Question about designing Chip amps

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hey! I was curious on how you guys design your chip amps, in ways of picking component values and putting filters, zobel/Thevinin networks as neeed. Do you guys just look at the datasheet or is there something else to it?

For example I just get confused knowning why do you need a capacitor in the feedback loop? and why use a filter in the input stage instead of just using one at the output stage?
 
Allow me to compare your question with why you put these particular ingredients in a food-dish.
The datasheet gives a first important information, through a standard schematic, about how this chip can be used. Further, the datasheet often describes in details certain details of the design and the requirements and limits for the design.
Understanding how to design the details of an amplifier requires a rather detailed overview of how amplifiers and their elements function. That can be learned by most with logic sense, patience and in particular interest in making the construction work (stubbornness).
But, it is not an overview you get in a couple of years. Most of those, on this forum, giving advise to others have decades of experience. They have also passed years of frustration with circuits that just didn't perform as intended.

Perhaps someone else can give a more precise reply?
 
Last edited:
I see what you mean, I am just so curious. for example, I want to build a LM1875 amplifier. Where do I even start? How do I even design a power supply that will get me my require wattage output? What filters will I put in? Where do I put in stabilize caps? etc etc
 
Last edited:
Can you further explain, I dont understand sorry :/

I think he doesn't understand.

You can certainly improve upon the datasheet circuits. Some data sheet circuits have rather high low frequency poles. You can improve this. You can put bigger bootstrap caps in for chips that use them. You can use premium parts. You can parallel or bridge them. You can configure them as a Howland current pump. You can use them in a two stage circuit (composite) with a premium op amp in the front end. You can compensate for their input capacitance.

Etc.

Half the people here have done at least one of these improvements.
 
Last edited:
I see what you mean, I am just so curious. for example, I want to build a LM1875 amplifier. Where do I even start? How do I even design a power supply that will get me my require wattage output? What filters will I put in? Where do I put in stabilize caps? etc etc

You seem very interested in DIY and you are clever to start with a rather simple project: the LM1875. Though simple, it has got very nice sound.

For your first projects, choose something existing and eventually a "kit" with a PCB and components you solder in yourself. Starting with your own design will probably just leave you with many problems and you will be fed up with DIY before you reach the end. When you have your kit ready (and it hopefully works), you start measuring on it and experiment with input, speakers and eventually the power supply and you gain feeling with how the circuit behaves. I guess we all started like this.
None of us (I believe) can just tell you something on this forum so you have the awakening - "aha, it has to be done like this".

We can help you choose a suitable kit to start with and help you with details in making it work. You can become one of those helping others with their problems, but have patience - it doesn't come in a moment.

Welcome to DIY.
 
Last edited:
It takes a lot of time to understand what components does what in a circuit as a whole. A very basic circuit is easier to follow than a more complex unit. Most common circuit kits follow the basic schematic presented in the data sheet of the amplifier chip..


I suggest purchasing a basic kit designed around the chosen chip. In this case, the LM1875..


Amazon.com: 25W Amplifier Module LM1875 (Kit): Home Audio & Theater


It appears to be a very simple kit using a simple single rail power supply. Dual rail power supplies (+ and -) may be a bit confusing for starters..


Information on the chip. *Note Single Supply Operation*


http://www.ti.com/lit/ds/symlink/lm1875.pdf


Build the kit. Study how it works and pique your interest in DIY!
 
Same as with other Arts and Sciences: first you start with the basic and mainly follow all the rules until you learn them and know why, they exist for a reason.

After you are fluent and comfortable with them, then you start noticing things which might be made in a different way, hopefully an improvement, and you start testing different approaches, which you confirm or discard by testing, experiment, measuring, listening.

Doing things different, just for the sake of it, *may* lead you to somewhere useful, but most often just by chance, and many times may prove frustrating.
 
I see what you mean, I am just so curious. for example, I want to build a LM1875 amplifier. Where do I even start? How do I even design a power supply that will get me my require wattage output? What filters will I put in? Where do I put in stabilize caps? etc etc

As a beginner, I'd highly suggest you to read some articles on esp website. Especially the beginner luck section and the design guidelines sections.

The power supply wiring guide is very good reading too.
 
I think he doesn't understand.

You can certainly improve upon the datasheet circuits. Some data sheet circuits have rather high low frequency poles. You can improve this. You can put bigger bootstrap caps in for chips that use them. You can use premium parts. You can parallel or bridge them. You can configure them as a Howland current pump. You can use them in a two stage circuit (composite) with a premium op amp in the front end. You can compensate for their input capacitance.

