Correct. The gain structure article should help with understanding why this doesn't boost the output signal level established by a passive pre.
No thank you. In most home audio situations 1W into 87 dB SPL/W yields levels where EU standards would mandate hearing protection if it were a workplace; I fall into the 90% who'd rather turn down the volume, dispense with the earplugs, and hence don't require rail voltages on the range being discussed. 100mW typical is rather on the high side but is consistent with post 1036 as well.
I appreciate your help in fielding the questions. Should you fellas need me, I'm only an email away. The Contact Us page on my website (neurochrome.com) is the best way to reach me.
Tom
I have 4 ch of Mod86v1. Thinking about paralleling them for "more power" as a stereo amp. Can 2 mod86 make 120W>4R like Par-86, or does that require higher PS voltage?
My speakers dip to 3.6@100hz. Would the paralleling the amps help with that?
I have not yet pushed the single Mod86s that hard yet on these new speakers, maybe the 65W>4R will be enough, but I appreciate the advice anyway.
Thanks
Rich
My speakers dip to 3.6@100hz. Would the paralleling the amps help with that?
I have not yet pushed the single Mod86s that hard yet on these new speakers, maybe the 65W>4R will be enough, but I appreciate the advice anyway.
Thanks
Rich
Putting two MOD86 boards in parallel is not recommended. Theoretically, it's possible to do with Rev. 1.0 of the circuit, but the process gets pretty involved. So I removed that option starting at Rev. 2.0.
If you want multiple chips in parallel, I strongly suggest using the Parallel-86. Putting chip amps in parallel is actually pretty difficult. In particular when you then wrap a control loop around the whole thing, creating a composite amplifier.
If you are driving the MOD86 to clipping and it still isn't loud enough, you need more power. You can try raising the supply voltage, but you may hit the current limit of the LM3886 with a 4 Ω*load if you go beyond ±28 V rails. I say "may" as a typical LM3886 can deliver 11 A but is only guaranteed to deliver 7 A.
If you're not driving your amp to clipping and you have the volume control all the way up, you need higher gain. The preferred solution here is to use a preamp with gain as this will lead to the best gain structure.
Another option is to increase the gain of the power amp. That can be accomplished by changing one resistor. For Rev. 2.0, you change R12. The default value of 2.21 kΩ gives you 20 dB of gain. 1.05 kΩ gives 26 dB; 732 Ω gives 29 dB.
For Rev. 1.0, the relevant resistor change is mentioned in the Design Documentation.
If your volume knob is at 10 o'clock and it isn't loud enough, you need to turn up the volume...
Tom
Last edited:
Most typically, 4A in (the transformer) = 4A from the chip shared between the speaker and heatsink, with up to 60% of that going to the speaker. We really won't even know about that 7A, during real usage. Linearity and thermals will come into play, well before the amperage limit, I think. Well, I certainly don't want to hear an amp that has been pressed beyond linearity capacity. That is not 7A. Usually, at near half that much, the linearity performance has fallen so that nobody is curious what it sounds like at even louder than that. That pretty much what you get with chips--the tiny outputs that happen to fit in the tiny casings, don't escape practicalities; and, instead, do have the same limiting factors as any other really tiny outputs. An excellent workaround is paralleling to improve most qualities.
That's right. Its a whole lot better. It has two more means to work. The doubled current capacity = doubled linearity, and linearity sounds fantastic (nearly the same as doubling the output device quality). The output ballast resistors involved in paralleling bjt's also happens to prevent that negative output impedance error that chips are prone to, and blocking that problem also sounds fantastic, and not obviously like a chip. So, yeah, that was good advice!
Sure, the voltage limit isn't different; so, the Parallel86 probably won't make anything louder than the Modulous86; however, the Parallel86 does have the electronic means to do louder in a much better sounding way. That would be linearity, and anyone who likes it loud, needs linearity capacity.
For louder, one needs to bridge the Parallel86. That did deserve a mention.
A bridged Parallel86, should be able to do, at most, as much output power as 4x Modulous86. However, the thermals and power are different, so I'll have to go with an actual claim of 3x, even though you might get more.
Then I ended up with a question: Where is the really high quality bridge adapter?
Those composite style amplifiers certainly do better than what I just illustrated.
Actually, the linearity is plentiful in expected size venue, at home.
Tom's amplifiers will do perfectly well in most situations.
In my previous post, I was talking about a "push for power" sort of situation, which is not likely inside a normal size house. Apparently, we're powering up some nightclubs instead?
Actually, the linearity is plentiful in expected size venue, at home.
Tom's amplifiers will do perfectly well in most situations.
In my previous post, I was talking about a "push for power" sort of situation, which is not likely inside a normal size house. Apparently, we're powering up some nightclubs instead?
Last edited:
I think some of what you're trying to say is 3.5A RMS is 7A peak and that clipping is a peak thing. Measured performance on the Modulus and Parallel is primarily limited by noise and the linearity of the APx525, not the linearity of the amps---the little kinks in the slope down to 40W you see in the THD+N graphs on the product pages are range adjustments on the APx525. The steep ascents at 38W 8 ohms and 55W 8 ohms is THD clipping.
Nightclub power is often class D amps of a few kW. Different design space indeed.
Nightclub power is often class D amps of a few kW. Different design space indeed.
OK, great. Thanks very much everyone for all the good info. I'm glad I asked to learn not to try paralleling Mod86.
Daniel, I tried Sympatico before Par86 was announced. It has a "big amp sound" that I liked, maybe that's due to the 4780 linearity benefits you mention.
