I see the pattern saw the pattern ages ago. The discussion is over whether the 3886 starts to behave differently below 4Ohms (and Ohm is a measure of resistance unless you specify its an impedance, so sorry for not following your differentiation correctly.
TI spec this to 4 ohms no lower. Below 4Ohms there are likely to be some other factors than a simple current limited output stage coming into account which I still maintain would be easier to measure than pontificate, which was my initial point. The graph line at 2 Ohms is around half an ohm wide so the margin of error is significant.
TI spec this to 4 ohms no lower. Below 4Ohms there are likely to be some other factors than a simple current limited output stage coming into account which I still maintain would be easier to measure than pontificate, which was my initial point. The graph line at 2 Ohms is around half an ohm wide so the margin of error is significant.
The variation is effectively similar between a single core and the shield in balanced cable.
Possibly my fault as I suggested to him the Parallel-86 would certainly work and your calculations show a brace of 3886s working very nicely should I need to melt my ribbons.
While that very well might be the curve for a particular unbalanced interconnect circuit that he measured, no way does it even resemble the audio frequency characteristic impedance of a coax cable.
The Calrec CB radio was an electric motor driving a car ignition system with a 1m rod aerial stuck on top of the spark plug. I could only run it after hours cos when I powered it up, every meter on site (including, I'm told, the passive VOMs ) would go blip blip .... It was an all band radio. Pity TV reception in Hebden Bridge.
The hardest part of the exercise was making a mike preamp with metering and headphone amp which didn't respond to it with 100m of mike cable and 150R on the other end.
RFI is usually a Yah or Nay issue.
The most common cause is rectification by input device base emitter junctions. If the RFI is below a certain level .. no problems. If it goes above, it's certainly NOT a GnatFart if you are doing a live broadcast.
OPAx134 which Self damns with faint praise .. often gives much better results than the newer uber OPAs for 2 reasons. It's very immune to RFI and also to poor decoupling on the power rails.
There's a marginal case that's sometimes encountered when you bomb-proof a device to RFI. You lose a sense of unease in the signal.
From your comments, I'm sure both you and AndrewT are versed in this art.
My original comment was in reply to Tom's post #1922 which implied that the caps should go to pin 1. I'm sure he didn't intend to imply this, but I'm concerned that us (pseudo?) gurus don't spread anything but clear words of wisdom.
This is only useful if p1 is tied to the socket with short leads, the socket makes a good connection to the chassis, bla bla ... and all the other good stuff that you & AndrewT have suggested.
The better XLR sockets usually have a tag to connect from p1 to the socket body. If you don't want to connect p1 directly to the chassis, a 100n disc ceramic bridging p1 and this tag WITH THE SHORTEST POSSIBLE LEADS .. works well.
The better PCB XLR sockets also have facilities to do all this properly.
I'll abstain from discussing the cases when you might NOT want to connect p1 directly to chassis
________________________
Even after my last statement, I can't help saying "Evil, EVIL! Truly EVIL"AndrewT said:
________________________
Anyone have EVIDENCE of this? ACTUAL speakers, units and 'fast changing Music Signals' please.billshurv said:
I spent a lot of time in my previous life looking for this unicorn without success.
And yes. I'm aware of Otala's work (??) and also Self & Cordell on the subject but I'm after EVIDENCE that it happens in 'real life'.
________________________
But on the subject of 2R loads, It's a really BAD idea to operate LM3886 et al in the region where the output is SOA limited. (below 4R at +/-35V & below 3R at +/-28V according to Fig 33 of the TI LM3886 2013 datasheet.)
Their SPIKE protection sounds VERY nasty when triggered. You can get some relief by using smaller PSU rails & bigger heatsinks but not a lot.
The main trigger isn't Otala's unicorn zillion A currents ... but the fact that reactive loads demand more SOA from the output stage for the same current. With SPIKE, this is thermal & signal history dependent too.
Bill, I'm really interested in how you get on with your 3R ribbons. Have you commissioned the system yet? Is this with a single LM3886 Modulus?
