Anti-imaging/low-pass filter for the VHM-314 Bluetooth device?

[Dan R. Hansen]

Greetings from Denmark.

I am working on a little project with a pre-amplifier, balanced outputs etc, and a VHM-314 receiver.
Everything else works fine, but when take a look at the output of the VHM-314 module with my oscilloscope, there seems to be a very "raw" output on the audio outputs.

1kHz test signal is stable but with the ladder steps, but if I go to 20kHz, the output goes crazy. Screenshots attached.

The VHM-314 module itself shares the same ground plane on it's board, and it is only connected with one wire to a common ground connection point, that everything else is sharing, so I don't believe it's/can be, a ground loop issue.
I thought it's a matter of making a simple low-pass filer for the output (22ohm resistor + 330nf capacitor), but I don't think the unstable behavior of 20kHz has anything to do with a missing filter? Is it a power supply problem?

What am I looking at here? 

[Dan R. Hansen]

Are there anyone out there, who knows what I am looking at, or have an idea to solve it?

[whollender]

That looks a lot like sampling images on the output which is expected without having a reconstruction filter in place.

If you're not familiar with sampling theory, the DAC output cannot really be considered valid unless the reconstruction filter (anti-image filter) is in place.

It looks like the DAC in this case uses a delta-sigma modulator architecture, so there is a lot of image rejection to begin with, but they are still present and need to be filtered out.

[Dan R. Hansen]

That looks a lot like sampling images on the output which is expected without having a reconstruction filter in place.

If you're not familiar with sampling theory, the DAC output cannot really be considered valid unless the reconstruction filter (anti-image filter) is in place.

It looks like the DAC in this case uses a delta-sigma modulator architecture, so there is a lot of image rejection to begin with, but they are still present and need to be filtered out.

Thank you for your reply. There reason I am extra confused is because of the unstable output I get a high frequencies (20kHz). I am not sure what is going on there?

[whollender]

I'm not sure I understand what you mean by 'instability' at 20kHz. 

I do see some triggering artifacts, where the traces don't appear to overlay perfectly because the scope is triggering on a different part of the waveform.

In all of your plots, there is a low amplitude ~350-400kHz tone that may be directly from the DAC output (possibly the sample clock), or it could be some other interference from probing, but, in any case, that will not be audible, and will be significantly reduced by any kind of low pass filtering.  This additional signal is likely causing the triggering artifacts that are causing the 20kHz signal to look different.

Outside of the high frequency tone, the output of the DAC looks very clean to me.

[Dan R. Hansen]

I'm not sure I understand what you mean by 'instability' at 20kHz. 

I do see some triggering artifacts, where the traces don't appear to overlay perfectly because the scope is triggering on a different part of the waveform.

In all of your plots, there is a low amplitude ~350-400kHz tone that may be directly from the DAC output (possibly the sample clock), or it could be some other interference from probing, but, in any case, that will not be audible, and will be significantly reduced by any kind of low pass filtering.  This additional signal is likely causing the triggering artifacts that are causing the 20kHz signal to look different.

Outside of the high frequency tone, the output of the DAC looks very clean to me.

I understand I messed up in the screenshot by setting the trigger at almost "zero crossing", but the fluctuating reconstruction of the frequency is "my problem". It doesn't behave as nicely as 1kHz and that is what made me think a filer might not work as optimal if there were other problem with instability or something. But I could totally be wrong and what i see on the scope is just a product of me setting things up the wrong way.

There is no filter other than what seems to be DC-blocking on the output?
The output of the Bluetooth module is biased with 50k on each channel vie the input of what it is connected to.

[whollender]

It's not that the scope is setup wrong.  The zero crossing is just happening at different times with respect to the 20kHz waveform because of the other signal that is present.

The traces for the other frequencies aren't moving around as much because the trigger instability due to the secondary signal is much shorter with respect to the overall signal period.

I really don't think there is an issue with the DAC output here, but it is easy to verify that the signal is cleaner with a low pass filter added.

[Dan R. Hansen]

It's not that the scope is setup wrong.  The zero crossing is just happening at different times with respect to the 20kHz waveform because of the other signal that is present.

The traces for the other frequencies aren't moving around as much because the trigger instability due to the secondary signal is much shorter with respect to the overall signal period.

I really don't think there is an issue with the DAC output here, but it is easy to verify that the signal is cleaner with a low pass filter added.

Maybe. I just noticed that if i set the trigger at one point, the plot on the screen would "behave", but when I got up to around 20kHz if would go crazy like on the screenshot. Regardless of the trigger postilion. That seemed like there was some sort inconsistency with the output.

