signal attenuation through EVAL-ADAU1452REVBZ analogue connectors

Hello Mark,

The EVAL-ADAU1452REVBZ has a full-scale analog input level of 1V RMS. The full-scale analog output level is also 1V RMS. I expect your test signal is clipping the ADC; this would look like attenuation on a level meter.

I usually set test sources at 0.9V RMS so that the input filter op amps are not close to clipping. If you adjust your 2.56V RMS source to 0.9V is the attenuation still there?

Also, if your tone source has high output impedance there could be attenuation through the evaluation board's input filter stage.

Joshua

PDF

Hi Joshua, thank you for your response.  Yes, I misread the data sheet and since my post (just this morning in Aus') have re-run my testing.  I am still getting attenuation.  The attached PDF indicates the steps I took this morning, I'll summarise as follows:

STEP #1: this was me basically calibrating the levels to and from my soundcard analogue inputs and outputs.

STEP #2:  I first generated a 1 kHz sine wave from within my measurement software (WinMLS) at -24 dBFS which resulted in a signal level of 0.69Vrms at the 3.5mm input socket terminals on my EVAL-ADAU1452REVBZ board.

STEP #2 Observations shows voltages measured, and some strange readings.  Can you respond to these?

I end up with just noise when I try and measure a signal through the EVAL-ADAU1452REVBZ board; the signal is attenuated too much...

Cheers,

Mark

Hi Mark,

690mV RMS is a good level for the input signal. It will read at -7dB on a SigmaStudio level detector.

There is some attenuation expected through the AD1937 CODEC, which on my board decreases the voltage from 690.2mV RMS at the input to 616.3mV RMS at the output. This is about 1 dB. The evaluation board does not attempt to compensate for this gain offset. The analog input and output stages are both unity gain.

However, I would not expect to see a decrease from 690mV to 330mV RMS.

You also mentioned the output is noisy. 690mV to 330mV is only few dB of attenuation, and 330mV RMS should be more than enough to hear a clean signal. So something is going on beyond simple attenuation.

Can you share your SigmaStudio project? And, are you using the evaluation board as-is, or have you made any modifications?

Joshua

Hi Mark,

690mV RMS is a good level for the input signal. It will read at -7dB on a SigmaStudio level detector.

There is some attenuation expected through the AD1937 CODEC, which on my board decreases the voltage from 690.2mV RMS at the input to 616.3mV RMS at the output. This is about 1 dB. The evaluation board does not attempt to compensate for this gain offset. The analog input and output stages are both unity gain.

However, I would not expect to see a decrease from 690mV to 330mV RMS.

You also mentioned the output is noisy. 690mV to 330mV is only few dB of attenuation, and 330mV RMS should be more than enough to hear a clean signal. So something is going on beyond simple attenuation.

Can you share your SigmaStudio project? And, are you using the evaluation board as-is, or have you made any modifications?

Joshua

Hi Joshua,

I've been able to narrow down one of my problems to being a faulty line input on my sound card, so at least now I am getting a reasonable signal back into my measurement software.

A few follow-up questions as I consider this in more detail:

1. You say it will read -7 dB on the SigmaStudio level detector.  What is the reference for that input?  (Assuming dBFS?)
2. Is there a way to adjust the analogue inputs gain on the AD1938 Codec?
3. Can I meter the AD1938 codec analogue inputs and outputs?
4. I'm still unsure why the Vrms level is always double across the ADC1LP & ADC1LN test points for analogue input #1?  e.g. 0.3Vrms at 3.5mm socket pins, but 0.6Vrms at those test points.
5. Since being able to actually make some IR measurements of the signal path through the Eval' board I've noticed variability in the IR each time I measure it (I also note that I lose around 1dB passing through the Eval' board, even with no processing blocks in the file).  Even if just a few seconds between each sweep.  And the difference is considerable.  (See following images).  Why is this so?  They should be the same each time if nothing changes...
   
   
   

A side note.  It is pretty frustrating when working with SigmaStudio on a laptop.  I added another IC to a new "dummy" project (an AD193x) to see if I could interrogate the settings of the codec on the Eval' board.  But when I clicked on the AD193x Register Controls window there's no scroll functionality so I couldn't actually select the "Get Current Settings From Chip" button (see below image).  I can only view all elements of that window when I enlarge it on my desktop monitor.  I'm assuming I'm missing something with GUI layout settings perhaps?  

Hoping you can help me!

Cheers,

Mark

Hi Mark,

Glad you found the source of the attenuation.

