Uriah, take a look at the schematic in the first post, there is a differrential pair in front of the LM318, this is not a MyRef RevC...
Clave,
You are correct.
This is a MyRef RevD Thread.
The circuit, at a quick glance, is a Howland Current Pump inside the opamp loop with the front end of the opamp adapted/modified to a discrete BJT LTP.
It is so closely based on the original Mauro circuit that Linux has used Mauro's thread title for his proposed design, rather than claim any credit for the probably unique topology in the Mauro circuit.
You are correct.
This is a MyRef RevD Thread.
The circuit, at a quick glance, is a Howland Current Pump inside the opamp loop with the front end of the opamp adapted/modified to a discrete BJT LTP.
It is so closely based on the original Mauro circuit that Linux has used Mauro's thread title for his proposed design, rather than claim any credit for the probably unique topology in the Mauro circuit.
Thanks, Dario - actually, the current version of the LF01 Class-A discrete/hybrid opamp module doesn't include the Rev D LTP stage. I might still include the Rev D LTP in a future version of the LF01, but nobody is really complaining about the noise floor of the LM318 in the MyRef, thus rendering the whole Rev D concept a bit moot.
OTOH, the Rev E (Dario's Rev FE compensation retrofitted into a Rev C) + LF01 Class-A discrete/hybrid opamp module is a *huge* improvement on the plain Rev C, so I'm concentrating on optimizing those right now. This will fit (with some caveats) on most V1.2, V1.3 and V1.4 Rev C boards, making it a relatively painless upgrade for existing Rev C boards. I'll post pic(s) of a Rev E + LF01 combo later.
Last edited:
Thanks Andrew
Oops...
I would be delighted to see it as JFET LTP, trying to follow Thornsten suggestion to get rid of C9...
Well, I'm curious to try them... for sure, though not approved by Mauro, the new compensation scheme works great.
Siva, you should try my new PS for LM318, it makes wonders...
OK Dario,
I don't care about all that amplifier stuff any more - I want to know what "Bailando Salsa en el Sietch"means. I tried Google translate and failed and I don't have the time to learn a new language.
No problem, the sentence is in spanish and it means 'Dancing salsa in to the Sietch'
What the hell is the sietch?
Simple, from Wikipedia List of Dune termininology:
Sietch – Cave warren inhabited by a Fremen tribal community; in the Fremen language, "Place of assembly in time of danger."
So the whole sentence should hint to what my nickname came from...
From two of my passions: Dune novels by Frank Herbert and salsa dancing (the Clave is both a rhythm and a percussion musical instrument typical of afro-cuban music)
This little OT satisfied your curiousity?
Last edited:
I don't see why a proper differential drive for the MyRef might be more complicated, except for the additional parts BEFORE the amp?
Forgive my innocence, but why wouldn't it work as a bridge?
What I mean is a differential driver, properly designed and using 0.1% resistors (which are now easy to find or hand select from 1% types).
The only problem I foresee is on the total power, which on a bridge should be 4 times its 8 ohm output. Dissipation should be critical and I'm not sure how the Spike protection may act.
I also do not understand the concern for the GFNB loop in the negative side of the signal. What causes the phase reversal is the previous differential stage, and it already goes in anti-phase when it's fed to the 318 and the amp.
My guess is you are thinking of the bridge National proposes on the Application Notes, which I don't think should be used for bridging purposes.
Take the "Jeff Rowland approach" on his power amp using parallel/bridged 3886s. On the first version he used a transformer and on the second a differential chip to feed the phase/antiphase signals. The amp stages are exactly the same.
Last edited:
It probably can be done with an additional phase-splitting stage before the LM318s. It might also require DC-coupling to the LM318s, so it will require some care with the offsets or a DC servo somewhere - either at the LM318 input, or at the phase-splitting stage input.
