I’ve been struggling to produce an attenuator for use in high power amplifier testing with the QA402 and have encountered yet another mystery.
My latest iteration is composed of a simple voltage divider to be connected in parallel with my 8 Ohm load. Testing showed some odd behavior when setting the Full Scale Input to values that should have worked fine. Much testing followed and I’ve simplified the test configuration down to just the QA402 and my divider network. I did this to remove any suspicion that my load resistor or amplifier was causing the problems.
I’m using the QA402 to generate an 18 dBV signal from its singled ended outputs. That signal is feeding a simple voltage divider composed of three resistors. The divider provides 40 dB of attenuation when the signal is monitored across R2, the sensing resistor. I’m then using QA402 differential inputs to monitor the signal across R2. The shells on the two differential inputs are left disconnected.
To illustrate the mystery, I’ll show two tests. One with the Full Scale Input set to 12 dBV, the other with the Full Scale Input set to 6 dBV. Both should yield reasonable results with the -22 dBV differential input sighal.
In the 12 dBV test case, all is normal. The gain shows a 40 dB attenuation, as expected with my attenuation network. THD and noise seem fine.
In the 6 dBV test case everything clearly goes to crap. It looks like the QA402 is complaining about a range violation and protecting itself from the -22 dBV input signal, something that should not happen.
I have only one theory as to what is happening. For some reason, the QA402 is monitoring the signal between one of its differential input connections and ground for protection purposes. The attenuation network would provide only 6 dB of attenuation in that case. The input signal would then be 12 dBV, more than enough to trigger a range violation on the 6 dBV Full Scale Input test, but not quite enough to trigger it in the 12 dBV case.
The voltage being applied to the unit in your setup is 3.93V on the ‘-’ input and 4.01V on the ‘+’ input. 6dBV is 2V, so the QA is doing it’s job and tripping the protection/OL.
Hi @Buckskin. The thing is strange. I tried to replicate what you (identical type of partition) did and did not encounter any problems. I attach screenshots at 0, 6 12 db and as you can see there is no problem. I am, however, using a QA403
I just did the same. No issues with my QA403 taking the tap in either polarity from the centre resistor.
0/6/12/18dBV all fine. Correctly reading approx 85mV (~-40dB) (I used a 220R).
Many thanks for replicating my test.Very interesting that the QA403 appears to perform differently. Perhaps my QA402 has a problem?
Before I run off and spend $600 on a new QA403, can you confirm that my connections are the same as yours? Photo below. Top connection goes to the QA402 output. The bottom two are the differential inputs on the QA402. As you can see the ground lines (shield) are left unconnected. Sorry to pester you with this, but I just want to make sure something silly isn’t going on. All three connections use BNC cables coming out/in to the QA402 with BNC to mini grabber adaptors going to the attenuation network.
I don’t know if your QA402 is working correctly, but I definitely know that what you are detecting is not correct. I saw from your pictures that you are using the CHR. It might be worth testing if the same situation occurs using the CHL and also the GEN2
Hi @Buckskin. It would be interesting if someone with a QA402 tried to replicate the measurements. It seems very strange to me that there is a design flaw in the QA402 (I would tend to rule it out). This leaves open the possibility that your QA402 is faulty, although it would be worth investigating further.
Hallo the picture you show i had on my 403 sometimes too on different Measurings, it happens sometime while i switch the Attunuation or other Settings (Not realy shure what i changed while Measuring running) in between an running Measuring but when i removed the Power from 403 and return it with hitting the Button “New Settings” it worked again fine.
Has you trried it???
To remove the Power and using the Button “New Settings” , Make all Setting new on the Software and start the Measuring again???
The most importand for me was the “New Settings” Button on File Tab!
Thats is what i do if something Strange happen on Measurings, Helpped me offten.
Not sure why, but i guess the Software loose probably the syncronisation with the Hardware or so on some Parts.
Hi @Buckskin, the QA403 and QA402 input stages are the same.
What does the time domain look like when you are seeing this signal? Do you hear the relay clicking?
If you remove the 200 ohm, does it still exhibit the same (with range set to 24 dBV)? How about if you remove the 200 ohm, drop the output level to 12 dBV with input set to 18 dBV?
What causes the relay to activate is the ADC sensing an overflow. If you switch to the time domain, you can usually see was the issue. For example, if I drive 18 dBV (balanced) with full scale input set to 12 dBV, I see the following in the time domain. You can see the signal tips approaching nearly +/-6V, and with the full scale input being 3.98Vrms = 5.61Vp, that causes the atten to activate.
As you can see, in the 12dB case, all is well. In the 6dBV case there is a troubling spike and the reported input levels take a stunning nose dive.
