Hi Guys,I have significant output noise issue with TAS5631PHD2EVM. First of all I modified the TAS5518 register 0x04 with the value 0x60 to have the TAS5518 automatic mute function disabled. With this configuration I measured what is on TAS5631PHD2EVM output with no input signal and I am really shocked. This is what I measured:
I performed also A-waighted RMS noise measurement in 20k audio band using NTi XL2 Audio Analyzer and got the result around 24mV! First of all I ask me what do I'm measured wrongly??? Why my EVM generate so high noise on the output???TI declares output noise voltage < 200µV. I do not hear significant higher noise having a speaker connected to the output but I guess it is because of the noise frequency.Could someone from TI help me to solve the issue please...Best regards,Tomasz
Sorry for the delay. Let me get someone to look into this for you.
Audio Applications Engineering Manager
Dallas, TX USA
Tomasz, I have finally been able to check this. I apologize for the delay; my travel schedule has been very intense lately. But it seems the problem must be measurement error caused by higher frequency components in the out of band noise of TAS5631 outputs. It is necessary to filter those to some degree or they corrupt the reading.
I measured 170 to 190 uV A-weighted at the outputs of a TAS5631 EVM, with register 0x40 = 0x60, no auto mute. This is in line with the typical 180uV in the data sheet. My setup was standard, 50V PVDD, 4ohm loads, digital input off. I have attached the differential output signal captured with my scope, and it looks similar to what you posted. I used an AP System Two 2322 to measure the noise, with A-weight filtering, bandwidth 22Hz to 22kHz, balanced XLR inputs, auto ranging.
But there is a trick: with the EVM outputs plugged directly into the AP inputs, the AP will not provide a measurement (it simply reads zero). We normally use one of AP's LC filters for class-D measurement to filter OOB noise, but I do not have one of these in my lab. Instead I added 34kHz RC low pass filters at each input, to reduce the higher slew rate out of band noise components, and then the AP read 190 to 200 uV. (The RCs were 100ohms series with 47nF shunt to ground.)
There is some error in my measurement from rolloff in my RC filters, but I think there is also still some residual error from the OOB noise. (When I increased the RC filter rolloff to 50kHz the AP again read zero output, so I have reduced the OOB noise just enough to get a reading.) I believe on balance the measurement is reasonably accurate.
I do not expect you to go out and buy one of AP's LC filters, as they are pricey. But can you try the RC filters and see if you get a more successful result? (It is even possible to cascade a pair of RC's to get better OOB attenuation, but be careful with that, and be careful with the component values. Input resistance of the AP is 100k to ground on each side, so there will be some error if the RC filter resistance in series with the AP input is too large.)
Hi Steve,Thanks a lot for your replay.At first, your scope screen shot is slightly different to what I measured and it is also one of very interesting behavior I observed. I guess you made your measurement using resistive dummy load instead of the speaker. If you would load the amplifier output with real loudspeaker you should get similar scope readout like my one. Look on the measurement I made. First time I measured the TAS5631 amplifier I used 4 Ohm Nubert NuBox311 Monitor speaker I normally use for audio monitoring working on my desk. The scope readout I get was like that:
The second time, when I tried to reproduce the measurement, I connected 4R dummy load, like didn't think about the differences. I have got then this result:
I was really surprised why I can't reproduce the first measurement having this same results??? After some tries I connected back my monitors and the I was really surprised - the output signal of TAS5631 depends of the load connected. With the real speakers connected I've got similar scope readout like at the firs time:
The difference looks to be significant. It means the TAS5631 output behavior depends on the connected load? Do you know how to explain the issue? What is the root cause of it? When I look on the measurement with real load screen shot I can see not just the base switching frequency but also some lower frequency signals like 4-6 time lower than the switching frequency. Unfortunately I haven't any frequency analyzer being able to measure between 50kHz and 500kHz but I have a feeling there is a lot of high power, uncontrolled noise coming out of TAS5631.Tomasz
I also tried to look directly on the PWM signal and it shows huge jitter. Look on the scope measurements I made directly on the TAS5631 outputs:
It would be interesting to made an FFT analyze of the jitter to see what is going on there. I guess the jitter shows this same high frequency signal I can measure after the amplifier low pass output filter.Do you know what is causing the HF noise???I can just suspect, the HF noise is a product of digital PCM to PWM conversion (Equibit algorithm). I know the Equibit converter works with 8-bit samples and strong high order noise shaping. I guess the noise shaping algorithm used in TAS5518 can produce precise signal in audio band but real garbage in the HF band... It is just my guess - I do know know how the equibit amplifier works in deep even if I am quite familiar with the PCT/DK97/00133 Equibit description.Could you explain what is the root cause of the HF noise coming out of the TAS5631 EVM?
Steve,The issue is definitely not present in analog input class-D amplifiers like TAS5630 and the whole family. I measured in similar way the TAS5630 and other class-D amplifiers output. The TAS5630 output looks like that:
You can just see the noise produced by switching the PWM.I have on my desk also other super simple class-d amplifier which I like very much because of having very quiet output looks like that:
Next time I will try to measure the TAS5613A I really like. At the moment TAS5613A is my favorite TI amp.I would like to investigate the equibit noise performance a little bit deeper but I think I make separate thread for that.Tomasz
Tomasz, you are correct that noise shaping will produce clean signal in the audio band and artifacts out of band. Of course, that is its purpose, to push noise into regions where it will be ignored by insensitive transmitters (speakers) and receivers (human hearing). But I think it is a good trade.
I think the most important point is that filtering is necessary with almost any analyzer to reduce the impact of the OOB noise elements enough to get a proper measurement. I hope you're able to try that and succeed with it.
If you want to know the details of the noise shaping, please place the post on the Audio Converters forum. Our converter people will be able to answer the question much more quickly than I.
All content and materials on this site are provided "as is". TI and its respective suppliers and providers of content make no representations about the suitability of these materials for any purpose and disclaim all warranties and conditions with regard to these materials, including but not limited to all implied warranties and conditions of merchantability, fitness for a particular purpose, title and non-infringement of any third party intellectual property right. TI and its respective suppliers and providers of content make no representations about the suitability of these materials for any purpose and disclaim all warranties and conditions with respect to these materials. No license, either express or implied, by estoppel or otherwise, is granted by TI. Use of the information on this site may require a license from a third party, or a license from TI.
TI is a global semiconductor design and manufacturing company. Innovate with 100,000+ analog ICs andembedded processors, along with software, tools and the industry’s largest sales/support staff.