It's possible, but is the additional power really needed? It will definitely need 2 LM3886s on each side of the bridge, or 4 in total for each channel. For stereo, that will be 8 x LM3886. That's a lot of additional complexity for a topology that's already adequately powered for 8-ohm loads (Where I live, 40W is *way* more than necessary to get the neighbours to complain).
DC offsets concerns are not more complicated on a differential layout as they are on an unbalanced one. No need to have a DC servo to handle anything, which would be tricky on the MyRef. The amp already has a 220uF blocking capacitor on the feedback, and that should be enough.
Whether the extra power is needed or not, that's for every one to answer. You just need two 3886 per channel, not four. I am not paralleling two more chips. 40watts is fine for small non-hungry speakers, but probably not for 10" or 12" subs, which some might be considering using.
Attachments
Carl,
Please explain your reasoning?
The National Chipamps are already crippled with low peak current capability.
They work well with high efficiency speakers and/or with high impedance speakers.
They do not work well with low and mid efficiency speakers particularly if they are also of low to mid impedance.
Bridged for extra SPL implies using low efficiency speakers and bridged forces one to adopt an effective low impedance. This combination is exactly what National Chipamps are not good at powering. I repeat again chipamps in general are current crippled. I cannot see how you arrive at your decision.
Post52, fig10.
Can anyone confirm for me what the gain of the single ended to differential pair is?
i.e. 1Vac single ended input gives ?? Vac differential output.
I am thinking Voutdiff = Vin * [R1+R3] / R2 ????
If 2k0 were chosen for all three resistors is that a gain of 2times (+6dB)?
Since each opamp is operating at effectively unity gain do both opamps need to be unity gain stable?
Can anyone confirm for me what the gain of the single ended to differential pair is?
i.e. 1Vac single ended input gives ?? Vac differential output.
I am thinking Voutdiff = Vin * [R1+R3] / R2 ????
If 2k0 were chosen for all three resistors is that a gain of 2times (+6dB)?
Since each opamp is operating at effectively unity gain do both opamps need to be unity gain stable?
AndrewT,
First of all it's no decision what I was proposing. Just thinking out loud about what could be done with a basically well sounding chip to extract more from it.
The idea came back after reading the excellent Erno Borbely article on balanced interfaces on an old The Audio Amateur magazine.
The basic idea is, yes, to cheat National's limitations for this chip. Jeff Rowland did, using, if I am not wrong twelve 3886 in parallel/bridge combination, and it seemed to work quite well as the amps had good reviews.
From your comments, which are certainly right, I'd say paralleling the chips is not an option before bridging, but a must. You need that to handle high currents and low impedance speakers. I don't see why you should avoid such speakers, particularly woofers, which are the great thirsty there.
The second part of my idea was to use cheap eBay 3886 assembled kits to assemble this project, just adding a good dual chip, also from eBay, and a healthy supply probably with +/-25v DC.
I did propose such a project here many years ago, using the inverted configuration of the original Gainclone for the amps and an AD balanced chip for the interface. But I never built it. The idea was to clone Rowland's amp as close as possible.
Carlos
First of all it's no decision what I was proposing. Just thinking out loud about what could be done with a basically well sounding chip to extract more from it.
The idea came back after reading the excellent Erno Borbely article on balanced interfaces on an old The Audio Amateur magazine.
The basic idea is, yes, to cheat National's limitations for this chip. Jeff Rowland did, using, if I am not wrong twelve 3886 in parallel/bridge combination, and it seemed to work quite well as the amps had good reviews.
From your comments, which are certainly right, I'd say paralleling the chips is not an option before bridging, but a must. You need that to handle high currents and low impedance speakers. I don't see why you should avoid such speakers, particularly woofers, which are the great thirsty there.
The second part of my idea was to use cheap eBay 3886 assembled kits to assemble this project, just adding a good dual chip, also from eBay, and a healthy supply probably with +/-25v DC.
I did propose such a project here many years ago, using the inverted configuration of the original Gainclone for the amps and an AD balanced chip for the interface. But I never built it. The idea was to clone Rowland's amp as close as possible.
