Expanding FR measurement beyond 10Hz-20kHz?


I need a way to measure FR on signal transformers we wind in-house, and we’s like to spec at least 10Hz-50kHz. Below 10Hz if that’s possible (it doesn’t seem to be).

How can I do this? Running on Win 10, sadly (hoping for an OSX version some day).

I’ve poured over every feature and menu option I can think of. I can’t figure out how to expand the FR measurement window past 20kHz.

I’ve used the gain function as a crude way to check FR at set frequencies below 20Hz and above 20kHz, but there seem to be dead spots, like 22kHz returns a -200 something reading, while 33kHz returns something usable. Nothing past that, though. Seems odd, since I’m sure this thing can measure past 33kHz.

Am I mistaken in thinking my 401 could measure FR above 20kHz? I’m pretty sure it could.

I know I can’t be the first to ask this question, sorry if it’s been covered already but I couldn’t find an answer.

Try using the amp freq response automated test which should allow you to change the frequency range:

Thank you for the reply! I was able to successfully set up and run the test on the 402. I’m getting different results depending on which test I run, and I’m trying to figure out how to interpret them. These are all 1:10 transformers. All are wound in-house except for the pink trace FR Plot (left)31, which is wound by Peerless/Altec.

Here’s a screen shot of test results:

If I run the test with the ‘FR’ button engaged, I get the flattest transformer response I’ve ever seen in my life. These are the red and blue lines FR plot (left) and FR plot (left)85. That doesn’t seem right.

FR Plot (left)43 and FR Plot (left)86 are the tan and green traces. The are the exact same transformer, same test run twice. Two very different results.

Depending on the order in which I run the tests, the measured output amplitudes differ. FR Plot (left)27 and FR Plot left(86) are two transformers from the same batch of winding. You can see the response traces look pretty much identical except for the amplitudes. If I flip the order of testing around, I get the exact same results except mirrored.

The last oddity is the FR test using the ‘FR’ button is different–it shows a flat response down to 10Hz, and using the ‘Gain’ test and manually sweeping through frequencies shows the same thing.

What am I missing here?

Here’s an example, same test run using the FR button.

Hi @mportnoy! Welcome to the forum! The FR test uses a chirp which can show erroneous results at the edges. The stepped tone @VAR showed is what you want.

it is often best to start at a single frequency and verify things are working as expected. For example, if you test at 1 kHz. are you seeing the turns ratio manifest as expected? You can also look at the time domain and verify you see the lead/lag that you expect. And then move to 10K, 20K, and then down to 20 Hz.

When testing with discrete tones, you can use the “IDLE GEN” button to generate a fixed tone, and that let’s you probe with a scope to verify the lead/lag, and also a DVM to probe the in/out numbers and confirm what you are seeing.

And once you have verified the single tones are making sense, then moved to stepped tones and verify the stepped tones at the discrete frequencies you measured are making sense.

And once the stepped tones are making sense, then you can move to a swept response. The swept response will be more sensitive to various settings (eg FFT size, input levels, etc).

At each step you’ll learn what settings matter (eg FFT size, minimum levels, etc) and can fold that knowledge into the next steps.

The good news is that, while it’s hard to debug this stuff in the frequency domain, you can always pus the TIME button to get to the time domain and that will quickly explain a lot.

So! With the above said, if you make a single measurement at 1 kHz using just a tone does it yield the result you’d expect? How about at 20k and 20 Hz?

Feel free to ask a lot of questions, no matter how basic.

I can’t figure out how to expand the FR measurement window past 20kHz.

To measure past 20 kHz, you’d first want to make sure you are running at 96K or 192K sample rate. The FR will automatically adjust the sweep for the higher sample rate. If you click on 192K and then right click on the Frequency Response button, you get the context menu. In there you’ll note the ending frequency of the sweep is 96K, and the sweep spans 10 octaves (93 Hz start).

And if you want to change X axis end frequency, you can right click on the X LOG button and change it to 100k, for example.

Thanks for the tips.

I think the biggest issue for us at the moment is the difference we’re seeing in the automated stepped tone tests vs manually sweeping through each tone. For example, we manually test at 10Hz (the gen shows 11Hz) and we get a response that is 0.5dB lower than the same amplitude tone at 100Hz, and again at 1kHz.

If we run the automated tests, the variations are much more pronounced; in some cases they are as much as 2dB difference in amplitude, and the automated tests don’t always accurately reflect the turns ratio.