Etc.

Half the people here have done at least one of these improvements.

I really appreciate that thank you!

It looks like all you guys are converging to the same conclusion, let me give you more background on me.

I have already built a LM1875 Amplifier. The problem I was having was horrible distortion, which I believe it was ruled out being on a breadboard, and not using non to shorter leads.

I am an electrical engineer with a heavy background in control theory, so I know about pole frequencies and what not. I love this DIY community, as I think the concept of making an amplifier is beyond word I can express.

If I were able to make my own amplifier I would cry ahhah, I just want to be knowledgable as this guy here. But he leaves so many questions unanswered, for example designing components (no LPF, or HPF), why he needs a low pass filter for open loop configurations, and in order to add zeros/poles within the amplifier you need to add it in the feedback loop? I haven't worked control with amplifiers so this is a new concept I usually do it for electrical systems or other systems. When he is looking at the output power of the amplifier he just looks at the datasheet?


In the end, I just want a good place to get started, in how to design components in an amplifier, how to know which components I need in an amplifier, how to do control theory on an amplifier, how to get my wattage out of an amplifier.

Thank you all for your kind words, hopefully this fills in more blanks
 
But he leaves so many questions unanswered, for example designing components (no LPF, or HPF), why he needs a low pass filter for open loop configurations, and in order to add zeros/poles within the amplifier you need to add it in the feedback loop?

That is a composite amplifier. The 3886 is controlled by the op amp. The 3886 is dog slow compared to the op amp. This will degrade the phase margin to the point of oscillation. The low pass filter "slows down" the op amp to reduce gain to below unity at the frequencies where the phase margin is degraded. This is no different than employing an interstage capacitor (often referred to as "Cdom") in a discrete design.

Where the closed loop gain crosses the open loop gain, you want the slope to be 20 dB/decade for stability. With an op amp driving a 3886 the slope is 40 dB/decade where you need it to cross. This will be unstable. But by limiting the bandwidth of the signal from the op amp, you move the open loop response curve to a point where the intersection is at 20 dB/decade.

When you superimpose the curves to examine phase/frequency response it's called a Bode plot. If ever a picture was worth a thousand words then this is it. You can research this and learn a lot. Walt Jung (my mentor) examined this topic exhaustively and presented it in a way that is very easy to understand. "IC Op Amp Cookbook" is highly recommended, if you can find it.
 
Last edited:
Thank you for telling us about your background and your LM1875 experience. Control theory is very relevant for amplifiers and your previous LM1875 experience should be replaced by an amplifier that works.

Your first posting left the impression that you were asking how to design an amplifier in all its details and, as I have replied, all details encompass many considerations. Then you looked at one of the most elaborate LM3886 constructions available, that of Tom from Neurochrome. You regret you cannot follow Tom's considerations in details and that may be intended since that knowledge has given the possibility for Tom to finance his life for years during studies through income from Neurochrome. Most of us do not understand all the details either. Tom's LM3886DR is no starter design if you look in the details. When you understand all he writes in details, you will help us. Fast Eddie gave you some detailed information you should appreciate for a start.

I suggest you start with an LM1875 design that works. We can help you making such an amplifier work and you can apply your control knowledge when we go along. But accept for a start that all control aspects will not be available to you in the sense that those designing the LM1875 chip issued a datasheet in which you have all the information you need to make it work, but not all information about loop details they have implemented inside the IC. Amplifier IC's are sold on "need to know" basis in order not to help competitors too much.

For a start, you need a power supply. Do you have a symmetrical power supply with 15-25V (+ and -) output?

On the chip: if this project is mostly meant to teach you about making an amplifier work, you do not need to make two for stereo. The second amplifier won't teach you much new. Do you have an LM1875 chip ready?

Do you have a reasonable selection of resistors and capacitors available such that yo do not need to buy every time a component is changed or added?

What type of breadboarding are you used to work with?
 
Thank you for telling us about your background and your LM1875 experience. Control theory is very relevant for amplifiers and your previous LM1875 experience should be replaced by an amplifier that works.

Your first posting left the impression that you were asking how to design an amplifier in all its details and, as I have replied, all details encompass many considerations. Then you looked at one of the most elaborate LM3886 constructions available, that of Tom from Neurochrome. You regret you cannot follow Tom's considerations in details and that may be intended since that knowledge has given the possibility for Tom to finance his life for years during studies through income from Neurochrome. Most of us do not understand all the details either. Tom's LM3886DR is no starter design if you look in the details. When you understand all he writes in details, you will help us. Fast Eddie gave you some detailed information you should appreciate for a start.

