If you take the time to understand the circuit you'll see selecting it would lead to higher distortion, not lower. Unfortunately here on DIY Audio it's common for folks to assume the performance of a composite amplifier is equal to that of the control amp. It happens this is rather dramatically incorrect. If you're wanting to reason about changes to the circuit familiarity with the loop gain and closure concepts discussed earlier in the thread is required to do so meaningfully. I recommend starting with section 6 of chapter 6 of Walt Jung's Op Amp Handbook, which you can download for free from Analog Device's website.
Incorrect over the majority of the operating frequency range. Refer to page 12 of the 49710 datasheet and figure 14 of the 1611 datasheet. And, again, you'll be in a better design position if you take the time to reason about what the specifications say about how the device interacts with the rest of the circuit.
Correct. But can you state why the OPA1611's ability to go to 15V rail to 15V rail up to 300kHz would yield better performance than the 49710's ability to do the same up to "only" 220kHz? If you can't that's a hint slew rate isn't the performance limitation of the system.
Also correct. Have a look at what that means for loop gain.
In what ways is it better in this circuit? In what ways is it worse? Why do its advantages outweigh its disadvantages? Please do be specific; as someone who's designed and built several composite amps and would chose the 49710 over the 1611 I'm curious if you're seeing anything I've missed. And hoping there's more analysis behind the quoted statement than reading through the datasheet and making a guess.
Yup; why I checked twice before posting that.
As usual, it pays to read the whole datasheet and not just the front page. The devil is always in the details. I have never built a composite amp but this one has made me want to try. Hasn't Tom hinted that the LM3886 is the inner loop? If so, the OPA1611 has less symbiosis with that chip because of the inferior HF PSRR performance. Tom says he used the 80KHz point in his open loop gain error correction so unless he decided to redesign the circuit for a higher reference point, there would be no net benefit. There are a few ways to connect the controlling amp, does he use the negative input on the LM3886? Do I even understand what I'm talking about? 
I think not just yet. But you're asking the right kind of questions. Carry on, mate.Do I even understand what I'm talking about?
Tom may prefer you start a separate thread for your own composite amp design, however.
Last edited:
Yep. Read the entire data sheet. Do some math. Run some sims. Then decide which op-amp is the one to go with.
That's public information, actually. Have a look at the block diagram in Post #1.
I do? Where? What matters most is the performance in the audio band. That's why I optimized the DC servo to minimize its impact on the LF THD performance and optimized for the highest possible loop gain at 20 kHz, thus, optimizing the HF THD performance.
I think you may have picked up the 80 kHz from the THD+N measurements. In the THD+N vs frequency sweeps, I used 80 kHz bandwidth on the Audio Precision SYS-2712 to get at least four harmonics at 20 kHz. That was done to show more of the THD component and less of the +N in the plots.
My suggestion would be to look at the references I list in Post #1. Walt Jung's Op Amp Applications Handbook is a good one as well. The paperback version is $60-70 on Amazon. As Twest points out, you can download the entire book from ADI's website if you prefer pixels over dead trees. Once you get a little further in your learning, start a new thread. I'll be happy to help out.
Should you get stuck, toss me a line. A little published fact is that I do a little consulting/tutoring for the DIY crowd in addition to my circuit design and board sales. I'll he happy to give help you out on a one-on-one basis should you need it. My rates are quite reasonable...
~Tom
Whoops! Yeah, in my enthusiasm I kind of forgot that this is his party. Sorry Tom. I'll move my meanderings to one of your other threads
No worries. A new "xxx opamp + LM3886 composite design" thread is probably better. Feel free to let people know you were inspired by my build here...
~Tom
I think I saw a thread were they were discussing how to correlate subjective impression to measurements. I can't find it anymore
Interesting discussions but too much food fight so I stay out. Anyway, three main impressions are pretty hard to confuse because they separate type: mental imaging, sound fidelity, sound timing. If someone make a listen to this amp, maybe make assessment with those descriptions. John Curl describe sonic deterioration when too many vertical stage for amplification. Halcro good example with ~5 amplification stages. Some like very much antiseptic sound, others feel like amp make everything sound too skinny (ie: sound fidelity). LM3886 have maybe two and half vertical stage by itself, maybe Tom can say if LME49710 add full vertical stage, none because stage operation is parallel or some other description.
The craziness of the audio world is expressed here - the listening tests obsess about flatness of FR and room resonances, influences, etc. But, the designer obsesses about reducing various types of electronics distortion, to microscopic proportions. And, the twain do never meet ...
There is a way out of this "madness": learn how to hear bad behaviour, caused by various types of non-linear distortion, occur in in a complete system, and then one starts to get a handle on how to sort things out better. Oh, no, say some, that would be dreadful!! - forever more, I would keep hearing that distortion, and I won't be able to "just enjoy the music" any more! Well, that's the dilemma: do I keep driving around in a car that can't handle poorer roads, because I'm comfortable with that; or spend some effort to improve the vehicle so that it can handle all roads?
