QA451B issue with external loads

[Mudjock]

I’ve used the QA403 with the QA451B for several years, measuring a wide range of receivers, amplifiers, and amplifier boards including car/mobile audio, home audio, diy, vintage, etc… and have generally found the capability and utility of the equipment to be extremely helpful. Once I gained confidence that my measurements were consistent with other published measurements for the same products, I began to publish some results in public forums, primarily Midwest Audio Club.

I have always known at some point I would want to test higher power amplifiers than that combo would allow and relied heavily on Matt’s article about using the 451 programmable load with external loads and his subsequent testing of a class D Behringer amp. I built a modular load based on the suggestions in that article. The load consists of 4 ea. 8 ohm “modules”. Each module is 2-100W Arcol HS100 4R f resistors in series. 1 module per channel for 8 ohm testing, 2 modules per channel connected in parallel for 4 ohm testing. I send voltage back to the 451 via the positive terminal on the load and a center tap between the two 4 ohm resistors to get an extra 6dB attenuation (as Matt’s article describes). Here is a picture of a module:

I ran my module array on an Extron XPA1002 plus, which had also been tested by Audio Science Review and got very reasonable agreement with those measurements either using the 451B by itself or with the external load. If anything, the external load measurements looked a little cleaner. Then I hooked up one of my big amps - a Samson SX 2400 rated to produce 550 W/channel into 8 ohms and 750 W/channel into 4 ohms. My initial ASR standard 1 kHz 5W 4ohm testing and frequency response looked reasonable. When I did the THD+N vs. Power sweeps at 1 kHz, I got quite a bit le

ss output than I expected:

My first thought was that the amplifier was either severely overrated or malfunctioning somehow (it is a class H design, so maybe the power supply wasn’t adjusting the rail voltages correctly). So, I dragged out my QSC CX902, which has similar power ratings and runs one of my subwoofers. I took my first power sweep at 4 ohms.

Nope… This looks like an almost identical max power as the Samson and not nearly enough. I went back to the drawing board with my load construction, wiring, settings. Nothing helped. Then, with much trepidation (now running 8 ohm tests so I only have to deal with one load module per channel), I decided to hook up my load (with only 6 dB of built-in attenuation) directly to the QA403. This sweep had to autorange and use the built-in attenuator extensively, so it isn’t as clean.

So now we are seeing the expected power output, but I don’t have the 451B low pass filter for class D amps or extra 12 dB of attenuation that would allow cleaner high power measurements.

Any ideas what is going on? I can build a new load with 32 ea. 1 ohm 25W resistors that would let me to get 18 dB attenuation on each channel… But I’m not thrilled to spend the extra funds on that when what I have theoretically should work and I won’t have the class D filter from the 451B, which would be nice for my upcoming IcePower 1200AS2 amp build and test…

[restorer-john]

@Mudjock

Samson SX 2400

I can think of two things that may be causing this.

Is the Samson a BTL/bridge output and have you added the additional 6db to the QA541’s 12dB (18dB total) in the QA software as an offset?

[Mudjock]

I did run the sweeps with a setting that had input gain of -18 dB. I don’t have a schematic of the Samson amp, but I don’t think it is BTL/bridge output. Per the spec sheet, it is rated for 500 WPC into 8 ohms and 750 WPC into 4 ohms and can be bridged for 1200 W mono into 8 ohms. The amp is described as Class H output circuitry.

I was initially suspicious of the amp, which led me to try the QSC CX902 and achieve very similar results. The CX902 is rated at 550 WPC into 8 ohms, 900 WPC into 4 ohms and 1500 WPC into 2 ohms. It has 2 tier class H topology according to their data sheet. That amp can allegedly put out 3000 W bridged mono into 4 ohms. I don’t see any evidence that the CX902 is a BTL amp.

[Mudjock]

The QSC CX902 is equivalvent to the DCA3022, which is the next higher model than the DCA2422 that Amir tested. They both claim the same topology.

[matt]

Hi @Mudjock,

It’s always helpful for me to start testing with just a tone and DVM to make sure everything is working as expected. For example, if I think my amp has 25 dB of gain, set the QA403 to -25 dBV output, connect the amp to the loads, set the QA403 software to a continuous tone via IDLE, and then probe the amp output directly with the DVM assuming balanced (that is, DVM across speaker + and -). And then repeat with DVM to + and ground, and DVM to - and ground. And then measure output of any attenuator. Note all those voltages down, and make sure everything is working as you expect. And then still staying working on a single tone, run a cycle on the QA403 and make sure that reading matches the DVM measurement across the +/- terminals.

