted (with minor modifications) into a Rev C board - any of V1.2 (Twisted Pear), V1.3 or V1.4, with no track cuts. Think of Rev E as an alternate BoM for the MyRef Rev C boards, with some audible improvements to the sonics:
1. More immersive and palpable soundstage.
2. Tightly-defined mid-bass, explosive percussion.
Everything else is more or less the same as Rev C, except that it simulates with higher THD20 than Rev C - which seems to be largely irrelevant to the audible sonics.
C32 changes the reactive character of Zo (i.e. the phase response of the output impedance). I'm quoting from memory here - at C32 = 150 pF, Zo varies from being capacitive to inductive at different places in the audible range (20 Hz to 20 kHz), but is mostly capacitive in the mid-range.
As C32 it is increased, Zo turns mostly inductive in the mid-range. (I'll post the frequency response graphs for Zo in due course - it's not with me now.)
ZL of most real speakers is inductive in the audio band (excluding resonant peaks), so the the damping factor (ZL/Zo) is closer to the resistive ideal suggested by Hiraga when Zo is inductive (which happens at different frequency ranges for Rev E/Rev FE (alternate) as compared to Rev C). That may explain why the sonics of Rev E/ Rev FE (alternate) are mostly better than Rev C in the mid-bass region (tighter definition), even though THD at 20 KHz is worse than Rev C.
Last edited:
Well, I have always made sure that the final speaker design has a nice and flat impedance. For example:
This has worked very good for me.
It is not reasonable to ask the amplifier to sacrifice performance by trying to accommodate problems the speaker designer should address properly.
Besides, as long as my measurements and simulation match, I will trust the simulations as a guidance for improvements. If they do not match, then I may go into deeper simulation. So we shall see.
An externally hosted image should be here but it was not working when we last tested it.
This has worked very good for me.
It is not reasonable to ask the amplifier to sacrifice performance by trying to accommodate problems the speaker designer should address properly.
Besides, as long as my measurements and simulation match, I will trust the simulations as a guidance for improvements. If they do not match, then I may go into deeper simulation. So we shall see.
Please, could you soongsc, Dario, linuxguru and anyone else involved in these two projects tell us what speakers/drivers are being used in your evaluations ?
My drivers are Fostex FE206E in the Dallas 2 BLH cabinet. I just re-read the X-Calibre thread and have been following ClaveFremen thread and linuxguru idea of replacing LM318 with LF01 and wonder what type of speaker they use to listen.
Did I read that this amp rewards higher efficiency speakers and punishes small 2-way with low efficiency ? I have a My_Ref revC from TP but it needs repair because I was stupid. I do not understand the science or have the equipment to repair it myself so my only choices are to find someone to repair it for me or desolder all the board and buy a complete new BOM.
If I am to buy all new parts, why not take advantage of these new ideas. So I ask about speakers to see what applies to my situation.
Did I read that this amp rewards higher efficiency speakers and punishes small 2-way with low efficiency ? I have a My_Ref revC from TP but it needs repair because I was stupid. I do not understand the science or have the equipment to repair it myself so my only choices are to find someone to repair it for me or desolder all the board and buy a complete new BOM.
If I am to buy all new parts, why not take advantage of these new ideas. So I ask about speakers to see what applies to my situation.
1. Fostex FE206 in a custom bass-reflex cabinet.
2. Jamo Compact 60 2-way bookshelves (3.5" mid-bass driver and ~1.5" tweeter crossed with a 4.7uH 1st-order capacitive crossover, for evaluation of mids and highs only).
Both are shown in the grainy pic below (taken with an entry-level camera-phone, excuse...) at a local DIYer/audiophile's place. There's an additional powered sub-woofer that's used when listening for fun (not evaluating) through the Jamos. There's a pair of 47 Labs Shigaraki DACs on the 3rd shelf (barely visible) which were used in some of the evaluations.
(My audiophile friend's ginger tabby prefers the warmth of the Marantz 6002 CD player. I can't disagree...)
Attachments
Last edited:
It's louder with large monitors, preferably in a bass-reflex cabinet. With small 2-way bookshelves, the SPL is lower at similar power levels, but that's not the fault of the MyRef, but related to the physics of small drivers. Small drivers/cabinets actually give better sonics for the mids and highs, because the attack and decay are quicker and any standing waves die out quicker in a small volume (the cabinet has to be stiff and acoustically well-designed for this to be effective).
BTW, small 2-way bookshelves with a powered sub- is a viable option for the MyRef (I know that 2.1 is heresy among audiophiles, but I've got to say it).
You have FE206 too !
Please help me ? Let me send you my dead My_Ref. You repair it, keep it around for a month or two and use it to help you with your ideas. A/B listen, experiment, what ever. An extra My_Ref for you to use. But just the two boards, my transformer would be expensive to ship to India.