What kind of filter would you suggest, assuming you are versed in such things? I have seen suggestions of sending the signal through a 22ohm resistor and a 0.33uf film-capacitor between the other end of the resistor and ground?

[whollender]

That kind of filter would work.  I would suggest starting with a higher resistance value and lower capacitance value because the output of module seems to be sensitive to capacitive loads (assuming I found the correct module, the microchip RN52 output has a capacitive load limit of 500pF).

I'd go with something like 2.2k Ohm series resistance, and a 3.3nF capacitor to ground.  However, if you have some components lying around, just try something that gives a corner frequency of 20-30kHz.  The formula is 1/(2*pi*R*C) for an RC lowpass.

[Dan R. Hansen]

That kind of filter would work.  I would suggest starting with a higher resistance value and lower capacitance value because the output of module seems to be sensitive to capacitive loads (assuming I found the correct module, the microchip RN52 output has a capacitive load limit of 500pF).

I'd go with something like 2.2k Ohm series resistance, and a 3.3nF capacitor to ground.  However, if you have some components lying around, just try something that gives a corner frequency of 20-30kHz.  The formula is 1/(2*pi*R*C) for an RC lowpass.

That is the thing, I do not have any capacitors of those low values. Nor resistors, though I can put some together to get close.
Wouldn't the the high resistance hurt the maximum amplitude?

[whollender]

If all you have is larger caps, you can try with more capacitance and lower resistance.

The resistance could affect the amplitude depending on what else is connected to the output.

If only the scope is connected, then it should not make any difference because the input impedance of the scope is very high.  If there is something with low impedance connected to the output (like a speaker), then the resistance may have a large impact on the amplitude.

[Dan R. Hansen]

If all you have is larger caps, you can try with more capacitance and lower resistance.

The resistance could affect the amplitude depending on what else is connected to the output.

If only the scope is connected, then it should not make any difference because the input impedance of the scope is very high.  If there is something with low impedance connected to the output (like a speaker), then the resistance may have a large impact on the amplitude.

The module in connected to one of 4 inputs on my DIY pre-amp board. Right now the pre-amp is running in dc-coupled input, but whatever input is selected, goes into a buffer op-amp with an "impedance/load" of 50k to ground. From there it goes into a PGA2311 chip and after that, there are also an option to feed the output into 2xDRV134 chips, for balanced output.
So the bluetooth module wont see a high load on it's output. And I guess I can compensate for the lower amplitude by programming the chip so boost the volume a little extra when the bluetooth input is selected. The PGA2311 can go over unity-gain if instructed.

[Dan R. Hansen]

A little update.
I added a filter consisting of a 22ohm resistor and a 0.33uf film capacitor on each channel. So signal goes through the resistor and then a capacitor is connected between the other side of the resistor and gnd.

The stepping is 95% gone and the wild unstable behavior is gone. However, there are still a problem.
If I look at the scope, I now see a slow oscillation of the 20kHz sinewave amplitude. I guess the original problem still persists, just manifested through another from due to the filter?

What could it be? 
Is it some sort of weird relation between the audio sample rate and the chips filtering?

[whollender]

Can you share a screenshot of the scope?  It would also be helpful to 'zoom out' the time axis to see if you can capture the overall behavior of the amplitude oscillation.

I wouldn't be surprised if the amplifier is oscillating, since the capacitor is much higher than the specified max capacitive load, and there is only 22 ohms of resistance isolating it from the output.

You may not even need a low pass filter here if your preamp design has enough filtering.  Can you share the preamp design?

I still think a higher resistance low pass filter would be a better choice here.  2.2k of series resistance before your preamp will only reduce the amplitude by about 4% (50k/(2.2k + 50k) ~= 0.96), or only about 0.3dB, which is likely insignificant.

[Dan R. Hansen]

Can you share a screenshot of the scope?  It would also be helpful to 'zoom out' the time axis to see if you can capture the overall behavior of the amplitude oscillation.

I wouldn't be surprised if the amplifier is oscillating, since the capacitor is much higher than the specified max capacitive load, and there is only 22 ohms of resistance isolating it from the output.

You may not even need a low pass filter here if your preamp design has enough filtering.  Can you share the preamp design?

I still think a higher resistance low pass filter would be a better choice here.  2.2k of series resistance before your preamp will only reduce the amplitude by about 4% (50k/(2.2k + 50k) ~= 0.96), or only about 0.3dB, which is likely insignificant.

I will try to make a screenshot that captures the whole event.
Is the perhaps a trick on budget Rigol scopes to stitch together a long recorded event while also being zoomed in to a more meaningful visual?