Modern class D can have some very strange topologies (and lots of unexpected DC at the output–sometimes half the supply rail), especially as they move to a single rail. And then, you have the issue of balanced and unbalanced attenuators and mixing those with balanced and unbalanced amps. If you use a balanced atten on an unbalanced amp, it will work, but it won’t let you use the input range you might expect. See more at the link below.

Measuring Power Amps

[Mudjock]

Okay Matt - I respect the request for some basic troubleshooting data. Here you go…

I set the Samson SX 2400 to maximum gain, which I confirmed is approximately 42V at the 5W test point I typically use to align with Audio Science Review for comparison purposes.

Next, I set the generator to -42 dBV and turned on the voltage readouts on the graph display.

I also inserted a voltmeter where I make my amp connections across the left channel load.

11 mV difference (about 1% of reading). This looks okay and my meter would not be considered calibrated.

Next, I measured the voltage at the left channel output of the QA451B

There, we are reading 0.126V, which calculates to -17.993 dBV. I believe that is close enough to confirm the basic setup is operating as intended.

[Mudjock]

For clarity, it is worth noting that neither of the amps mentioned in my original post are class D.

[Mudjock]

Next step is higher power testing of the same. 100 Watts per channel into 8 ohms.

Here is the same display we did at 5 Watts per channel and 0 dBV:

Based on the power sweep graph from the first post, we are already starting up the steep distortion slope (with increasing power/voltage). The voltmeter across the load agrees reasonably well with that shown on the graph.

Coming out of the 451B, we also get what we expect with -18 dB of attenuation:

[matt]

Thanks for the data, @Mudjock. Are you sure the amp gain is 42 dB? That’s a lot, and the first spec sheet I can find suggests that 34 dB is expected.

Also, I note in the photo you provided that it looks like your output is hooked up balanced, which means you are delivering to the amp 6 dB more than you think. On the QA403, because there’s not a dedicated BAL and UNBAL port, you have to indicate if you are making a balanced connection via the DBV context menu and the Output Gain textbox. Note there’s a button to preset 6.02 dB, which is what you’d want when making a balanced connection. And once you do that, you’ll see this appear on screen:

That means that when you indicate, say 0 dBV, the actual output will be -6.02 dBV on L+ and on L- (ref to ground aka BNC shell). But if you measure across L+ and L-, you’ll see exactly 0 dBV.

So, in your first picture, your shown output is -13 dBV, but if the connection is balanced then the actual output level is -7 dBV. And so your gain would really be 42-6 = 36 dBV, which makes a bit more sense.

Agree your outputs look as expected.

The 6 dB discrepancy will explain errors with gain calcs. But as you demonstrate, with 100W out (28.32Vrms), your THD is very poor. The SX2400 appears to be able to work in balanced mode, but presumably for this test you are not in balanced mode.

The spec on this amp is 0.03%/-70 dB distortion, and so at 5W with 3H at -80 you are probably close to that. But at 125 mW out you are probably falling short of the spec based on 3H.

Additionally, you have a lot of 60 Hz running through all the measurements, rendering your THD+N not good.

In the 100W measurement, your 2H and 3H are both right at -70 dBc, so your THD is probably around -60 dB = 0.1. You should add a THD tile. But the 2H and 3H are dominating here, so your THDN is probably representative.

Just curious, but what do you expect to see for THD at 100W? Spec says 0.1% at rated power.

[Mudjock]

Thanks for providing your insight Matt. I wasn’t aware that I needed to set a different output gain when running a balanced output to get an accurate gain measurement. I reran the 100W, 1 kHz test tone with that corrected. I agree the gain makes more sense now.

You are also correct that this amp is capable of running with balanced output (bridged mode).

With regard to my expectations, this is a pro audio class H amp (linear power supply with a big toroid) design more for sustained high power than minimizing distortion. My experience and decades of looking at amplifier measurements suggested that I would have some AC noise (60 Hz plus harmonics) for this type of amp and the 60 Hz noise sitting around -90 dB seems reasonable at 100W per channel. THD I would be expecting to be primarily in the -60 to -80 dB range for this amp. Many times, class H amps will start to produce increased distortion once the rail voltages start to change at higher power levels to prioritize efficiency. The real issue from a distortion perspective isn’t what we see at 100 WPC, but rather what happens as we move up from there. By the time we are a couple of dB up from this, THD is way up to about -20 dB or 10%. I would not expect to see that further increase until at least 600 WPC or so.

