Hi everyone,
I’m contemplating building a little class-D amp based on the TPA3125 chip. I have a question about the inductor-capacitor network used at the output. The datasheet specifies 22 uH inductors and 680 nF caps for 4 ohm speakers, and 33 uH and 220 nF for 8 ohm speakers. Can someone run me through the concepts and calculations that are involved in coming up with these numbers?
My practical motivation for asking is a follows: right now, I only own one set of speakers, and they are 6 ohm. The datasheet has no values for 6 ohm speakers, so how would one calculate the ideal values there? Further, these speakers are pretty ordinary and I’m hoping to replace them eventually. I do not know in advance whether I’ll be replacing them with 4 or 8 ohm speakers. I’m wondering which case would be preferable, an amp built with the 4 ohm components powering 8 ohm speakers, or an amp built with the 8 ohm components powering 4 ohm speakers?
I understand that the purpose of the inductors and capacitors is to act as a bandpass filter, so that the high frequency switching noise is strongly attenuated while the music signal gets through (at least, I hope that’s right, otherwise I really have no idea what’s going on). This is an LC circuit, right? I’ve skimmed through the Wikipedia article on those and think I have the gist of it, but of course there are no speakers in those examples. Where does the speaker impedance come into play? After the inductor, I guess the capacitor and the speaker present two parallel impedances to earth, correct? Do I need to compute the total impedance of those two parallel paths and make sure to keep the product of that and the inductance constant? How can I compute the center frequency of the bandpass filter for a given capacitance and speaker impedance? If know that then I can assess which values are best for building an amp that is an acceptable "all rounder" with regard to speaker impedance (if such a compromise is possible!).
I’d appreciate any light that anybody can shed on this!
I’m contemplating building a little class-D amp based on the TPA3125 chip. I have a question about the inductor-capacitor network used at the output. The datasheet specifies 22 uH inductors and 680 nF caps for 4 ohm speakers, and 33 uH and 220 nF for 8 ohm speakers. Can someone run me through the concepts and calculations that are involved in coming up with these numbers?
My practical motivation for asking is a follows: right now, I only own one set of speakers, and they are 6 ohm. The datasheet has no values for 6 ohm speakers, so how would one calculate the ideal values there? Further, these speakers are pretty ordinary and I’m hoping to replace them eventually. I do not know in advance whether I’ll be replacing them with 4 or 8 ohm speakers. I’m wondering which case would be preferable, an amp built with the 4 ohm components powering 8 ohm speakers, or an amp built with the 8 ohm components powering 4 ohm speakers?
I understand that the purpose of the inductors and capacitors is to act as a bandpass filter, so that the high frequency switching noise is strongly attenuated while the music signal gets through (at least, I hope that’s right, otherwise I really have no idea what’s going on). This is an LC circuit, right? I’ve skimmed through the Wikipedia article on those and think I have the gist of it, but of course there are no speakers in those examples. Where does the speaker impedance come into play? After the inductor, I guess the capacitor and the speaker present two parallel impedances to earth, correct? Do I need to compute the total impedance of those two parallel paths and make sure to keep the product of that and the inductance constant? How can I compute the center frequency of the bandpass filter for a given capacitance and speaker impedance? If know that then I can assess which values are best for building an amp that is an acceptable "all rounder" with regard to speaker impedance (if such a compromise is possible!).
I’d appreciate any light that anybody can shed on this!
There are no production class D amplifiers on the market that give an option. Usually 470nF and 33uH is acceptable. Some use 220nF but others don't. The reason is to allow the HF to dissipate within the L and the overkill to be removed with the C. Hypes however, have a heat issue with their chokes.
I'm sorry, can you clarify what you mean by "give an option"? Do you mean give the option to use either 4 or 8 ohm speakers, i.e. class D amps are by necessity built to only drive one particular impedance?
For my knowledge there are much more 8 ohm than 4 ohm speakers on the market. If I could choose, I would prefer 4 ohms, because you get twice the power than with 8 ohms at a given supply voltage. Concerning the output LC-Filter I would prefer the 4Ohm-values for 4/8 ohm speakers. All in all I do not think that these values are critical.
I think you're right about the relative frequency of 4 and 8 ohm speakers, so I suppose going with the recommended 8 ohm values would be the safest bet. What makes you think the values are not critical? I presume that if TI bother to put two different value sets in the datasheet, it must make some measurable difference to the specifications in said datasheet.
I've been trying to find some clues about this by looking at datasheets for other class D amps. It hasn't been terribly enlightening.
The IRS2092 specifies 22uH and 0.47uF as the only combination and does not mention any variation of this with speaker impedance. It also doesn't talk terribly much about permissible speaker impedances, but an application note seems to imply that 4 and 8 ohm are both acceptable. So here's some evidence that these values do not need to be terrible speaker-dependent.
The RT9108NL specifies exactly the same as above, and the datasheet implies that 4 and 8 ohm are both acceptable. So, more evidence that speaker impedance doesn't matter much.
The TDA7491HV has a table specifying separate values for 4, 6, 8 and 16 ohm speakers! Strangely enough, their recommendations for 8 ohm speakers exactly match TI's recommendations for the TPA3125, but recommendations for 4 ohm speakers do not match. This is evidence that L and C *do* need to be matched to the speaker impedance, *and* evidence that the matchings are not device independent, which would make this whole process of sampling many datasheets a bad idea...
