Hi Sven, I tried to see if I could measure those spurious signals with my spectrum analyzer + RF preamp and antenna, but I didn’t find them. This makes me think that they are somehow generated by the QA403. I may be wrong and I’ll wait to see the measurements of Matt and others on the forum.
Where do you put the low pass filter and whats the corner frequency?
Here’s a quick look at 9 kHz to 1 MHz.
You can see in marker 1 above the 10 kHz that the QA403 is generating. Marker 2 is 100 Khz, and marker 3 is 768 kHz. Note the spectrum analyzer is showing dBm and the QA403 is dBV, but roughly, they are about the same. So, we can look at the Marker 2 and see it’s a few dB better than the fundamental. And marker 3 is is about 10 dB worse than the fundamental.
Marker 3 dominates, and so roughly, we might estimate the out of band total to be around -85 dBV
I added a shunt 6.8nF, but the board that had the 6.8nF picked up too much hash to be useful.
Out to 1 GHz, it looks as follows (red=off, blue is generating at 10 kHz and -100 dBV)
Later I’ll try 6.8nF/NP0 directly at QA403 output pins inside the case.
The low pass filter is an RF filter that has an impedance of 50ohm and an attenuation from 12 Mhz to 1000 Mhz greater than 60 dB, from 0 to 10Mhz it has an insertion loss less than 2 dB. It is a filter that I use for other radio frequency purposes but that I wanted to try in unbalanced output of the QA403.
1 Like
Thanks Matt, your out measurements at 1Ghz confirm mine and Sven’s indirect measurement was right. I didn’t think the QA403 could generate spurious signals around 500Mhz. You never stop learning 
Here’s a quick measurement on the Topping E70 Velvet. Red is off, Blue is on. You can see a similar spread around 300 MHz, peaking around -80 dBm, and what looks to be a clock around 450 MHz. There’s a lot of programmability on these DAC parts, so if that is a clock, the difference in clock freq (466 versus 450) isn’t a surprise. But I think the conclusion here is it’s in the DAC part. The part is a sigma delta, which means it’s oversampling and moving noise around the spectrum via DSP…
Hi Matt, I had the same idea as you to try with a DAC and I made this measurement with a Topping D50III that mounts a dual ES9039Q2M. The behavior is similar and I agree with your conclusions.
Hi @Giuseppe, thanks for the additional data point! I will try some series ferrite + shunt c inside the case, and also some shunt c by themselves. I don’t think the ferrite will hurt THD, but series L will unless it has a high saturation current.
Hi Matt, if you will have a good solution to insert the rf filter within the box, could you descibe it that I can retrofit my unit?
Hi, measurement update. Had ordered a BNC inline filter LP 200 kHz. With this one all RF noise is gone. My ST 1710A shows now 0.003 % THD at unbalanced input with a 0 dBV 1 kHz sine. Selecting the internal 80 kHz LP filter of the ST 1710 THD goes down to 0.001 % whis is the lowest of the instrument. Without a Signal from the QA403 the Sennheiser UPM550 shows 60 µV noise at bandwidth 10 Hz - 100 kHz unweighted. Therefore I will leave the QA403 untouched and if necessary I will use the external filter.
Hi @Sven, thanks for the posting the update.
I am behind in posting an update, but your post reminded me. Below is the RF spectrum of the output, conducted. Blue trace is QA403 off, red is L+ with no filter, Green is R+ with shunt 10nF inside the case, and teal is R- with a ferrite bead and shunt 10n
Stripping back a few traces, you can see L+ with no filtering and R- with maximal filtering (bead + shunt). There is slight improvement. But mostly it’s the same.
Additionally, with the lid off the QA403, touching the crystal on the processor allowed the spectrum can be pulled. This is because the finger loads the xtal, changes the frequency and thus you see the entire spectrum shift a bit.
Combined with the fact that the bead + shunt c didn’t show much improvement, I think this suggests most of this energy is radiated inside the case from the processor to the BNC connectors after the components. And from the connectors, the signals are then conducted out to the DUT.
So, the correct solution here isn’t an internal filter–the signal will creep in after the filter. The solution is probably a metal can over the processor if this is to be improved in a future product.
And the inline BNC also solves the problem nicely as you have demonstrated.
1 Like
Hi Matt, thank you for your update. Interesting that it is hard to attenuate the RF noise internally. So it is as it is. For my purposes the external BNC filter is fine.
Oh, during testing I saw that above 20 kHz the software does not calculate THD even when sampling rate is 192 ks/s and the display extends to 100 kHz. Also the measured level is displayed as 186,64 dB whereas under 20 kHz it is diplayed as -06 dB. Why does this happen?
Hi. To set the THD measurement interval, you need to access the “THD Option” context menu and then set the “Measurement Start frequency” and “Measurement Stop frequency” fields. To access this menu, right-click on the “THD” button on the QA40X sw.
Hi, today I received a DC-100kHz BNC in-line filter. Now is the residual noise from the output almost gone. My ST 1710A shows now a THD of 0.0015 % which gives me now the oportunity to verify my Audio-AC voltmeters and distortion measurement instruments. As already stated this distortion level is sufficient for my purpose and amplifiers.
1 Like
Hi @Sven, thanks for the update. Where did you source this part from? Seems like a good thing to have in the drawer.
I agree. I was going to ask the same question. I’d like to have an inline filter like this as well. Thanks
Got it from China on the German Ebay website Low Pass Filter LPF Filter With BNC Connector 100K For RF Ham Radio Uses | eBay.de
There are offers also for other frequency limits like the 200 kHz which I also have.
1 Like
Hi,
Interesting discussion about the 403, I have been following it from the beginning. I had a look inside the 403 when I was designing an external input signal attenuation circuit (0dB,-6dB, -10dB to enable mesuring higher powered PA’s) and noticed that there is a good isolation between the USB powered part and the isolated analog circuitry. However, I noticed at the same time that the case is connected to the incoming USB GND. The case surrounds the floating analog circuitry and there is a capacitive path that can leak high frequencies to the floating part. I have also noticed spurios very high frequency noise on the outputs. My plan is to do a small grounding modification (ground the chassis to the analog part and leave the USB-poriton “floating”) and see if it has some impact.
Hi, interesting finding. I hesitate to open the box but am curious what you will detect.
Hi to matt. Can you enlighten us regarding the case grounding?