for your perusal http://www.ap.com/download/file/148
good luck trying to recreate their THD measurements.
you can bet TI does use linear PS for all its testing. if you've ever done muli-tone linearity testing you know the effects of down sampled spurs (mixed spurious inband)
(my conclusion > new class D amps need new ways to make them appear cleaner than they really are)
THD curves at shown at one frequency ( curves will be different for other frequencies ) this is an example of what 'downsampling' does. Now throw in real world music signals and switching power supplies!
good luck trying to recreate their THD measurements.
you can bet TI does use linear PS for all its testing. if you've ever done muli-tone linearity testing you know the effects of down sampled spurs (mixed spurious inband)
(my conclusion > new class D amps need new ways to make them appear cleaner than they really are)
THD curves at shown at one frequency ( curves will be different for other frequencies ) this is an example of what 'downsampling' does. Now throw in real world music signals and switching power supplies!
Attachments
Last edited:
Yes I was thinking that too, the distortion rise happens far earlier on for the 80k bandwidth measurements than it does for the 20k. Clearly something else is happening here besides just harmonic attenuation.
The distortion minima is lower for the 20kHz bandwidth as would be expected due to lower inherent noise, but the rise only happens at around 2-3k for the 20k limit rather than 100-200Hz for the 80k.
If you extrapolate from the 20k bandwidth graph what the distortion at 10kHz would look like then you end up at around 0.05% for the 75 watt plot. Coincidentally this is exactly what the 10kHz distortion level actually is when measured with 80k bandwidth.
If you look at the 1 watt figures what's odd here is that the 20k bandwidth is good till around 6.7kHz and yet it shows no rise whatsoever. The 80k one though shows a rising response from around 3kHz.
The only thing I can think that would cause this is way out of band noise or spurious components.
The AES17 filter is largely a brick wall imposed at 20k, whereas the AUX-0025 filter rolls off slowly beyond 20k and then brick walls at just above 100k.
It could quite easily be a SMPS that they use too. Unless newer bench PSUs are still built from linear transformers. There are some videos out there of TI measuring a TAS5630 I believe so you can probably snipe a look at what power supply unit they use from that.
I would have actually thought the opposite, newer class D amps are able to function will less severe filters than their predecessors. Owing to lower EMI emissions and in some cases amplifiers that do not even need an output filter, mainly because of the switching scheme used.
There is nothing inherently wrong with a SMPS if it is designed for the job and real world signals or IMD can be predicted from single tone distortion testing.
I used SMPS from "Delta Elektronika", they are great designed, silent bench supplies.
http://www.delta-elektronika.nl/upload/dts_es150.pdf
http://www.delta-elektronika.nl/upload/dts_es150.pdf
Last edited:
Actually in the TI video I was thinking of they use the SMPS they designed to work along side the amplifier.
I don't doubt SMPS for class D power stages, b/c they create their own ripple in great amounts. but I think they (TI) may not like to have power supply spurs show up on the data sheets. it's expensive to have to retake a weeks worth of data. AFAIK DC ripple and "other SMPS stuff" > 20KHz shoots straight through to the audio outputs
Last edited:
This isn't right. The TPA will provide 140 watts at 1% distortion into a 4 ohm load. At this point inductor non linearities are going to be the least of your concerns.
140 watts = 24VRMS = 6A.
Ipeak would therefore = 8.46A.
For the 10uH version the 1D17A can handle 31A DC before it sees a 10% drop in its inductance. This isn't even close to having its RMS current rating pushed. A DC current of 6 amps will cause the inductor to heat up by 40 degrees over a 25 degree ambient, which is something completely different.
For all intents and purposes the 1D10A, which can handle 7.5 amps DC, should be more than enough for this amp.
I have a handful of both the 1D10s from TPA3118 builds and the 1D17A from some TAS5630 builds, so we shall see.
I'm going to be building a 4 channel version and have the PCB already etched and half of it is soldered up, trouble is no TPA3251 stock
My numbers are the same as yours, I'm going by 36V / 4 ohms = 9A peak (definitely OK there), / sqrt(2) = 6.36A RMS. So if you amplify a sine wave that's powerful enough to bounce the amp off the rails, you'll technically exceed the 6A mark by a whopping 6%
I'll probably change my card to the 1D17A's actually, to make more room for a bigger heatsink without making the PCB larger.
I'll probably change my card to the 1D17A's actually, to make more room for a bigger heatsink without making the PCB larger.
This isn't right. The TPA will provide 140 watts at 1% distortion into a 4 ohm load. At this point inductor non linearities are going to be the least of your concerns.
140 watts = 24VRMS = 6A.
Ipeak would therefore = 8.46A.
For the 10uH version the 1D17A can handle 31A DC before it sees a 10% drop in its inductance. This isn't even close to having its RMS current rating pushed. A DC current of 6 amps will cause the inductor to heat up by 40 degrees over a 25 degree ambient, which is something completely different.
For all intents and purposes the 1D10A, which can handle 7.5 amps DC, should be more than enough for this amp.
I have a handful of both the 1D10s from TPA3118 builds and the 1D17A from some TAS5630 builds, so we shall see.
I'm going to be building a 4 channel version and have the PCB already etched and half of it is soldered up, trouble is no TPA3251 stock
Yup TI have run out of stock, both on their free samples and on their paid ones. This is at least as far as my account (ie non business, non educational) shows me.
The funny thing is if I am not logged in to the TI website and I request a sample or request to purchase then it shows me the item as being 'in stock' however as soon as I log in it shows no stock.
If I am requesting a free sample it allows a back order, however with paid stuff that doesn't seem to happen.
My design uses two chips for a 4 channel implementation but I am not going to look a gift horse in the mouth. Normally I would just wait until stock is back up but everywhere I look shows lead times that place availability around December
If you have a chip or two to spare then I will gladly accept your offer. Please PM me and I will provide you with my address.