I suggest you start with an LM1875 design that works. We can help you making such an amplifier work and you can apply your control knowledge when we go along. But accept for a start that all control aspects will not be available to you in the sense that those designing the LM1875 chip issued a datasheet in which you have all the information you need to make it work, but not all information about loop details they have implemented inside the IC. Amplifier IC's are sold on "need to know" basis in order not to help competitors too much.

For a start, you need a power supply. Do you have a symmetrical power supply with 15-25V (+ and -) output?

On the chip: if this project is mostly meant to teach you about making an amplifier work, you do not need to make two for stereo. The second amplifier won't teach you much new. Do you have an LM1875 chip ready?

Do you have a reasonable selection of resistors and capacitors available such that yo do not need to buy every time a component is changed or added?

What type of breadboarding are you used to work with?


I do appreciate it, I appreciate every comment you guys leave trust me!

I agree with your philosophy on how I should get started.

I pretty much have a lab here, and I do have a LM1875 Chip on end. I have a basic breadboard kit.

Dont have a split power supply only problem, I have a DC power supply bench withe one terminal
 
Last edited:
Interesting para-discussions. Advice, strategies, references, linkies.

But interestingly, not much about “how — in principle — to do it right”. Well perhaps I might bear upon a few receptive readers to offer principles.

(1) BREADBOARD your first attempts.

You can make all kinds of learning mistakes without really doing much damage to anything except your Ego. You'll likely learn some about the following recommendations.​

(2) TIGHT PACK your layout.

Can't over-stress the sensitivity of the feedback-which-creates-gain-and-stability loop, and how any long lines inside that feedback loop becomes pickups for RF noise, for self-oscillation signal, yada, yada, yada. The more tightly you pack your chip amp and its critical feedback/control loop components, the happier your circuit will be.​

(3) DON'T STRETCH the gain of the stage.

There might be a few “rules of thumb”, but in essence, do not attempt really high gain one-chip amplification. This is what a preamp stage is good for. Just let the chip-amp stage be something that'll raise “nominal standard input” (0.775 volts RMS) to “nominal rated output” (which might be 20 V RMS, as an example). Do the math: 20 ÷ 0.775 = 26× gain. This isn't much.​

(4) GET USED TO THE MATH that's involved.

Use your calculator. Make spreadsheets. Don't keep them! Just toss 'em. Make more. The idea of “repetition makes the mind grow keener” is actually really true. Make a bunch, make some more, and you'll be able to do the formulæ from memory!​

(5) OBTAIN CHEAP-BUT-GOOD test equipment.

Minimally, I advise having at least 3 multimeters. 2 of them, digital. 1 if you like, an older “needle meter”. Its surprising how useful they are at averaging varying input signals. You don't need terribly high accuracy you know, whilst experimenting. Ballpark is OK. Ballpark on a orders-of-magnitude kind of scale. ±3% is fine for all but the most demanding “matching required” designs.​

Also get ahold of a USB based oscilloscope module + its software. Sure: there are gazillions of absolutely great full-sized near-antique analog and digital scopes on EBay. But seriously? A $200 USB module with 2 or 4 channels is going to get everything done you want done. And its compact. No downside.​

Third, you need a standard signal source. You can use the output jack of the same computer that's running the digital scope software interface! Kind of convenient. There are hundreds of signal generator software apps for … very little money. Worth the investment? Pretty much!!! Or again, you can “EBay” yourself into a fine, fine old analog signal generator or two. They look heck-of-impressive on your test bench.​

Fourth, drop the coin on some nicer soldering and “3rd hand” tools. A Weller-style low wattage precision soldering iron is a must. A all-direction clamp is good. Even wooden clothespins are great at times.​

Fifth, drop a “relatively tiny” amount of money into plussing-up your box-of-standard-components. NOTHING is more dementing than to be at 2:30 AM, bleary eyed, and need a 2.7 kΩ resistor. And not have one. Nothing. The old “5% series” of resistors, the 20% series of capacitors, a bunch of standard diodes (1N4007 & successors) is your meal ticket. Get them. 5 to 25 each. More, especially, of the middle-range values.​

And perhaps zeroth … get a BREADBOARD. It doesn't need to be über-fancy. Just a nice 100 mm by 250 mm socket-board. It'll need a bunch of variously colored pieces of pre-cut hookup wire (necessarily mostly solid core), so get some rolls of that. I found a place that had a 10-pack of 100 foot rolls, solid core, 24 gauge. Great for breadboards.​

Lastly, your hand tools. Tiny needle-nosed pliers. Tiny wire cutters. A real, honest-to-good wire-stripper that isn't a piece of ACE hardware junk. A magnifying glass. A bunch of electrical tape.​
_______

I'm sure subsequent posters will offer more advice along this line. But it is good. You need to MAKE (sig gen) precise signals to SEE (oscilloscope) the evolution of the signal from input to output, from output back to negative feedback input. To SEE the ripple on your power supply, and the lack of ripple on the output due to good chip-amp CMRR (common-mode ripple rejection).