Measurement is good - but, learn to measure with your ears, as well as test gear ...
Do you mean this thread or this one? Tom, you pinned it. I read the 80KHz number and assumed it was related to the feedback. Since the PSRR is so strong, it would seem this design is fairly immune to any kind of power supply crud. What made you settle on the 22mF per channel? Have you tested smaller caps with a higher volt rating? At 8A instantaneous with 30V, it seems like even 6mF could discharge enough to do the job. By the way, will an smps provide power faster than a standard linear supply on the modulus-86?
Last edited:
That's what I had available. They're $3-4 at Mouser, so shouldn't blow the budget for most builders.
I'm fairly sure the electron velocity doesn't depend on the type of power supply chosen.
A regulated supply (SMPS or linear) will usually lower output impedance than a non-regulated supply. This means the supply voltages will droop less under heavy load with the regulated supply. Hence, for a regulated supply, the amp will still run on ±28 V rails even at peak output power, whereas with the linear supply, the supply will have drooped to ±26 V (measured using the 2x22000 uF on my amp). This in turn means you'll get slightly higher peak output power with the regulated supply.
The most obvious benefit of a regulated supply is that you get rid of most of the power supply crud. The Modulus-86 is largely immune to the supply crud, but most other amplifiers are not.
The other advantage of an SMPS is weight and size...
I actually have an SMPS300RE (±30 V version) on the way from Connex Electronics right now. I'm looking forward to testing the amp with it. I will report results. Including shipping to the US, the cost came to about $85. An unregulated supply would set you back about $70-75...
~Tom
On the power supply side. It looks like you have four diodes on the board plus the big filter caps. Is this for use with a center-tapped secondary transformer rather than the dual-secondary-with-two bridges that you frequently see for bipolar power amps?
I'm also curious what the maximum input AC voltage voltage would be? Fairly conservative, as I have no desire to be at the mercy of our already-high household voltage (125) creeping higher.
thanks,
Skip Pack
I'm also curious what the maximum input AC voltage voltage would be? Fairly conservative, as I have no desire to be at the mercy of our already-high household voltage (125) creeping higher.
thanks,
Skip Pack
The Modulus-86 board takes DC in. The rectifier and supply caps are external to the board. The board itself contains 2x1000 uF bypass cap where the power enters the board as well as optimized bypassing by the LM3886. Post #125 shows the supply schematic.
Note that a dual secondary transformer can be turned into a center tapped one by connecting the two secondary windings to form a center tap. I've shown this in Post #125. I see no reason to use two rectifier bridges...
The diodes you're referring to are probably the protection diodes for the low noise regulators used to power the opamps.
~Tom
Note that a dual secondary transformer can be turned into a center tapped one by connecting the two secondary windings to form a center tap. I've shown this in Post #125. I see no reason to use two rectifier bridges...
The diodes you're referring to are probably the protection diodes for the low noise regulators used to power the opamps.
~Tom
Last edited:
Basically, yes. Note, however, that the output impedance is generally measured as a small-signal parameter. In case of a power supply supplying a class AB amp, such as the Modulus-86, the energy storage in the supply caps matters as well.
That depends on your definition of happiness. If the amp is delivering 20 Hz at full output power, there will be significant droop on a supply with only 6000 uF of reservoir cap. There will be less droop on a supply with 22000 uF supply cap. Hence, the amp will be able to deliver higher output power with the 22000 uF caps than with the 6000 uF caps. In terms of perceived loudness, there's probably little difference between the two, but you will certainly be able to measure a difference in output power. You will also be able to measure an increase in THD at LF at lower output powers with the smaller supply caps.
If you make some assumptions about the output current (say, it's constant at the peak value), you can perform a back-of-the-envelope estimate of the ripple voltage, hence, supply voltage droop.
C = Q/V = i*t/V <--> V = i*t/C. t = 1/(2*f). i = V/R.
Let's assume ±28 V as the output voltage. 8 Ω load. i = 28/8 = 3.5.
V = i*t/C = i/(2*f*C)
Ripple voltage @ 6 mF: V = 3.5/(2*60*6E-3) = 4.86 V.
Ripple voltage @ 22 mF: V = 3.5/(2*60*22E-3) = 1.33 V.
So if the supply delivers ±28 V at idle, it would droop to about 23.1 V with 6 mF and 26.7 V with 22 mF. Assuming 1.6 V drop across the LM3886, the max output swing would be 21.5 V with 6 mF of supply cap and 25.1 V with 22 mF supply cap. The max RMS output powers would be:
6 mF: 21.5^2/(2*8) = 28.9 W
22 mF: 25.1^2/(2*8) = 39.4 W
So about 10 W difference. That sounds like a lot, but it's actually only 1.3 dB. That's a barely perceptible difference (the auditory change threshold is 1.0 dB).