1. The level meter would read -7dBFS, that's correct.
2. The evaluation board is not set up for variable analog gain; you would need to rework different feedback components into the op-amp filter circuit. The schematic is in the user guide if you would like to do this.
3. Yes, the ADCL1P and ADCL1N (etc.) are the analog input pins of the AD1938. Similarly the DAC1L (etc). pins are the analog output pins.
4. The AD1938 features differential inputs, so the eval board uses a single-ended to differential converter made of op amps. The noninverting and inverting stages both have a gain of 1, so if you measure differentially across the ADC1LP and ADC1LN pins you will see twice the amplitude. I think this explains what you found. The schematic for the filter is in the user guide. I would like to make some notes about probing:
   • If you use a scope to measure ADC1LP with a ground reference, the scope input should be AC-coupled. The ADC1Lx pins are biased at 1.5V which would offset RMS audio levels.
   • Be careful of grounding if you want to measure differentially across ADC1LP and ADC1LN with a scope. The evaluation board is typically connected to earth ground through the USBi+PC, and most scope probes are earthed as well. So connecting a ground clip to a biased ADC pin could cause some problems. 
5. I don't usually see results like this. The frequency response measurement should be repeatable. Are you using a sine sweep to measure the response or something else?

You mentioned you wanted to query the settings of the ADC with SigmaStudio. By default the AD1938 is in standalone mode; all the SPI pins are grounded. You may remove the 0ohm resistors and attach an SPI interface to the test points in order to read or change registers.

SigmaStudio should show scrollbars when content doesn't fit. I see them on my machine. You might check in Windows display settings whether "Automatically hide scroll bars in Windows" is checked.

Please let me know if you have any further questions!

Joshua

Thanks for your responses, Joshua!

Your responses to questions 1 thru 4 are great!

Regarding question 5.  Yes, I have been using a sine sweep for my measurements.  I regularly measure lots of different audio equipment, including electronic measurements, but I have not seen such variability in responses when simply running the same measurement over and over...  When no settings have been changed of course...  Is there perhaps some dynamic processing in the AD1938?  What else could cause this?BasicLoopbackTemplate.zipI've attached one of my basic loopback testing projects so you can see that there is no other processing in-line...  

Oh, I figured that since I could place an AD193x IC into my SigmaStudio design that I might also be able to interrogate the chip on the eval' board.  I'll try and work something out.

Regarding scrolling in windows.  All my other TABS and settings windows have scroll bars, but the Hardware Configuration tab for the AD193x register controls does not.  It's the only one that doesn't...  Shame...  (Yes I confirmed Windows settings).  Any further ideas why this may be?

Kind regards,

Mark

Hello Mark,

Let me chime in here since I support the AD1938 and the ADAU1452. 

One of the changes we did on the Eval-ADAU1452RevBZ verses the older Eval-ADAU1452MINIZ is that it is possible to change the codec to be controlled from SigmaStudio. I detailed those modifications in the user guide for the eval board. It is a bit involved of a mod but it is possible. 

The codec will come up with the default settings with only a few exceptions when in standalone mode. The differences in standalone mode are detailed in the datasheet if I remember right. The volume controls will be up at 0dB with no attenuation when it powers up. Standalone mode will not change this at all.  

Regarding the scroll tabs for the GUI. This is an older GUI and there are not very many controls for this part. Obviously the programmer did not enable scroll bars in the window object which I agree is an oversight. I have not noticed this since I usually have enough room on my screen. You might have to close out the output window or the Tree Toolbox window to make some room. 

I reported this in our work request system for the programmers. 

Now regarding your test inconsistency. I suggest you try to isolate if it is the input to the DSP or the output. 

I have not looked at your project but you can do two things. You can setup an oscillator in the project to send out to the DACs. The oscillators are at 0dBFS so you can see how you are measuring this externally. You can change the frequency and test the performance. 

Then you can use the meters to look at what level is coming into the DSP. Keep in mind that the meters in SigmaStudio are not meant to be precise reference level meters but they still can do a decent job of reading the level. The only thing that I wish were different is that the programmers set it up so that a 0dBFS signal will read as -3dB. So keep that in mind when looking at the incoming levels. 

Dave T

Hi, Dave!  Thanks for your responses too!  That is very helpful.  

I'm a little nervous about ruining my Eval' board, however the instructions in the updated User Guide for undertaking the modification seem pretty clear.  If you look at the example file I put together in the attached Zip file, I'm assuming that this is how I configure the software (after implementing hardware mods) to then be able to also interrogate the AD1938 IC?  1727.BasicLoopbackTemplate.zip

I can try your suggestion with oscillators to test the DACS, however, I'm not too concerned if there is an overall level difference through the device, what is very important is consistency.  I can easily reference the signal path through the Eva' board to overcome any frequency dependent anomalies, however, what I am seeing is a different IR every time I make a measurement through the board.  Therefore, referencing is not possible, and I can't confirm operational consistency. 

I've also included a collection of IR measurements I have made through the Eval board, using the included simple SigmaStudio file.  Every measurement is different! 

XLSX

The similarity between concurrent measurements would ideally be identical, or almost identical (I'll also included some loopback measurements through my soundcard (SC) for reference.  My main focus is on the pre-ring in the IR's having significantly different amplitudes, and I'm wondering where the ringing is coming from (and hence my need to retrieve the Codec parameters so I can see what filtering etc is active to see if that is contributing).