QA40x and Transformer Measurements

[matt]

There have been some questions on how to measure transformers using the QA40x. The transformer used here is THIS from Amazon. This transformer is designed for transforming a high-z tube output to a low-z suitable to drive a 4 or 8 ohm transformer. The specs indicate a 5KΩ input Z and a 4 or 8Ω output Z.

A 5KΩ primary and 4Ω secondary related to the turns ratio via:

\frac{{{Z}_{2}}}{{{Z}_{1}}}={{\left( \frac{{{N}_{2}}}{{{N}_{1}}} \right)}^{2}}

And so the turns ratio is the square root of the impedance ratio:

\sqrt(5000/4)=35.35

And in dB, this is

20*\log_{10}(35.35)=30.96 dB.

Now, we know the large voltage on the high-z primary will be transformed to a smaller voltage on the secondary, and so the 30.96 dB will be a -30.96 dB (since it’s getting smaller).

Let’s look at the setup to drive the transformer:

The QA403 output is driving directly into the primary. This is fine, because we expect the primary is high-Z. But we might not know exactly the impedance, and the 100Ω output impedance of the QA403 may or may not influence the exact amplitude we’re driving across the transformer. For this reason, we take the output voltage across the primary and run that into the right channel. If we specify that we’d like to use the right channel as the reference in the measurements, then the loading on the QA403 won’t matter–it will all be referenced to the right channel.

On the secondary, note a 4 ohm load is connected. The transformer needs load to do its job. For this measurement, a resistance substitution box works well because it allows you to easily play around with the load and see how that impacts the flatness.

The test setup is shown below. Note the BNC T’s on the L+ and L- output. Scope probes are using to drive the primary balanced (note that ground clips aren’t used), and the secondary goes into the resistor substitution box and is measured single-ended into the left channel.

A sanity check was made with a DVM. A 1 kHz tone at 16 dBV was generated, and the output across the primary was measured at 11.84Vrms, and the output across the secondary was measured at 258mVrms. Note these measurements were made in IDLE mode to give a constant tone the DVM can measure. This gives:

20*\log_{10}(0.258/11.84)=-33.23 dB

Returning to the measurement using the QA403 instead of the DVM, we get:

Using the reported QA403 values we get

20*\log_{10}(0.2576/12.02)=-33.38 dB

From this, we know our setup is correct. The values measured by the QA403 are more reliable than your handheld DVM. The goal in checking the with DVM is to ensure something big hasn’t been missed, like a 6 dB correction factor some place–these are easy to miss when making balanced measurements.

We can then run the Automated Test AMP Frequency Response Chirp with the following settings. we’ll drive at 0 dBV (which is actually about 6 dBV since we’re driving balanced) and we’ll plot the gain and the phase.

The resulting plot is below. We can see the output has rolled off about 3.5 dB at 20 kHz. We can also see the phase is 0 degrees at 1k, meaning we have the measurement polarity correct.

The gain at 1 kHz is -33.36 dB, which is within 0.02 dB what we measured using tones.

Now, recall above the transformer impedance winding ratio suggested the turns ratio would give a voltage transformation of:

20*\log_{10}(1/35.35)=-30.96 dB

But we measured -33.36 dB. The discrepancy here is likely the assumed primary impedance. We’ll measure that soon.

[matt]

The transformer input impedance was measured using the MISC Input Impedance plugin. The same setup as above was used. For this test, an early version of a QA462 was used. The QA462 is like the QA461. But while the QA461 had a 0.02 ohm current sense resistor with a gain of 50X INA (giving 1 ohm effective output impedance), the QA462 has 0.02ohm, 0.2 ohm and 2 ohm (button selectable), followed by a 50X INA (this gives effective sense resistor values of 1, 10 and 100 ohms). The QA462 uses the new-ish INA849 from TI as the current sense instrumentation amp. The INA849 is laser trimmed, delivering killer gain matching and fantastic CMRR. It has >8 MHz of bandwidth at G=50, which improves considerably the ~70 kHz current sense bandwidth of the QA461.

Using the 100 ohm effective current sense resistor range, you can readily measure the input impedance of higher-impedance inputs, including amp input stages.

From the above, we can make a table of the input Z at 1 kHz:

Transformer Load	Impedance (Ω)
4Ω	6.9kΩ
8Ω	5.7kΩ

Above we used the equations to determine the turns ratio:

\frac{{{Z}_{2}}}{{{Z}_{1}}}={{\left( \frac{{{N}_{2}}}{{{N}_{1}}} \right)}^{2}}

And we relied on the Amazon-seller’s claim the input Z was 5k to estimate the turns ratio. But with the impedance measured at 6.9k for 4 ohms, we can re-calc the turns ratio:

\sqrt(6.9k/4)=41.53 = -32.36 dB

Note the sign on the log gain was inverted to match the measured above. And this agrees reasonably well with with the -33.36 dB we measured above based on the impedance and turns ratio relationship.