That’s the area where we’re looking for the most clarity to start.

That’s the area where we’re looking for the most clarity to start.

OK, got it! Take a look at the post below as the physics of low frequency measurements will require careful attention to settings. Can you first try the suggestions there and see if you can confirm the automated low frequency stepped tone sweeps in loopback, and then introduce your DUT and see how things change.

But just to double check, it sounds like static (non-stepped) measurements at 1 kHz and low frequencies are making sense is that correct?

BTW, I just ran loopback on the QA402 with the following parameters at 3 different FFT sizes:


Note that sample rate was 48K, and rounding was enabled (right click on Gen1 and ensure “round to eliminate leakage” is selected) and also Flat Top windowing was used (it has a wider bandwidth) and also 0 dBV input range

In the graph you can see there’s not much difference between 32K and 128K FFT. Both give very good results for measurements at 10 Hz. Note the input on the QA401 is AC coupled, with a corner around 33 mHz (4.7uF series and 100K shunt)

Thanks for the tips, Matt. I appreciate it.

Right now, we’re having more strange issues–transformers that are known to have ideal frequency response curves are now measuring with huge dips below 500Hz and above 10kHz. Like a big parabolic curve instead of a flat line.

This issue happens on newly produced parts, parts that were measured already and passed QC, AND parts from other manufacturers that we know are good. It’s very strange. The parts pass every other stage of QC (including listening tests) but fail here.

We suspect some issue with the 402. The issue comes and goes, but it’s been consistent for the last two weeks.

Any thoughts? We’ve tried relaunching the platform, and rebooting the computer. Sometimes that works but sometimes it doesn’t.

Hi @mportnoy, any chance you could send a transformer? Just to double check, your tests in loopback (output straight to input) with the same setting should be consistent and very flat. Can you confirm that is true? You can do thousands of measurements and just watch the screen and make sure there are no glitches to verify that.

And just to double check, you are using an amp of some kind to drive the transformer, is that correct?

Loopback tests show flat response. That was one of the first things we tried when we started getting strange results.

We are not using any type of buffer to test these transformers–we connect to them 402 directly.

Loopback tests show flat response.

Hi @mportnoy, OK, so the hardware in a known state is as expected

We are not using any type of buffer to test these transformers–we connect to them 402 directly.

OK, got it. What is the inductance of the primary of the transformer? Are are you driving single ended or differentially?

Primary inductances vary, because there are several different models. In this case, inductance is 9H. We drive single ended, grounding one leg of the transformer on both input and output. Here’s an example of what happens when we test:

These are seven passes, all on a 1:10 transformer that is known to be good. Input is +6dB. You can see that most of the tests look like the transformer has severe LF attenuation, some worse than others. The yellow trace is the result we would expect to see.

The issue comes and goes. Sometimes the tests work, sometimes they don’t. I hope I can get this solved.

Hi @mportnoy, OK, so at 10 Hz the impedance should be about 550 ohms, which is OK to drive directly but be aware the QA402 has 100 ohms of output impedance, so you’ll see some divider action at the low end. But that doesn’t explain the inconsistency.

And just checking: With 2Vrms in and 1:10 ratio, you’d expect 20Vrms out of the transformer (26 dBV), but it looks like you are seeing about 16 dBV in or around 6.3V on the expected trace. Is that what you expect?

If you set the generator for +6 dBV and 10 Hz, and make a normal (non-swept) measurement, what does it look like? If you let it run over and over, do those measurements looks consistent? They should be rock solid.

Below is a measurement at 10 Hz in loopback. This should be rock solid both in loopback and with the transformer connected (with your transformer, the gain would be >3 dB or so).

If you can confirm the static measurement at 10 Hz (and 11 Hz, and 12 Hz, etc) is rock solid, then we’ll look into what might be different in the swept test.

“so you’ll see some divider action at the low end”

This could explain the contrast between what we’re seeing and what we’re hearing. We’ve got one test example that consistently shows LF rolloff, but in blind audio testing (part of our QC) it has THE strongest low end by far. It has everyone scratching their heads.

I’ll test some more and report back. We just moved into a new workspace and everything is hectic ATM.

If you can send a transformer or even a similar off-the-shelf transformer that is showing similar results, I’d be happy to measure it while you were moving to the new workspace. Just let me know.