The pre-amp design is of the simplest possible.
7805 + 7905 linear power supply, input to buffer biased to ground with 50K, unity-gain OPA2134 buffer -> PGA2311 -> "unity-gain (+6db)" DRV134 -> amplifier.
I have bypassed the DC-blocking capacitors at the moment, so it all runs on DC-coupling right now.

[whollender]

There maybe a way to capture a long time period on your scope, but if not, you could see if the problem is the same at lower output frequencies, and get a better view of the issue that way.

Have you looked at the output of your preamp connected to the module, or listened to the output?

The instability you see without the low pass filter may not be visible or audible at the output of the preamp, although given the wide bandwidth of all the components it may be visible on a scope.

[Dan R. Hansen]

There maybe a way to capture a long time period on your scope, but if not, you could see if the problem is the same at lower output frequencies, and get a better view of the issue that way.

Have you looked at the output of your preamp connected to the module, or listened to the output?

The instability you see without the low pass filter may not be visible or audible at the output of the preamp, although given the wide bandwidth of all the components it may be visible on a scope.

I have actually only looked at the output through the pre-amp. It seems kind of odd not to not have a looked the the output directly from the module now that I think about it. I guess the reason was to utilize the 50kohm bias load that was on the pre-amp already.
The behavior only comes forth when I reach the higher frequencies. I will create a more extensive test of frequencies as sine-wave wav-files, created in Audacity.

I haven't listened to the sound. I am sure I can't hear the imperfections now. Originally I just wanted to address the instability. Which I guess I mostly did already. I am not even sure the "oscillating?" amplitude would even be a problem under normal audio playback.

[Dan R. Hansen]

Can you share a screenshot of the scope?  It would also be helpful to 'zoom out' the time axis to see if you can capture the overall behavior of the amplitude oscillation.
I still think a higher resistance low pass filter would be a better choice here.  2.2k of series resistance before your preamp will only reduce the amplitude by about 4% (50k/(2.2k + 50k) ~= 0.96), or only about 0.3dB, which is likely insignificant.

There maybe a way to capture a long time period on your scope, but if not, you could see if the problem is the same at lower output frequencies, and get a better view of the issue that way.
The instability you see without the low pass filter may not be visible or audible at the output of the preamp, although given the wide bandwidth of all the components it may be visible on a scope.

Here is what I see. There is no difference at the source, after the filter ,and at the output, other than the extra dBs added by the DRV134 chip.

I have tested a filter consisting of 22ohms and 0.33uf and a filter consisting of 1kOhms and 0.33uf.

The instability varies "randomly" and seemingly there must be multiple factors? All power involved with the pre-amp, 5V logic included, are supplied by linear regulators, so the power is very clean and stable.

With the 22ohm I see some loss in amplitude as the frequency goes up. With the 1Kohm the reduction is so significant, that it there were no point in measuring any further. I assume the cheap module just doesn't have enough umph?
I also added a measurement of 1khz+5khz tone.

[whollender]

Hi Dan,

It looks to me like the signal amplitude in the longer captures is pretty stable, but maybe I'm not understanding what you mean by 'instability' in this case.

The change in amplitude at higher frequencies is to be expected.  An ideal filter with 22 ohms of resistance and 330nF of capacitance will have a 'corner' frequency at 22kHz, where the amplitude is 3dB lower than at DC.  The amplitude will be pretty consistent below the corner frequency, but it will decrease rapidly above it.

With the series resistance increased to 1kHz and the capacitance kept constant, the corner frequency of the filter will drop by roughly a factor of 50 (1k / 22 ~= 50), so the 3dB point would now be 400Hz, and the amplitude will drop very quickly above 400Hz, which is why the signal amplitude at high frequencies is so low.

[Dan R. Hansen]

Hi Dan,

It looks to me like the signal amplitude in the longer captures is pretty stable, but maybe I'm not understanding what you mean by 'instability' in this case.

The change in amplitude at higher frequencies is to be expected.  An ideal filter with 22 ohms of resistance and 330nF of capacitance will have a 'corner' frequency at 22kHz, where the amplitude is 3dB lower than at DC.  The amplitude will be pretty consistent below the corner frequency, but it will decrease rapidly above it.

That was the math suggested to me after I put in the values of components I had laying around. I had some 22ohm resistors because I sometimes use them on the GND connections of audio inputs, as a way suppress potential ground-loops.
In your opinion, is there actually any point in pursuing better performance of this little module? I mean, it starts to seem like you would have to change so much, that you might as well buy a better Bluetooth module/board to begin with?