We can also use the MISC R,L,C plugin to graph the impedance and inductance of the transformer input with a 4 ohm load:

Note from the plot above we can see the impedance at at 1 kHz is again 6.9kΩ. But with this plug-in we can also display the reactive component of the measurement, which at 1kHz is 115 mH.

[restorer-john]

I really want to get hold of a QA462 as soon as it is available.

[Giuseppe]

Hi Matt, would it be possible for those who own the QA461 to make a modification to have the performance of the QA462?

[Dan]

@matt thanks for a great series of posts!

[ScottRichardson]

I used this method to measure the frequency response of several very different output transformers, all different grades (large high-end, standard, and inexpensive). I would have expected them to be vastly different regarding frequency responses, but the primary differences were in the high end with lower-priced ones dropping off somewhat above 10kHz. My prediction was that the smaller, less expensive ones with less iron would have less output at lower frequencies, but that was not the case because they only decreased slightly. I suspect that the transformers will behave differently with more power and higher voltages than in this test. Thoughts?

[MarkT]

At low frequencies saturation is the limiting factor, so lower frequencies and higher voltages will be more demanding as V.dt = L.dI, and lower frequencies mean larger dt.

[ScottRichardson]

Therefore, given that lower frequencies and higher voltages are more demanding and cannot be replicated with a QA403, will the test results with the QA403 provide an accurate analysis of an output transformer, in particular its ability to handle low frequencies? The ability to handle low frequencies is usually one of the critical factors in evaluating an output transformer.

[MarkT]

I think you need the power amp to test an output transformer - but that can be driven by a QA403…

[Frunse]

Hey Matt i ordered an QA461 at 28.05.2024 on Saelig but it is not in Stock.

When you will release the QA462 ?

I guess an easy Update isn’t possible because the INA 199 (SC70 or UQFN) has an other Footprint to the INA 849 (SO-8 or VSSOP) even if i would use my own Sensresistors externaly Only swaping the board would do it.

Is it possible to get only the Board without Enclousure directly from you any day (not soon, but if i want to swap)???

Is a bit hunting the white Rabit but why not asking…
Also QA451B Ordered also not in Stock any suggestion how long i need to wait ???
At least they send the Ordered QA403 and QA 472 this day, so i get soon something i could Play with.

Robert

[ScottRichardson]

Yes, a power amplifier would be best to accurately evaluate an output transformer, but a customer has a numerous loose transformers from different units that he wants to evaluate. There’s no practical way to install each in an amp to test.

[MarkT]

I think to characterise without power requires measuring the magnetic circuit cross-sectional area and the numbers of turns, and knowing the magnetic material used…

[ScottRichardson]

Unfortunately, that is not something that is physically or economically practical to do. I guess the QA403 measurement will be all I can do.

[Frunse]

Any News on release Date for QA462???
Or what about QA455? will this Device happen?

Days, Month, Years ???

no Hurry, only to know Proximitytime for my Enclousure Project to go further.
Still waiting for my ordered Powersupply Boards from Audiophonics, also not arrived yet.

Have still Projects to work on. So I try to stay cool here under the Roof @ 30°C.

LOL

Robert

[matt]

Hi @Frunse, yes both are still being worked on. But no idea on dates. Sorry about that!

[THDaniel]

Nice! This is exactly what I got my 461 for (and was disappointed to figure out it’s mainly useful to get T/S parameters and not a full bandwidth measurement). Out of curiosity: would you mind posting an impedace graph of a fixed resistor (e.g. 16R) with the 403 set to 192k sample rate?
Thank you very much.

[matt]

Hi @THDaniel, here are some plots of the QA462 driving into different resistive loads (1, 10, 100, 1k, 10k, 100k). For each load, there are one of three sense resistors on the QA462 that can be selected: 0.02 ohm, 0.2 ohm and 2 ohm. The sense resistors current is then gained up 50X in an INA849. So, the effective sense R is 1, 10 or 100 ohms. The 1 ohm would be used for higher current situations (smaller z loads) and the 100 ohm for lower current (higher z loads)

In the plot below, you can see the load is 1 ohm, and the readings you get with a 100 ohm, 10 ohm and 1 ohm sense resistor. For this, drive was -30 dBV into QA462, which means -10 dBV out of QA462. This is 316 mV of drive voltage into 1 ohm, which means 316 mA of load current, which means 316m * 100 = 31.6V sense current, which is non-sensical. So, there’s some consideration required when picking the sense resistor. As the graphs below will show, you want an effective sense R that is about the same magnitude as your load.

Note the flatness of the 1 ohm sense resistor up to 100k. Much better than QA461.

And now with 10 ohm load. You’d want the 10 ohm sense resistor here:

And now with 100 ohm load. Both the 10 ohm and 100 ohm sense to fine here:

And now with 1k ohm: The 100 ohm is preferred here as you can see the noise creeping up around 20k on the 10 ohm:

And 10K load. 100 ohm for sure.

And finally, a 100k ohm load. The 100 ohm sense is useful to 20k or so:

In the end, I think this will be a good improvement over the QA461. It was bothersome the QA461 couldn’t measure the input Z of the QA472, for example.