I will openly admit that I made a lot of assumptions in above math. That's why I called it a back-of-the-envelope estimate. If you want to get closer to reality, I suggest setting up a simulation.
The Modulus-86 will perform to spec within the range of ±20 V to ±42 V. Higher supply voltage --> higher max output power --> higher power dissipation --> bigger heat sink needed. The sweet spot is ±24 V to ±28 V if you want to drive both 8 Ω and 4 Ω loads. You can get up to ±35 V if only 8 Ω loads are to be driven. The upper limit is set by the thermal management and the LM3886. The lower limit is set by the drop-out voltage of the low-noise regulators used to power the opamps in the Modulus-86.
~Tom
Last edited:
Need advice on grounding
Hello Tom,
I need your advice on grounding. I hope that you will spare some time from your busy schedule to help me out.
Kindly refer to fig. 4. in the article on Composite amplifiers in Radio Electronics.
If I were to build such an amplifier:
If I have separate transformers, one that generates +/-15V that will power the pre-amp section and the op-amps in the composite amplifier and other transformer that will give me say +/-30V for the output chip amp, where will the 'TIE' for the grounds be best located?
Will it be correct to have the return traces of all the points except the loud speaker and the zobel network return to the star point of the +/-15V supply since the components are drawing current from the +/-15V rail.
Your thoughts on this will be greatly appreciated.
Thanks in advance and regards!
Hello Tom,
I need your advice on grounding. I hope that you will spare some time from your busy schedule to help me out.
Kindly refer to fig. 4. in the article on Composite amplifiers in Radio Electronics.
If I were to build such an amplifier:
If I have separate transformers, one that generates +/-15V that will power the pre-amp section and the op-amps in the composite amplifier and other transformer that will give me say +/-30V for the output chip amp, where will the 'TIE' for the grounds be best located?
Will it be correct to have the return traces of all the points except the loud speaker and the zobel network return to the star point of the +/-15V supply since the components are drawing current from the +/-15V rail.
Your thoughts on this will be greatly appreciated.
Thanks in advance and regards!
I suspect he uses only one supply rail for his boards but I think I understand the motivation for what you're doing. Are you planning on having the pre corrected by the precision op-amp? Tom, are you using a voltage divider to split the rail for the precision op-amp? Modeling the cap behavior requires accurate measures of the cap and I don't have the equipment to do this. Which caps do you feel meet their specs the best. I know you mentioned some on your website but a short list here would be helpful.
Last edited:
My pre will be a SigmaDSP processor that can feed a composite amplifier as shown in the referenced article.
My pre already has +/-15V from which other supplies for the DSP are derived.
If I were to use the same supply from the pre-amp to power the op-amps that has the chip-amp nested in its loop, then I feel that the return lines need to be referenced to the supply it is drawing the current from.
This leaves only the speaker, the de-caps, bulk-caps, the reservoir supply caps and the zobel network that are associated with the currents from the higher voltage supply.
In such a case, do we need to TIE the two grounds at all? If yes, should the tie be near the +/-15V supply source?
If you start a separate thread on the topic, I will be happy to help you with your project to the extent that I have time. Should you prefer to get my help on a 1-on-1 basis with a tighter schedule, I will be happy to be your tutor/consultant. My rates are quite reasonable.
Eventually, I will add this grounding information to my Taming the LM3886 page.
My best suggestion at this point, is that you set up a simulation where you model the ground impedances. Inject an error current into the system and see where you have the quietest ground. That's your best ground reference. This is exactly the methodology I used for the Modulus-86 (now back on topic...
)Is the plus-minus not making it through? I've been saying ±20 to ±42 V with ±24 to ±28 V being the sweet spot. That's two rails. One +28 V and the other -28 V. The THAT1200, the OPA277, and the LME49710 are powered from on-board, low-noise regulators creating a ±15 V supply.
In general, I only buy components that come with a data sheet. If I can look up the component on the manufacturer's website and get a data sheet that shows the ESR (or loss tangent) and ESL (or SRF) I'm interested. If the data sheet shows the impedance of the capacitor versus frequency, I'm even more interested. I've been quite pleased with the Nichicon electrolytics for low ESR. The ones by CDE and Panasonic are good too. For film caps, Panasonic and EPCOS are my go-to manufacturers. For ceramic caps, Murata, TDK, and Kemet tend to be well-documented.
The only boutique brand of capacitors I recommend is Solen. I recommend Solen because they're reasonably priced and meet my criteria listed above. Solen actually publishes data for these parts (concept!).
~Tom