I do have enough gain with +12dB tube preamp. I have 2 mod86s at 20dB and 2 at 26db. I'll crank up the Mod86 and see what happens. I do have huge heatsinks on them.
I just realized I can biamp B+MT if I change them all to the same gain.
Thanks
Daniel, I tried Sympatico before Par86 was announced. It has a "big amp sound" that I liked, maybe that's due to the 4780 linearity benefits you mention.
I do have enough gain with +12dB tube preamp. I have 2 mod86s at 20dB and 2 at 26db. I'll crank up the Mod86 and see what happens. I do have huge heatsinks on them.
I just realized I can biamp B+MT if I change them all to the same gain.
Thanks
Thanks! That seems like it.
Well, Tom's amplifiers do have balanced input and that's potentially useful for those professional installs that drive many speakers nicely (well within capacity for fidelity--like a fixed install). That's much different than overworking a few speakers (straining--like a portable rig). And, the sound is much different as well.
I wonder how much woofer amp it would take to keep up with a/each Parallel86 that is only driving midbass, midrange and smaller speakers?
Assuming a medium/small theater pack, consisting of two 91db 8 ohm midbass run parallel for 4 ohms (tames the rising peak, halves the crossover costs, moves twice the air), a wide dispersion midrange, and a tweeter that matches dispersion pattern with the midrange (so that harmonics aren't unaccompanied), then in this medium/small scale scenario, then how much woofer amp is needed, approximately (per each)?
If we also knew the venue size for that example, then we could figure that 25% as much power will work fine on a venue half that size. If instead of the 91db pro speakers, the smaller venue has 88db home speakers, then re-scaling is straightforward, half with half.
Although a coarse estimation, that could provide a pointer on watts per square foot. Then, one can measure the room size to figure how much power one did or did not need.
That could result in a handy chart.
The larger scale examples would all be "bi-amp" in some form.
Last edited:
I had mean to bi-amp with large Class D on bass and the Parallel86 on the midbass and higher frequency tasks.
Then the question is approximately how many Parallel86 are needed for accompanying 2KW of bass amplifiers?
You'll need to do the maths for the drivers, XO, and PSD in question. But probably it'll come up in the range of a couple dozen.
Oh! Right. Lowest for male voice is approximately 85hz, regular guitar at 82hz and we don't want to put that through the subwoofer, because we don't want those to sound like drums. So, the midbass (and up) amp in my example, needs approximately 2/3rds as much power as the bass amp itself. Hadn't thought of it.
Assuming a 3db bass boost setting, a 1.2KW bass amp (half its power was used by that tiny bass boost, which is very likely to be actual usage) and then 4 of Parallel86 driving 4 ohm speakers (sum is 400W). That's a little easier to look at. It is, still, sort of 600W with 400W, that 2/3rds proportion, presented a little more practically for easier discussion.
Last edited:
The LM4780 is two LM3886 dice in the same package. You can convince yourself of this by dissolving the package in hot nitric acid and examining the dice under a microscope. There's no reason why the LM4780 should have any advantage over an LM3886.
If you're using two LM3886 (or the two halves of an LM4780) in a bridge configuration, such as the Sympatico, I suggest doing the math first. You'll find that you hit either the output current limit or the thermal limit with 8 Ω load at ±28 V and 4 Ω load at ±20 V.
If you want more power than a single LM3886 can deliver, you're better off putting two channels of LM4780 in parallel.
I'm not sure if the two dice in an LM4780 are better matched than two LM3886es from the same batch, but I got much better performance (20 dB reduction in THD) by using two channels of LM4780 in parallel rather than two LM3886es is parallel.
On the output current: The LM3886 (and LM4780) is specified to 7 A minimum (across process and temperature). This is in the spec table of the data sheet, which means it's tested in production. The chips you buy will meet this spec. I'm not sure why this is even a topic of discussion. If your chips don't meet this spec under the conditions specified in the data sheet, you should take it up with TI. However, as twest points out, 7 A peak is 4.95 A RMS. Perhaps that's the source of confusion.
Tom
Last edited:
Yes Tom, that 7A minimum and 11A typical are effectively peak output current values.
Many Builders do not read datasheets and many more do not read what the message is.
For sinewave testing the minimum AC current capability becomes 7Apk/sqrt(2) >= 4.95Aac
Maximum power output is I²R for a resistive load.
i.e. 4.95² * 4r0 = 98W into a 4ohms resistive test load. But the chip will overheat quite quickly trying to do this. Typical supply rail voltage to achieve this would be 7Apk*4r0 + ~4Vdrop =~±32Vdc, or slightly higher.
Driving a resistive test load to the chips maximum capability stresses the internal parts to severe thermal stress. That's what will fail first.
Driving a reactive load is a different ballgame. Reactive load driving is much more stressful, the SOA will be overstressed first. There may be little sign of overheating until just before it blows up.
Many Builders do not read datasheets and many more do not read what the message is.
For sinewave testing the minimum AC current capability becomes 7Apk/sqrt(2) >= 4.95Aac
Maximum power output is I²R for a resistive load.
i.e. 4.95² * 4r0 = 98W into a 4ohms resistive test load. But the chip will overheat quite quickly trying to do this. Typical supply rail voltage to achieve this would be 7Apk*4r0 + ~4Vdrop =~±32Vdc, or slightly higher.
Driving a resistive test load to the chips maximum capability stresses the internal parts to severe thermal stress. That's what will fail first.
Driving a reactive load is a different ballgame. Reactive load driving is much more stressful, the SOA will be overstressed first. There may be little sign of overheating until just before it blows up.
Last edited: