Measuring Power Amps
Measuring power amps can pose a challenge, because their signals level can readily exceed the maximum input of the analyzer. The QA403 and QA402 both allow inputs up to 32 dBV = 40Vrms. And into 8 ohms, this would be 200W, and into 4 ohms this would be 400W. A few years ago, these would have been enormous amps. But with class D and the associated high efficiencies and advanced signal processing, it's not uncommon to find "prosumer" equipment advertising 1000W. A bargain amp used in clubs is a Behringer NX3000D, advertising 3,000 W bridged into 4 ohms.
There are several amp topologies to be aware of. Many of these topologies can be tracked to a period in time when a certain technology dominated. Below is a summary of the different topologies, divided into two groups: Unbalanced and Balanced. Unbalanced here means the negative terminal of the amp is ground. Balanced means the positive and negative terminals are equal in amplitude but opposite in polarity. Commonly, this is called "bridged" mode. But as the "Floating Transformer Coupled" example shows, you don't need to be bridged to get this behavior.
The reason the Unbalanced versus Balanced dichotomy matters is because it determines the type of attenuator we need.
Below, we'll delve into the different topologies in more detail.
A single ended amp generally ties the negative output of the amplifier to chassis ground. The positive output is then driven. With your DVM, you can play a tone from the amp and measure the positive output relative to chassis ground. And then, check the negative output. If the negative output is essentially zero relative to chassis ground, then you are probably dealing with a single-ended amp (although the output could be floating--see below). This is typical for solid state amps from the 70's and later. Modern chipset amps like the LM3886 fit into this category.
Amps using tubes (including guitar amps) will have a transformer output stage to translate the higher impedance outputs of the tube power stage to the lower impedance needed by the speaker. When one side of the output transformer is grounded, we can consider this an single-ended or unbalanced output stage. If you don't see any continuity between the speaker outputs and ground, the outputs might be floating--see below.
If the amp uses an output transformer and neither side is grounded, then we can consider the output "balanced" in the sense that the speaker is driven by two equal but opposite signals.
A bridged amp will have the positive and negative terminals driving in equal but opposite directions. This is very common on modern chipset class D amps like the TPA3255. The Behringer NX3000 normally has a left and right output, but can be put into "bridged" mode. In this mode, the right channel is inverted. The right channel then drives the same speaker as the left channel, doubling the power to the load. That is, in stereo mode, you get 2 x 900 W into 4 Ω (sqrt(900 * 4) = 60Vrms). In bridged mode you get 1x3000W into 4 ohms (sqrt(3000*4)=109Vrms). The bridging allows you to double the drive voltage level to the speaker (60 → 109Vrms).
The TPA3255 mentioned above has another tweak to the topology that allows it to run from a single DC rail. For example, if you are running the amp from a 50 V DC supply, then the + and - outputs idle at 25 V. As the + rides up to 26 V, the - rides down to 24 V.
- Determine a ground reference on the amp. Usually, this can be the RCA outer ring that is used for the signal input.
- With the amp off, check for continuity between the ground and the positive and negative output terminals. If you see a dead short on the negative output, you probably are looking at a single-ended output which is the left side of the chart above (unbalanced).
- With the amp on and speaker connected, measure the DC at the positive and negative outputs relative to the ground established in step 1. If the DC is non-zero and equal in magnitude on both terminals with the input shorted, then it's probably a Bridge-Tied Load with DC Offset. This will be common in more modern amps and home theater.
- If DC on both terminals is zero, apply a test tone such that you get 1-2Vrms across the speaker terminals. Next, measure the + output relative to ground, and the - output relative to ground. If they are equal in amplitude, you are dealing with a balanced amp. If the negative terminal (relative to ground) is much less than the positive terminal (relative to ground) then it's an unbalanced output.
Most measurements will be made into a resistive load. You can measure into a speaker if you like, but most prefer measuring into a fixed resistive load. A good candidate for loads are planar resistors such as these. You can parallel and series in just about any combination to get the power handling and impedance you need. These planar resistors can handle 100 W continuous, and 500 W peak. For 4 Ω, we'll use two in parallel which pushes power handling up to 1 kW.
The resistors are +/-5% tolerance, which is a bit looser than we might like. The resistors are wirewound, but there's not much else the manufacturer publishes about their construction. We can measure the inductance around 20 µH or so:
While the QA403 and QA402 inputs can handle up to 40 Vrms, it's generally a good idea to give yourself some headroom: 6-12 dB will ensure the THD of the analyzer isn't degraded by excessive inputs. It's also a good idea, from a noise perspective, to target an input range where the attenuator is off. That is, 18 dBV and below.
If, for example, you wanted to measure a 1000 watt amp into 4 ohms, we know
And so this suggests 63 Vrms = 36 dBV. And if we wanted to use the 18 dBV input (for noise) with a bit of margin, that suggests around 24 dB of attenuation.
Above we looked at different amp topologies. How we build the attenuator matters for the amp topology. Let's take a look at this in SPICE. Below, we have a 100Vrms source called VG1. There are two voltage-controlled voltage sources. Note that one of the sources has a gain of 500m, and the other source has a gain of -500m. This is TI's preferred way of generating balanced sources in TI-TINA. We then take that balanced source and drive into an attenuator made up up 5.11K and 680 ohm resistors. This should provide 20*Log10(680/(5.11k*2+680))=24.1 dB
If we run the transient analysis, we get the following. We can see that the balanced source (VM4) is 100Vrms. VM1 is 6.22V, and VM2 and VM3 are 3.11V. And this should make sense: We have used a balanced attenuator on a balanced source, and that gives us both legs being equal and opposite. And the sum of those 3.11Vrms sines are a 6.22V. And 20*Log10(6.22/100) = 24.1 dB attenuation.
Now let's add a ground to the lower leg to simulate a single-ended (unbalanced) amp. That appears as follows:
And when we run the transient we get the following. Note that VM4 is now 50 Vrms (because we grounded one leg). And we can see VM1 is 3.11 Vrms. Half of what we saw in the circuit above. So, it appears this is working as expected.
But wait! Note that VM2 and VM3 are huge AND also in-phase. VM2 is 23.4Vrms and VM3 is 26.48Vrms. In the balanced example above each leg was 3.11V and out of phase with each other. Here they are around 25Vrms and in phase with each other. While the result is correct, the "heavy lifting" of subtracting the sources now falls onto the audio analyzer and that requires large input range to cope.
Fortunately, we can tweak the circuit by shorting out the lower 5.11K resistor. This convert the divider from a balanced divider into a single-ended ended divider:
And with that change, we get the following. Notice now that VM3 is 5.85Vrms. That is, our 50Vrms signal has been knocked down to 5.85, giving us 20*Log10(5.85/50) = 18.6 dB of attenuation. Of course, we can calculate that from the resistors directly: 20*Log10(680/(680+5.11k))=18.60.
A small control board was built to allow 3 attenuator settings, and another jumper to select Balanced or Unbalanced operation:
The finished board is shown below. There is an 8 Ω load on the top and also bottom of the mounting plate. These are paralleled to form 4 Ω. You can see there are screw terminals (marked LOAD+ and LOAD-) on the right side for connecting to an amp. The terminals are HERE and can handle up to 12AWG wire and 16A. There is an 8A fuse in a 1206 package. The fuse can handle 8 A (which translates to 256W) indefinitely, and will pass 16A (1024 W) for a few seconds without opening. The resistors are rated at 500W each for 5 seconds, so this means 1000W, which is 15.8A. If you are going to be routinely testing at 1000W bursts, it might make sense to change from the 8A fuses C1F8 to a 12A fuse that can handle the repeated bursts at 16A. For example, here's a fuse with a 125VAC 15A continuous rating that can avoid errant trips.
Remember, too: Tripping a fuse in the speaker cable isn't something to be taken lightly. When a fuse trips, the di/dt becomes huge, and combined with the cabling stray inductance a massive voltage can be induced across the amp's output stage. This can easily kill the output MOSFETs unless the they avalanche rated and the amp designers contemplated something like this.
The attenuator settings are shown below. The attenuation you get depends on if the attentuator is operating in Unbalanced or Balanced mode.
| Position | BALANCED GAIN | UNBALANCED GAIN |
|---|---|---|
| POS A | 0 dB | 0 dB |
| POS B | -18.0 dB | -13.0 dB |
| POS C | -24.1 dB | -18.6 dB |
The resistors are rated for operation at 250C. However, that resistor is so large that it is much too hot to contemplate on a desktop. So, they max resistor temperature for safe operation has been limited to 70C. This is in line with temperatures consumers can experience around space heaters. There is a stick-on temp sensor on the upper resistor to help you keep tabs on temperatures. But it's up to the user to limit the temperature.
Actual attenuator measurements were made and are recorded below. Even though the attenuator board was built with 0.1% resistors, the measurements below factor reflect the 100 k input impedance of the analyzer.
| Position | BAL GAIN | UNBAL GAIN |
|---|---|---|
| POS A | -0.44 dB | -0.45 dB |
| POS B | -18.08 dB | -13.14 dB |
| POS C | -24.15 dB | -18.70 dB |
Using the NX3000D in bridged mode, with the load in Position C (gain = -23.70 dB), the NX3000D measured 1.34 kW of output power. The RMS voltage out of the amp was 73.35 Vrms (37.31 dBV) with the channel A gain at max. The atten knocked the 37.31 dBV down to 37.31-23.70=13.6 dBV, allow us to the use the +18 dBV input range. This gain was measured as 47.31 dB, which squares with the GEN1 output at -10 + 47.31 = 37.31 dBV.
If we back off 1 dB on the input level, the distortion drops an order of magnitude:
Reading some forums, it seems Behringer likes to quote peak power when they speak of power. So, take the reported average power and multiply by 1.41, which would take 1.34 kw to 1.89 kw.
Their manual also doesn't list a distortion we could expect at the peak. There are 3 LEDs for each channel. The first indicates -40 dB (signal present), the next is -6 dBV, then -3 then 0 dBV. At -8 dBV in the 0 dB LED will illuminate. This is showing 1.7 kW with 9.2% distortion.
The input to the NX3000D was adjusted to give 194W of power. The FFT size was set to 64 k and 48 ksps, yielding a 1.3 second burst with the default processing delay between bursts (around 200mS). The 70C temperature was indicated at 90 seconds, and 90C was indicated at 140 seconds.
The Arcol planar resistors are rated for 500 W each, and so two in parallel should allow momentary operation up to 1000 W. During testing of the NC3000D the levels went up to almost 2 kW, or 1000 W per resistor. It's not recommended to exceed the manufacturer's recommendation, of course. And in order to hit these power levels, the fuse will need to be upgraded.
A Hypex NCx500 class D was measured, as these are quite a bit more refined than the Behringer measured above.
Because the NCx500 drives the output + terminal and the output - terminal is held at ground, the load config was set to UNBALANCED and POS C (-18.74 dB of attenuation) was selected.
The QA403 was used in differential output mode. The L+ and L- BNC outputs were converted to RCA, and then an RCA to 1/4 Phono jack cable was used to connect to the NCx500 module via a 1/4" TRS to XLR. The actual unit used for testing was a Buckeye Amps NCx500 Monoblock. The Buckeye pre-amp was bypassed using the provided jumpers, and the Hypex module was set to "Buffered", which should provide between 26.6 dB and 27 dB of gain.
The QA403 settings to achieve the input gain (due to the attenuator) and the output gain (due to the use of balanced analyzer outputs) are both set in the dBV Content Menu:
With these settings made, we can measure the gain, which is a bit higher than the expected max of 27 dB:
We can disable the QA403 output and take an RMS measurement (unweighted, 20 to 20 kHz). We see the noise is 55 uV. The NCx500 spec indicates buffered mode noise at 20uV (typical). Since it's not indicated, we might assume that is unweighted. Note this figure is equivalent to -85.13 dBV, and with 27.7 dB of gain, the input referenced noise is -112 dBV. This is getting close to the output noise of the QA403 DACs, and so we want to dig a bit more here.
First, let's change the Load jumpes to POS A and update the input gain. The measured noise doesn't change, suggesting the measurement is valid.
Next, we can short the XLR Hot and Cold together (and optionally connect both to Chassis Ground). This causes a big rise in hum and noise:
So, we might take the 50 uV noise figure (unweighted, 20-20 kHz) as the actual measured noise.
Next, let's do a sweep of THD versus Frequency at 10, 100 and 300W. For this, we need to revert to POS C (-18.70 dB attenuation). For this, we'll use the PWR THD versus Frequency Automated Test.
Sweeping at 300W, 20-10 kHz and 3 points per octave yielded 27 test steps. At 64k FFT, each step took 1.4 seconds. The load was around 70C after the test had completed.
A link to the various design files, including gerbers and dxf, is located HERE.
It's important to understand the amp topology you are dealing with, as that has an impact on the attenuator you design. For balanced amps, you will need a balanced attenuator. And for single-ended amps, you will want a single-ended attenuator. Of course, in a perfect world there is no difference. But mismatching attenuator types can put a heavy burden on the analyzer and limit in your input options.