Last edited:
If you run the Modulus-86 on ±15 V ~ ±16 V rails, you can support 2 Ω load. That should give you around 30 W out, accounting for some voltage drop across the LM3886.
If you want higher power, you need to use the Parallel-86 as a single LM3886 cannot source enough current to drive a 2 Ω load beyond 14 V peak. Also note that the voltage drop across the LM3886 is significant at high output current. If you want power into low impedance loads, your best option is the Parallel-86.
Tom
Last edited:
Curious: Where does this 150 % number come from? Do you have a reference or some math that supports this?
Tom
There used to be an Overture Calculator spreadsheet floating around. It was published by National Semiconductor. That actually took the voltage drop across the LM3886 output devices into account when calculating the output power. That voltage drop is what causes the output power to drop sharply at high currents. I think you can work your way through it with the information given in the data sheet, but the calculator was rather convenient.
Tom
So you want me to look at the imbalance in the input impedance of the two inputs of the THAT1200 to see if they can counter-balance some arbitrary interconnect? Why? So you can make one perfectly balanced amp + cable combination? How is this relevant for everybody else who will use a different interconnect cable and a different sample of the THAT1200?
Sorry dude. While I would like to help, I am only one person and I have to spend my energy where it's more likely to benefit the most people.
I do admit that I'm slightly amused to see this "requirement" of yours drift as the discussion moves on. Your initial request was incredibly vague.
Tom
Last edited:
If the load is reactive, you need to include the phase or imaginary part. The unit for impedance is ohm (Ω) regardless of it's real or complex.
Tom
I doubt that. See Figure 11 of the LMV851 App Note. Maybe the designers of the OPA134 got lucky. I'd prefer to see data before jumping to conclusions, though.
My original comment was in reply to Tom's post #1922 which implied that the caps should go to pin 1. I'm sure he didn't intend to imply this, but I'm concerned that us (pseudo?) gurus don't spread anything but clear words of wisdom.
No worries. Thanks for clarifying for everybody's benefit.
Tom
the link to an article is on this Forum.
I can't remember the authors, nor the year.
The article describes the test procedure, explains what the procedure is setting out to prove and shows a variety of test results confirming that music signals do present signals that result in very significantly increased current demand by the speaker.
They selected three very different speakers assumimg they were of severe, medium and easy reactive loads.
It turned out that all three, even though reputedly of easy to severe loading, imposed output currents of the amplifier approaching and in some very rare cases exceeding 5times what the equivalent resistor would have demanded.
That is where I was aimimg but Billshurv abandoned listening to my long-winded explanation of how to read the Fig35 graph.
I have the NS spreadsheet somewhere on an old HDD.
How would I post it here?
The TI AN doesn't identify the typical standard OPA.
The most common mechanism is rectification in the input stage. Bipolar OPA are much more susceptible than FET i/p OPAs but OPAs very widely in susceptibility.
5532 is better than loadsa BJT OPAs but OPA2134 is the best I've tried.
My 'data' is a couple of designs where simple substitution of OPA2134 for 5532 cured an RFI problem with the Calrec CB radio. It's an extreme case but some users want 'total' immunity from RFI.
Strange, whenever this comes up, no one can remember the authors bla bla either.
If anyone digs up the carcase, I'd appreciate a look at a unicorn.
In more than half my life working on speakers, I know of only one situation where a speaker will demand more current from an amp that what's suggested by the low-level impedance curve. It's a fault condition and when it happens, you'll have other stuff to worry about.
The 5534 has pins into the input pair collector loads.
There are sch showing how to convert the 5534 to jFET input.
Would this conversion make the 5534 more tolerant of RFI?
Could the jFET conversion exceed the RFI performance of the 0134?
No, I think it is essential to evaluate whether cable imbalance will cause variation in sound quality. There are lots of unknown issues which effect sound quality. I just ask for information I feel valuable. Since lots of discussion went into balanced input filtering, I think the impedance variation of the cable really shows that the more significant issues are not addressed.