Can this amplifier be used with this dac?

Other Hardware Audio
1

Hey guys,

I want to add sound to my latest project, and since I'm going surface mount I figured I should upgrade to a better set of sound chips than what I've been using. I'm aiming for 2W @ 4 ohms, because Pololu has a nice little speaker which can handle that.

Previously, I was using the same dac as the audio shield, seen here:

1480×1104 26.5 KB

And I wouldn't mind continuing to use that same dac since the 8bit version is super cheap, except the opamp used with it is only capable of 0.25W @ 4 ohms:
http://www.st.com/internet/com/TECHNICAL_RESOURCES/TECHNICAL_LITERATURE/DATASHEET/CD00001188.pdf

So I found this amp, which says it can handle 3W @ 3 ohms. Which I think works out to around 2.25W @ 4 ohms:

And it looks like a really sweet amplifier. And it's got the negative feedback loop already built in. I Iearned about that while trying to figure out how to connect the + and - pins to the DAC, here:

But those + and - pins still present a problem.

It seems that there's two ways to use this chip. I can use it the same way the opamp in the audio shield is used, with what is called a single ended output. There I would just tie the - input to ground, and the + input to the dac which would output 0~5v. Or, I can use it in differential input mode, where the dac seems to need to output -5~0v on one pin and 0~5v on another.

Unfortunately it doesn't look like that dac has a differential output ability. And that's bad, because the datasheet for the amp says that differential output would allow me to put 4x as much power into my speaker. I assume that when they say it can drive a 4 ohm speaker @ 2W they meant in this mode.

So if I use this chip in single ended mode, now I'm stuck with a measly 0.5W output. Better than the wave shield, but not by much. Doubling the power only increases the output by 3db which if I understand decibels right is only around 30% louder. In contrast, the full 2W I want to output would almost double the volume.

So what's the verdict? Can that dac be used in a differential configuration or is it single ended only?

If I can't use that dac, then I'm kinda lost as to what I should use. Going through the dacs on digikey is really confusing, it's hard to tell which ones would be good for audio and which ones wouldn't. And I want one which I can communicate with over the SPI bus.

Since TI seems to have such good datasheets though, I went straight to their site and looked for audio solutions (searching for dac didn't seem to bring up many hits) and I found these dacs, which seem to be the lowest price ones they have:

They're like 3x the cost of the Maxim one though. And I don't really need 24bit fidelity. I guess it wouldn't hurt though. I was looking for an 8bit dac in the first place cause I thought I might be able to send it less data over the SPI bus but that appears not to be the case.

More importantly though, these don't seem to be geared towards differential output either. It looks like they have a left and right channel output, both +.

So what should I do? Find a different amp? Is there such a thing as an amp which takes a single ended input and has a differential output so it can drive the speaker at maximum voltage swing?

2

While looking for amplifiers which did not require a differential input to give a 2W output, I found this one:

The interesting thing is this amplifier still requires a differential input, but they show a schematic for something called a "bridged" mode, where the output is fed back into one of the differential inputs, so only one single ended input is needed from the dac.

I wonder if the other amp I found earlier could be configured in a similar manner to give the full 2W output?

3

I think you are mixing up the differential input of the amplifier with the output.
What is wrong with using the TPA6211A1 with a differential output and a single ended input. I can't see how that is going to cut down on the power.

4

Hm, I guess I did. And spent hours reading datasheets for other amps and dacs as a result. :[

So you're saying I can just wire my circuit up as in Fig. 30 and I'll get the full 2W? That's great if that's the case.

I like this amp cause it doesn't require a ton of capacitors like many of the others. Less for me to worry about, less expense, and less space, which is important cause this board needs to be tiny. The lm4991 required three caps with one being a tantalum. It also output 2.5W and I didn't know how to limit it to 2W to prevent accidentally burning up the little speakers I wanted to use.

Looking back over both datasheets, it seems like both amplifiers use this bridged setup on the load. The first one just refers to it as a BTL configuration.

I guess I only have one other question right now then...

In Fig. 30, why do they show Cs as being an electrolytic cap, when two pages later they say:

Decoupling Capacitor (CS)
For higher frequency transients, spikes, or digital hash on
the line, a good low equivalent-series-resistance
(ESR) ceramic capacitor, typically 0.1 ?F to 1 ?F,
placed as close as possible to the device VDD lead
works best. For filtering lower frequency noise
signals, a 10-?F or greater capacitor placed near the
audio power amplifier also helps, but is not required
in most applications because of the high PSRR of this
device.

?

Am I right in assuming I can use small SMT ceramic caps for all three caps in the schematic?

5

In Fig. 30, why do they show Cs as being an electrolytic cap,

Because you can't easily get a 10uF ceramic capacitor. In doesn't say it has to be a ceramic.

The lm4991 required three caps with one being a tantalum.

No it doesn't have to be a tantalum an aluminum one will do.

Am I right in assuming I can use small SMT ceramic caps for all three caps in the schematic?

10uF ceramic caps are not very good.

6

Because you can't easily get a 10uF ceramic capacitor.