Carlos
Last edited:
yes, 3 chipamps in parallel to get good current performance into 4r0 loading.
Then bridged to get good voltage and current performance into 8r0.
For stereo you end up with 12 chipamps costing much and quite complex to get right.
Every chip had a multi-turn pot to zero offset on each output, so you had a total of 12 pots to set. I'm not sure that is too complex to achieve.
The main thing to assemble are the 3886, which I was considering getting on eBay. Only problem is all kits offered on eBay are quite certainly non-inverting, which was not what JR used on their amps.
There are kits being offered, not so cheap, with three 3886 in parallel.
What I wonder if other chips (like TDA7293) are as reliable as a 3886 or as good sounding for a project like that.
Carlos
The main thing to assemble are the 3886, which I was considering getting on eBay. Only problem is all kits offered on eBay are quite certainly non-inverting, which was not what JR used on their amps.
There are kits being offered, not so cheap, with three 3886 in parallel.
What I wonder if other chips (like TDA7293) are as reliable as a 3886 or as good sounding for a project like that.
Carlos
If you intend to bridge for 8 ohm speakers, 2-chip parallel can do it with MyRef.
A 2-chip parallel MyRef could also gracefully drive 4 ohm speakers (if the amp isn't bridged).
These examples can be done with LM4780, although pairs of LM3886 could have a better quality thermal interface.
This proposal is slightly smaller scale than the Jeff Rowland example; however, its a lot more cost effective and a lot easier. Particularly noticeable is that MyRef and nested designs override/correct the tone of LM3886 and so excessive paralleling isn't required for lavish tone since the power amp isn't contributing as much noise. Even when scaled up to parallel, that basic difference remains applicable.
A 2-chip parallel MyRef could also gracefully drive 4 ohm speakers (if the amp isn't bridged).
These examples can be done with LM4780, although pairs of LM3886 could have a better quality thermal interface.
This proposal is slightly smaller scale than the Jeff Rowland example; however, its a lot more cost effective and a lot easier. Particularly noticeable is that MyRef and nested designs override/correct the tone of LM3886 and so excessive paralleling isn't required for lavish tone since the power amp isn't contributing as much noise. Even when scaled up to parallel, that basic difference remains applicable.
I'm not sure why you'd want a china kit instead of a Decibel Dungeon point to point design. Nick and Joe have some really nice photos and documentation to make this very easy. You can be up and running faster. Its also do-able without running power all over small signal like the cheaply made costly noisy boards on ebay. No need to spend extra for that.
Other topic.
For TDA7293 and TDA7294, their design is different and so their optimal support circuit is different (and definitely not in the datasheet). That doesn't make a bigger amp. It makes a smaller amp. Its difficult to underrun TDA7293/4 and would favor lower voltage use by giving you higher resolution (greatly depending on power circuit choices). In such case, you'd have an excellent lower powered amplifier that could safely run 4 ohm speakers from a single chip without problem clipping. These can make delightful guitar amplifiers.
ST's current limiter doesn't do exacerbated clipping like LM3886, so you can easily run ST chips on much lower voltage.
But we weren't looking for a lower powered amplifier.
The thermal interface of TDA7293/4 isn't twice as big as LM3886.
Therefore, please proceed with Parallel LM3886.
Rev E + LF01
LinuxGuru
I'm very interested in what you are doing with Rev E + LF01. I'm hoping to buy a kit or two from you when you are ready. Please keep us informed on the component optimization. I sent you a PM before I had dug in to understand the details. I will await your progress with anticipation.
Lehmanhill
LinuxGuru
I'm very interested in what you are doing with Rev E + LF01. I'm hoping to buy a kit or two from you when you are ready. Please keep us informed on the component optimization. I sent you a PM before I had dug in to understand the details. I will await your progress with anticipation.
Lehmanhill
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.