The TPA3200 has separate values for 4 ohms (15uH, 1uF) and 8 ohms (33uH, 0.47uF), and the datasheet actually specifies that these both yield a cutoff frequency of 41 kHz. So at least now I am confident that this is all about tuning bandpass filters. I'm trying to reverse engineer a formula to derive this frequency from the values involved...
I suppose that ideally if one has to make a single design that will apply to multiple speakers, one should err on the side of a higher cutoff frequency for the non-optimised speaker impedance, rather than a lower one. A lower cutoff could dig into the audible range, ruining the treble response. A higher cutoff will let through more HF switching noise, but I don't *think* this is too big a problem because several other chips I looked at datasheets for actually stated that running filterless was not really a problem unless you're having EMI problems or have speakers with a very low power rating.
The IRS2092 specifies 22uH and 0.47uF as the only combination and does not mention any variation of this with speaker impedance. It also doesn't talk terribly much about permissible speaker impedances, but an application note seems to imply that 4 and 8 ohm are both acceptable. So here's some evidence that these values do not need to be terrible speaker-dependent.
The RT9108NL specifies exactly the same as above, and the datasheet implies that 4 and 8 ohm are both acceptable. So, more evidence that speaker impedance doesn't matter much.
The TDA7491HV has a table specifying separate values for 4, 6, 8 and 16 ohm speakers! Strangely enough, their recommendations for 8 ohm speakers exactly match TI's recommendations for the TPA3125, but recommendations for 4 ohm speakers do not match. This is evidence that L and C *do* need to be matched to the speaker impedance, *and* evidence that the matchings are not device independent, which would make this whole process of sampling many datasheets a bad idea...
The TPA3200 has separate values for 4 ohms (15uH, 1uF) and 8 ohms (33uH, 0.47uF), and the datasheet actually specifies that these both yield a cutoff frequency of 41 kHz. So at least now I am confident that this is all about tuning bandpass filters. I'm trying to reverse engineer a formula to derive this frequency from the values involved...
I suppose that ideally if one has to make a single design that will apply to multiple speakers, one should err on the side of a higher cutoff frequency for the non-optimised speaker impedance, rather than a lower one. A lower cutoff could dig into the audible range, ruining the treble response. A higher cutoff will let through more HF switching noise, but I don't *think* this is too big a problem because several other chips I looked at datasheets for actually stated that running filterless was not really a problem unless you're having EMI problems or have speakers with a very low power rating.
Trevor Marshall - Class D Audio Amplifier Design - TDA7498 Output filters
and in this you find this link:
Understanding output filters for Class-D amplifiers | EE Times
why do you contemplate the old tpa3125? newest generation is tpa3116 family I think.
8 ohm loudspeakers can (often) have impedance above 16 ohm for high frequencies...
and in this you find this link:
Understanding output filters for Class-D amplifiers | EE Times
why do you contemplate the old tpa3125? newest generation is tpa3116 family I think.
8 ohm loudspeakers can (often) have impedance above 16 ohm for high frequencies...
cheaper??? check ebay for diy kit tpa3122 (dip) $13 and 3110 $10 and others completed, alieexpress/taobao will show completed amp boards tpa3116 for usd $7......Sure, SMD is cheap if I buy someone else's PCB off eBay, but where's the fun in that? Then all I'm doing is soldering stuff where somebody else told me to. With through hole stuff I can buy super cheap perfboard/stripboard etc. and then lay it out myself. This means I have to study the schematics, I have to think about layout issues, how to make everything fit comfortably while keeping things that should be close close, doing proper grounding, etc. It makes the hobby interesting and challenging and educational (but also, of course, sometimes very frustrating). If I want that experience with SMD I need to get my own boards made, which can quickly cost more than all the components.
I did some spice simulations with output LC-filters.
The idea is to achieve crititcal damping of the LC-circuit by speaker resistance.
Given the speaker, the L/C ratio is given by
sqrt(L/C) = Zspeaker
Then you chose a LC resonant frequency in the range of 30..50kHz and your LC-filter is determined.
So if you double speaker impedance the L value doubles and the cap reduces to half value.
Behind this is the assumption that an 8Ohm speaker exhibits 8R real impedance at 30..50kHz. Obviously this is being far from reality, and that is the reason why LC-filters in pre-filter feedback systems can be calculated exactly for ohmic dummy loads but not for real loudspeakers.
It is just a coarse approximation imho.
The idea is to achieve crititcal damping of the LC-circuit by speaker resistance.
Given the speaker, the L/C ratio is given by
sqrt(L/C) = Zspeaker
Then you chose a LC resonant frequency in the range of 30..50kHz and your LC-filter is determined.
So if you double speaker impedance the L value doubles and the cap reduces to half value.
Behind this is the assumption that an 8Ohm speaker exhibits 8R real impedance at 30..50kHz. Obviously this is being far from reality, and that is the reason why LC-filters in pre-filter feedback systems can be calculated exactly for ohmic dummy loads but not for real loudspeakers.
It is just a coarse approximation imho.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.