?
I just checked Digikey, and they literally have like 500 million in stock, and they're less than 15 cents apiece.

10uF ceramic caps are not very good.

In what way?

I'm really squeezed for space on this board, and sound quality is not a top priority. This board is basically a toy for adults that lights up and makes sound. I want it to sound good, but it doesn't have to sound spectacular. Most of the time it will be connected to tiny speakers that won't give great sound quality anyway.

So do you think using a ceramic cap would have a very noticeable effect on the sound quality?
And do you feel that I might as well not use any cap at all if I were to choose a ceramic for this purpose?

7

Decided to do some research into what people had to say about using ceramic in audio situations.

Haven't come to any conclusions, but here's some posts I found about them:

Y5V capacitors should not be used at all. Their high cap/size ratio is more or less marketing scam. This got big attention in a couple of recent years, as soon as capacitor vendors started to publish Cap/Voltage plots. They will have ~5x smaller capacitance in real application, compared to part value.

X5R and X7R should be OK to decouple voltage rails for analog path. Peso effects would be hardly noticeable if they are used together with electrolytic caps. Vibrations should have big amplitudes at low frequencies, where el caps and voltage regulator feedback is capable of suppressing piezo effects in ceramic caps.

The curve in the ceramic cap capacitance vs voltage has a slope related to the voltage of the cap. So I think if you use 200v or 1000v ceramic caps, which are pretty common, you won't see much capacitance variation over the 1.6v signal you might get on an audio input signal.
I've violated this Iron clad rule and am not going back. I used 50V 5uf ceramic Z5U caps as input caps to the first transistor in my power amp. I like the resulting sound a lot better than 5 uf tantalum caps I took out

COG MLCC are the best for RF decoupling, small size with minimal parasitic inductance is what you want.

COG ceramics for values up to 330pF are the best part for RF filtering and are well behaved with very low temperature and voltage coefficients, no problem at all for audio.

C0G capacitors have excellent dielectric characteristics which compete with better film varieties: low distortion, low DA, low DF. X5R capacitors, and its siblings, are another matter entirely having dielectric characteristics that compete with the worst of electrolytics for Worst Capacitor Ever.

I found some of those C0G capacitors they're recommending on Digikey, but those ones aren't offered in anything over 0.47uF

8

I also found this page with some remarks from an Analog Devices engineer:

Thank you for your inquiry about capacitors in the audio path around our SigmaDSP parts. In general, there are 3 types of capacitors that will be available in the values that are appropriate as AC coupling in most signal paths: electrolytic, tantalum and ceramic. Each has strengths and weaknesses.

Electrolytic capacitors are generally the best performing for this purpose. They are very linear when biased by the Common Mode voltages present in most SigmaDSP audio paths and their low ESR (Equivalent Series Resistance) makes them attractive as an 'invisible' component when designing a circuit. They are fairly low cost as well. However, electrolytic capacitors have quite a large footprint and they will suffer from degradation of value and performance over time - especially in a high temperature environment. They literally 'dry out' and become a high pass filter in the signal path. High temperature versions are available and will hold up well over a long period of time.

Tantalum capacitors are a reasonable alternative, however they are more expensive than electrolytics and are not available in as many values. As polarized capacitors, they must also be biased, however I have not found them to be as linear in an audio path as electrolytic capacitors. Some people say they "don't sound as good," however static single-tone distortion tests might not reveal any differences. On the positive side, tantalums do not dry out and so are appropriate for a long term high temperature environment. They are generally smaller than electrolytics for the same value.

Both of the non-ceramic types will hold their capacitive value under use condition, unlike ceramics....

Ceramic capacitors are generally not recommended for use as AC coupling in audio. They are attractive because of the size, cost and low ESR. However, when most ceramic capacitors are biased, the capacitive rating can decrease by as much as 50%! The higher grade (X7R) types suffer less from this, but you should be aware that a 1 uF might be acting like a 0.68uF in the circuit. We have also found that ceramic capacitors suffer from microphonic effects: the audio passing through the part will actually physically resonate and cause distortion! I would only use ceramic capacitors in the audio path where performance is not an issue, but size and cost are your higher priority. NP0 and C0G are the best of the ceramics, however I don't think you will find these types available in appropriate values.

You will also see some references to film capacitors for use as AC coupling in audio. While they are stellar performers, they are very large in size for the rated value. I have also had a great deal of trouble with the SMD versions of these parts surviving assembly and reflow using standard automated assembly techniques.

So based on that, it seems that since size and cost are a major factor here, I might want to look at those X7R caps as a possibility, with the understanding that the audio quality may be impacted somewhat.

[edit]

Oh, and I just noticed someone linked to this paper at the bottom of that page:
http://www.edn.com/contents/images/6430345.pdf

Gonna have to read through that, looks like it has a lot of good information.

9

Remember that these people are talking about hi fi audio. That is through expensive speakers mounted correctly. I am not sure if you will notice much difference in your application.
Basically the large cap was used in decoupling in your circuit. That will not add to distortion, but might not be as high a value as you think. You could always put two in parallel if you have any decoupling related instability.