Impedance matching issues

Hi

If the Filter is a high pass set at 10khz it will pass 25khz without loss. Putting the Filter in Off will bypass the filter and it will Pass ALL frequencies.

Can you see if the frequencies below 10khz are now gone?

Thanks

Brett

Hi Brett,

When I set the filter as OFF, I see that the signal at 25 kHz is about 3 mV. When I remove the DSP and look at the signal directly on the lock-in, the signal at 25 kHz is 60 mV. I think that just putting in the DSP (with the filter OFF) is reflecting the 25 kHz signal due to impedance mismatch.

Thanks,

Vivek

With regard to the frequencies at 5 and 10 kHz, they go away when I turn the filter ON. Comparing these frequency component amplitudes without the DSP versus with the DSP (with filter OFF), there is some signal loss due to reflection at these frequencies as well, but not as much as the loss at 25 kHz.

Hi

Its not a reflection as such. Its just the Accuracy of the ADC and DAC gains and Post analog Filters when working at "beyond audio frequencies"

You can add a "Gain" block in the DSP and Trim this to be exact gain.

Something to also note:

1) the input level is really quite Low in order to take advantage of the full dynamic range of the ADC-DSP-DAC.

you may need to amply the input signal using the PGA. and attenuate it to the desired level Post DAC.

2) what sample rate are you using ? you really need 96k to pass 25khz audio

Thanks

Hi,

I am using a sampling rate of 96 kHz. Could you please tell me more about what you meant by Accuracy of ADC and DAC gains, and post analog filters ?

I will try using a gain block to amplify the input signal. Figure 25 of the manual says that the input impedance is a function of the PGA gain setting..

Thanks,

Vivek

Hi

I think you are loosing some input signal because of the high output impedance of the PMT sensor.

and the input impedance of the ADC-PGA is not a 1M ohm input impedance like the "Lock IN" input.

However this probably really does not matter (within reason). If you loose a bit of amplitude because you can adjust gain in the ADC-DAC and even in the DSP to make up any variation.

Now using the PGA... The PGA inverting input has a gain that varies with Gain Setting.

The NON Inverting input is always 105k input impedance at all gains (as per the data sheet. Page 4, I have not actually measured it).

So I would use the Non inverting input. Then you can put a "Level Meter" on your DSP schematic and adjust the PGA Gain to get a reasonable signal. Say peaking at -10dbfs.

Then on the Output adjust the Level control on the "Left Out Vol"(or what ever output you use) on the "Playback output signal path" tab to the desired level you need to feed the "lock in" input.

These "analog output" level controls are in 1db steps, so if you need Finer control add a digital gain block . Set the "analog output" level controls to the closest setting and only use the digital gain to trim to exact.

Depending on your requirement, actual exact level may not be critical.

Thanks

Brett

Hello Vivek,

Try a test if you can. Try to bypass all the digital sections of the part. You can see in the register settings tabs in SigmaStudio, that you can take the feed of the input mixer and send it directly to the output mixer bypassing, and shutting off all digital paths through the DSP. Then see if you still see the roll-off.

One other thing you should check is that all the blocks in SigmaStudio are set to 96kHz. So at the top of the SigmaStudio window, set it for 96kHz and then press the button to the right to propagate the sampling rate to all the blocks. Then Link/Compile/Download.

Dave T

Hi Brett and Dave,

Sorry for my late response. I was trying a number of things and now it seems like I was able to resolve the impedance mismatch issues by using a transimpedance amplifier after the PMT and before the DSP. The output impedance from the transimpedance amplifier is very low, so it is easily compatible with a > kOhm input load of the DSP. I decided to use the LAUX input and LHP output (pin numbers 6 and 20 respectively). The 10k Ohm highpass filter in the DSP seems to work well, I do not see any signal losses that I was seeing previously.

However, I am wondering about a different problem related to a DC offset in the output signal of the DSP now. I see a DC offset in the output from pin20, which goes into LHP output. The offset is about 1.6V and I believe its coming because this pin is biased at AVDD/2, the AVDD in my case is set to 3.3V. Looking at pin21 (MONOOUT), the virtual ground for capless headphone does not seem to have this biasing, but I am not sure. Do you have any suggestions to get rid of this DC offset ?

Thanks a lot for your suggestions!

Hello Vivek,

The MonoOUT pin is also biased to the AVDD/2 voltage. If you are going out to something that is referenced to ground then you must use coupling caps. If you are going into a differential input then you can simply use the AVDD/2 as a reference for the differential input stage.

The DC bias cannot be removed from the pin, the DAC needs it to properly function.

Dave T

Hi Dave,

I haven't gottena chance to try out the differential input yet, but just to make sure, do you think if I use the AVDD/2 as the reference for the differential input, the output after the DAC will not have the DC offset ? Or it will have a DC offset in any case ?

Thanks,

Vivek

Hello Vivek,

The DAC outputs will always have a DC offset. The DAC has to be able to swing from a no input signal condition (0VAC) to a full scale positive or "negative" from there. Since the DAC can only swing from ground to +3.3V (and not the full range at that) the only way to have a full wave AC signal is to have it sit at AVDD/2, with no signal, and swing up and down from there.

The Capless headphone out simply uses the low impedance drive of the MonoOut to provide the reference for the level when there is no signal. It is not ground but a virtual ground. So the headphone speaker will not pass current if both sides of the speaker is sitting at AVDD/2. If you have to interface with another device that is referenced to ground then you must cap couple the outputs to remove this offset.

Dave T