I have a customer who has some issues hooking up a 5410A McBSP to a TLV320AIC23B CODEC I have also posted this issue a few weeks ago in the Audio convertor forum - http://e2e.ti.com/support/data_converters/audio_converters/f/64/p/83170/286414.aspx#286414.
I have listed some email exchanges I have had with the customer below that give some more details. I've asked the customer to try a couple of things
1. He only waits 3us from resetting the CODEC to programming the regs - I've asked him to increase this to 1ms - it's not clear in the datasheet how long this time should be.
2. With the output changing due to temp this could indicate a timing issue on the the serial link for the data. Can you make sure you are transmitting from the DSP on the rising(or falling) edge of the serial clock and having the ADC receiving data on the falling(or rising) edge of the clock.
My hope is that you have drivers or a code base for the 5410A McBSP side or you have seen this issue before with the CODEC and have some insight. I notice there are some drivers available for the CODEC but none for the 5410A, I did get Linux based ones for the OMAP but if there are any drivers for the 5410A(6713 would probably work too) then that would help.
I have also attached the two plots the customer refernces below.
Customer description on what is happening.
I have had some time in the past day to look a little deeper into the CODEC/DSP problem that is plaguing us. I created two tone generators, sending their respective outputs directly to the left and right channels of the CODEC. I can control the level of each generator from zero to full scale. Here's what I've found:
1) I can generate full scale tones (at the CODEC output, this is ~ 2.34 volts p-p) through the left channel regardless of the output content of the right channel.
2) I can generate full scale tones through the right channel, but only if the output content of the left channel is 0. When there is any signal in the left channel, the right channel is limited to around 12 bits before spikeys appear on the CODEC output. The amount of signal in the left channel (full, half, quarter, etc.) doesn't affect the distortion level on the right channel - only that it is greater than 0. The amount of right channel distortion is proportional to the level of the right channel tone generator. I've attached a couple of scope shots of the distortion - 157spikey3.jpg at the level that the spikeys first start appearing, and 157spikey4.jpg with a full scale signal.
3) The distortion threshold appears to be temperature sensitive - one moment the right channel is clear (that is, outputting a clean sine wave [albeit limited to 12 bits] without spikeys), then it will begin to distort without having changed the level.
4) You'll notice that the distortion always appears on the top peaks of the waveform.
The data order emanating from the DSP is LEFT followed by RIGHT. As far as the right channel is concerned, it appears to operate correctly when there is no data in the left channel slot preceding it.
I'm deducing from this (as I've had suspicion all along) that I have not correctly programmed one or both of the ports (most probably the McBSP, sicne there are only three possibilities in the CODEC). As you can see in my source files, I've tried various combinations of interface settings for both the McBSP port and the CODEC registers.
My hope is that the above information might give you some idea of where I should concentrate my efforts to resolve this issue.
1) I can generate full scale tones (at the CODEC output, this is ~ 2.34 volts p-p) through the left channel regardless of the output content of the right channel.
2) I can generate full scale tones through the right channel, but only if the output content of the left channel is 0. When there is any signal in the left channel, the right channel is limited to around 12 bits before spikeys appear on the CODEC output. The amount of signal in the left channel (full, half, quarter, etc.) doesn't affect the distortion level on the right channel - only that it is greater than 0. The amount of right channel distortion is proportional to the level of the right channel tone generator. I've attached a couple of scope shots of the distortion - 157spikey3.jpg at the level that the spikeys first start appearing, and 157spikey4.jpg with a full scale signal.
3) The distortion threshold appears to be temperature sensitive - one moment the right channel is clear (that is, outputting a clean sine wave [albeit limited to 12 bits] without spikeys), then it will begin to distort without having changed the level.
4) You'll notice that the distortion always appears on the top peaks of the waveform.
The data order emanating from the DSP is LEFT followed by RIGHT. As far as the right channel is concerned, it appears to operate correctly when there is no data in the left channel slot preceding it.
I'm deducing from this (as I've had suspicion all along) that I have not correctly programmed one or both of the ports (most probably the McBSP, sicne there are only three possibilities in the CODEC). As you can see in my source files, I've tried various combinations of interface settings for both the McBSP port and the CODEC registers.
My hope is that the above information might give you some idea of where I should concentrate my efforts to resolve this issue.
Cheers
Calum