Eliminate the 200 ohm resistor, leaving everything else unchanged and the reported input levels drop to just a few mV. No drama with any relays clicking. 18dBV input or 12dBV input, range set to pretty much anything you like and the same thing - near zero reported input levels and no relay clicking. Time domain clean - no spikes. Frequency domain shows lots of noise down around the -100dBV level, the 1kHz tome sitting at about -56dBV. Time and Frequency domain plot below for the 12dBV input and 18dBV range requested test case.
Please note that in all my test cases I’m using the singled ended output of the QA402, not the balanced outputs.
In an interesting twist, I decided to feed the attenuation network from the balanced outputs of the QA402. The only change was that I set the output level to 12dBV instead of the reference test case of 18dBV since I was using the balanced outputs. Everything performed just fine. I can go down to 0dBV input ranges with perfectly nominal measurements. No range violations. No clicking.
Frequency domain plots below when setting the balanced output to 12dBV and 0dBV input range. Looks good.
So… Broken QA402? The only serious change with moving to balanced outputs is eliminating the ground. Hence it smells like a grounding issue inside the QA402.
Why is this signal not symmetric about zero? For the 12 dBV case it indeed looks symmetric.
Maybe you could feed in a single-ended signal right at the full scale input limit and look at the time domain for each. That is, set it to 0 dBV full scale, feed in a 0 dBV signal (from L+ out to L+ in, and ground L- in), show the time domain plot.
Then feed in a 6 dBV signal while in 6 dBV full scale, show the time domain plot, And again for 12 and 18 dBV.
Taking a guess here that you wanted a loop back between the single ended output and a single ended input (negative input terminated). I’ve removed the attenuation network and run the four test cases. Time domain plots below.
Did you see the final test case I ran above. With the balanced inputs?
OK, so these look fine. Can you put this back in the config that isn’t working (6 dBV input) and measure the DC (relative to BNC shell) you see on the L+ and L- inputs?
The key is to figure out why an AC coupled signal isn’t symmetric.
Yes, I see the balanced output at 12 dB. Is it possible that your 6 dBV input is just firing a bit earlier than the other channels for overvoltage? That is, if you run 5 dBV into the problem config, does it work?
Once again, thank you for your time. I appreciate it.
Here is what I did.
Reconfigure to original test setup. 18dBV single ended output, attenuator in, differential input across the 200 ohm resistor input and 6dBV input range.
Observed that the problem is still there. Relay clicking. Spike in the time domain.
Connected my Fluke 87 DVM (on DC mVolts setting) between one of the differential inputs and its shell.
Observed that the DVM was jumping around. Usually around about 20 mV, but sometimes getting as high as 150mV.
Just for grins, I zoomed in on the spike. Shown below.
Got a bit confused here. The generator is set to 18 dBV in my test case. With the attenuator in place, the analyzer sees -22 dBV across its differential inputs. No problems on the 12 dBV input scale. Problem show up on the 6 dBV input scale. In any case, I pressed on with the suggestion to dial down the generator output a see what happens.
Now it gets interesting. With the generator at 18 dBV, the input range at 6dBV and the QA402 clicking away, I started dialing down 1/4 dBV at a time. When I got to around 16 dBV on the generator, the clicking stopped and I got this in the time domain.
Question. But is this DC offset present at the output of the DAC or is it introduced by the input and conditioning stages that precede the ADC? To check this, you need to use an external oscilloscope and check the output of the DAC (BNC).
Hi @Buckskin, you can of course send it in any time you wish, but in 99% of these cases, there’s not a problem found–instead it’s an assumption that was being made that was invalid (either by the operator or by the software)
Plus, you should be able to verify most everything yourself: Your DAC outputs are identical in amplitude, they are correctly inverted (for L- out compared to L+ out). And things are measuring as expected in single ended mode. I know above you checked L+ out to L+ in (with L- grounded). But you should also be verifying L+ out to L- in (with L+ grounded). And L- out to L+ in (with L- grounded). In other words, confirm that every output is doing what is expected, and ditto with every input.
Can you draw again the precise connection diagram (don’t leave anything out–for example, scope connections or other stuff)? I’ll order leaded resistor from digikey in the same values you are using to replicate. Or, if you want to send it in the with the resistors, I can take a look.
In the plot you posted, the following states can be seen:
You can see the attenuator is getting turned off in the middle of an acquisition. This is likely because it wants the atten off for the measurement (<= +18 dBV full scale input), but the protection circuitry runs async from the acquisition.
So, can you try this: Set everything up with but don’t plug in to the L+ and L- inputs AND make sure the QA402 is stopped. And then, plug in L+. If that causes an overvoltage, you’ll hear the relay click. But it shouldn’t because the DAC isn’t running. And then, plug in L-. That again shouldn’t cause a relay to click.
And then do a control + space–that will run a single cycle. And share that time domain plot. do this in balanced loopback–just your 10K series R from the DAC outputs and the 200 ohm balanced R. that will let us see what the initial overvoltage event looks like.