And don't feel back just “tearing it up” (wholesale removing everything from the breadboard) and starting afresh. Stuff happens. Don't try to build a castle using a deck of poker cards.
_______

Welcoming you to your inevitable success:
GoatGuy
 
I do appreciate it, I appreciate every comment you guys leave trust me!

I agree with your philosophy on how I should get started.

I pretty much have a lab here, and I do have a LM1875 Chip on end. I have a basic breadboard kit.

There exists a type of very popular wire boards where you enter each component in holes in a plastic board (with contacts underneath the holes) and connect the components with wire jumpers. These boards are less suited for amplifiers because the critical components are often not located close enough to the chip to avoid self-oscillation.

Do you have a symmetrical power supply?

I suggest you start making the amplifier (in mono version) described on page 2 of the LM1875 datasheet. It is named "Typical Applications".
Try to leave the leads of the components a bit long such that they can be re-used, EXCEPT for C3, C4 and R4. They need to be placed very close the the LM1875 with short leads, else you risk self-oscillation!
You must find a piece of aluminum profile or plate to fasten on the LM1875 as heatsink. Make sure it does not touch anything else because the LM1875 cooling pad is not insulated from one of the 5 pins.


I will shortly describe the function of each of the components:


R1/C1/R2 form the input circuit. C1 is called a coupling capacitor and serves to bring the AC signal (sound) to the LM1875 input, even if there is a DC difference between the input from the source and pin 1 of the LM1875. R1 makes sure C1 is charged to ground level on the input side, even without a source connected. R1 is not critical and may be 100K or so if that’s available. R2 is a bias resistor recommended by the IC manufacturer and serves to bring the IC input potential to a suitable level (here probably ground level). Together with C1, R2 determines the low frequency cut-off point for the amplifier, thus, how low frequencies the the amplifier will pass through.


C2/R3/R4 form the feedback circuit. C2 has got a high impedance compared to R3 at very low frequencies such that the amplifier will not try to amplify very low frequencies. In the audio range, the values of R3 and R4 determine the amplification. With 20K and 1K the amplification is 20K/1K=20 times. Somewhere in the datasheet it is described that the gain must be minimum 10 (20dB) for stability. Thus, you could leave R4 as 10K and the gain would be 10K/1K=10 times. If you use 4K7 as R4 the gain would be 4.7 times but then the manufacturer warn you the amplifier may oscillate. These components are always connected to what will act as the inverting input (here pin 2).


C3/C4/C7/C8 are power line decoupling capacitors. The small 100nF capacitors are very important for HF decoupling and should be placed as close to the LM1875 pins as possible to avoid self oscillation. The 100nF capacitors are often chosen as ceramic capacitors because they have good HF properties. C7 and C8 are not critical in value as long as they are minimum 100uF. 1000UF, 2200UF or even 4700uF are often used. They serve to keep the charge ready for the amplifier when the amplifier needs to reproduce a high volume signal at the output. They are buffer capacitors and of the electrolytic type.


R5 and C5 form what Neurochrome calls a Zobel network. It serves to compensate the amplifier for certain changes in the speaker impedance. On the Internet you can read more about Zobel networks. Why these component values? The manufacturer has studied the LM1875 needs carefully and concluded these values to be the best. We believe him.


That was it and you now have a first overview of the design.


One very useful rule being valid for an amplifier designed as a power Operational Amplifier (like the LM1875): The amplifier will put its output such that there is no voltage difference between the two input pins (pins 1 and 2). With your knowledge of regulation systems, you can try to make simple calculations of how the output will vary when the signal in the input (pin 1) is varied.


If you connect the components as described in this schematic, it is 99% certain you have made your first working DIY amplifier.


Let us know when you have made the circuit and we will advise you how to test it in a rather safe way.
 
Last edited:
Just a correction of my calculations of gain above. It is a non-inverting coupling so the gain will be 1+R4/R3, not just R4/R3.
With 20K, I come to 21 and with 10K, it should be 11. With 4K7, the gain should be 5.7 if not oscillating.
Sorry for the inconvenience.
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.