Then upgrade it to what sounds best with your FE206. When your done, tell me an amount to paypal to you for parts, effort & shipping it back.
Please help me ? Let me send you my dead My_Ref. You repair it, keep it around for a month or two and use it to help you with your ideas. A/B listen, experiment, what ever. An extra My_Ref for you to use. But just the two boards, my transformer would be expensive to ship to India.
Then upgrade it to what sounds best with your FE206. When your done, tell me an amount to paypal to you for parts, effort & shipping it back.
The last time I agreed to do this for a friend, his board sat around for 7 months before I got around to fixing it (a day or two) and shipping it back.
You're better off in getting it fixed in the US. Ask troystg or udailey or others on this forum for help - if they can't do it, they'll at least give some pointers.
I'll start with my 2c: Check R11, the 1-ohm lift resistor between the small-signal ground and PGND. If it's open, it's likely that merely replacing it will fix the MyRef.
I'll be happy to supply the parts for the Rev E mod, as well as the LF01 modules, after your boards are fixed.
🙁 Oh well, If I had not asked the answer would have automaticly been no.
R11 ? By open, do you mean putting a meter on either side and measuring 0 ohms when it should measure 1 ohm ? Or infinite ohms ? Sorry for my lack of understanding, the kit was well labled and I simply followed pictures, it was like "paint by numbers".
a little history. The amp was working and driving 5.25 sealed 2-way so it needed a pre. I bought and assembled a TP Kookabura but hooked up the transformer wrong and over-volted the pre while it was hooked up to the My_Ref, killing both.
R11 ? By open, do you mean putting a meter on either side and measuring 0 ohms when it should measure 1 ohm ? Or infinite ohms ? Sorry for my lack of understanding, the kit was well labled and I simply followed pictures, it was like "paint by numbers".
a little history. The amp was working and driving 5.25 sealed 2-way so it needed a pre. I bought and assembled a TP Kookabura but hooked up the transformer wrong and over-volted the pre while it was hooked up to the My_Ref, killing both.
Last edited:
Squalor, buy two new boards from Linuxguru. I believe he is the only person currently selling new boards. They'll be slightly different from the TP boards, but you can use all the original passive parts (someone correct me if this is wrong) that are still good in your blown up boards, or make a small investment in a few better quality critical parts. I'm not sure what you mean by "over-volted", but I suspect that you destroyed one or both of the chips, so you'll need to buy those. Even if they weren't both destroyed, it's very difficult to remove them without causing some damage.
Once you get your amps up and running again, you can worry about the additional refinements potentially offered by Linuxguru's and Dario's upgrades. The MyRef C built with high quality components is a superb amp.
Peace,
Tom E
Once you get your amps up and running again, you can worry about the additional refinements potentially offered by Linuxguru's and Dario's upgrades. The MyRef C built with high quality components is a superb amp.
Peace,
Tom E
With the Kookaburra I bought the 15+15 trafo that TP had for it. I had a picture of somebody's preamp that had a simular trafo. As best I can remember (it's been 4 years) there were 4 AC mains wires, I used one pair to AC. There were 4 wires for the imput to the pre, I took + from the first pair and - from the second pair to hook up and tied the middle two together. It didn't go "BRZZZ and then stink", it just stopped working. I don't remember hearing the relay; wasn't that supose to protect it from too much DC ?
After the death I emailed BrianDonigan and explained what happened and he said something like, "Oh, you sent it too much. Too bad for you". Sad thing is I probably didn't even need the Kookaburra, just a good pot like a Noble or one of those kewl powered/ remote ones.
I liked what little I heard. me wifie keeps saying,"Just go to BestBuy and get a Sherwood, it's only $100" but I know the My_Ref is what I want. People tell my with my Fostex BLH I should get a 300B SET but I can't begin to afford the tube game. Russ White said in post# 2906 of this thread " In the RevC the LM3886 is a current pump in the feedback loop of a single ended opamp. The current pump configuration is a big reason for its unique sound. RevC tends to sound "mellow" or warm especially with low impedance speakers because of it low damping factor. I have some transmission line speakers which I think the low damping factor of the RevC benefits. I use the amp for those speakers all the time". The guy who designed the Dallas II, Ron Clarke said these cabinets blend the action of a horn with the action of a TL.
I've boogered up this thread too much with my problems. When I've got a few chores finished and have my boards and trafo on my desk and ready to start metering things I will start a new thread. I hope you and others can help me find what I can salvage.
Infinite ohms - low-valued film-resistors tend to fuse open (high resistance) rather than short, though it's not always the case. If it's open, the odds are about 50:50 that replacing it will revive the MyRef (it could still have blown other parts as well).