[matt]

Hi @Mudjock, yes, that looks much better! Generator is at -7 dBV, gained by 36 dB of gain to 29 dBV = 28.13 Vrms = 99.29W into 8 ohms. It all makes sense. So now with the single tone working, you can run the sweep you were after before and it should make sense.

Agree on the bnc confusion. If there were more panel space and a higher price point, it’d be better to have XLR out for balanced and a nearby BNC for unbalanced. And when you specify “0 dBV” both the balanced and unbalanced have the same level. But space and cost constraints dictate the shared connectors and so it does require a bit of extra thought when connecting.

Yes, agree on 60 Hz too. Usually, when the supplies start to sag (due to increased current demand) you’ll see the 60 Hz as grass way up in the 1 kHz region and beyond. And sometimes even more pronounced IMD where the 60 Hz products are shown off of each harmonics. Below is a plot from @VAR’s measurement of MC2300 at around 300W and you’ll see the same phenom. And yes, agree, when the grass is ~100 dB below the RMS it won’t be perceived.

PS. What would be cool with the class H is to use the amp dynamics plug-in and see how quickly it can go from, say, 10W listening power to 500W plus. What does it do with a 10mS bump, 50mS, 100mS, etc??

Using the QuantAsylum QA40x (pt2) to measure a McIntosh MC2300- Vintage Audio Review Episode #115

[Claudio]

@matt

…. Agree on the bnc confusion. If there were more panel space and a higher price point, it’d be better to have XLR out for balanced and a nearby BNC for unbalanced. And when you specify “0 dBV” both the balanced and unbalanced have the same level. But space and cost constraints dictate the shared connectors and so it does require a bit of extra thought when connecting.

Hi @matt. I completely agree with what you say. However, I think that if in the QA40x software, in the “Generators” section next to the “GEN1” and ‘GEN2’ buttons, there was a small button “Use gen in balanced mode” (or if there were a checkbox in “Signal Generator Setting”) that, when pressed, would automatically set the output gain to +6.02 dB, this would greatly help to avoid misunderstandings and trivial errors due to simple forgetfulness. Of course, this is just a suggestion. Thank you.

[matt]

Thanks @Claudio, it is a good point.

[Mudjock]

I also like @CIaudio ‘s suggestion.

Let’s get back to the issue that started this thread. We’ve now traced through my setup and everything seems in order. The 100W into 8 ohms measurement showed some increasing distortion when the voltage signal from the external load is running through the QA451B. I’m not sure this really outside of expectations though. The real evidence that something is amiss comes when I try to drive the amp a little harder. Below is 2 dB up from the 100W results.

Distortion is now at the point I am essentially at the ceiling of what I can work with. If this is indicative of amplifier performance, then the amp is signficantly underperforming. I really wouldn’t want to report that publicly without being quite sure, so that led me to do some testing with the QSC CX902 (see THD+N vs. Power level graph from the initial post), try different wiring connections, try different load resistors, etc. None of those made a difference. I still couldn’t run at 175 W per channel into 8 ohms without seeing massive distortion. I even confirmed with a voltmeter that less than 5 VRMS is being output from the QA451B to the QA403 at this test point.

At this point, the only reasonable thing to try was to take the QA451B out of the loop.

I do have an observation about that. The output of the QA451B is a single BNC connector for each (left and right) channel. Those connect back to the left and right positive inputs on the QA403. I tried to mimic this when connecting my external load and the gain was 6 dB too high. When I connected the external load using both the + and - inputs for each channel, the gain returned to normal… I only have 2 bare wire to bnc adapters laying around, so the measurements from here will only include the left channel (right inputs have 50 ohm shorting caps).

Once this was sorted, I replicated the 0 dBV test point.

There’s maybe a few 10th’s of a dB more noise than with the 451B, so probably close enough. Next, let’s use the more common 5W test point.

This still lines up pretty well with expectations and what was seen with the 451B in the loop. The 8 ohm load is two 4 ohm resistors in series with a center tap to give 6 dB attenuation, so about 3.2 VRMS is going into the QA403. If we’re going to test high power this way, we’ll have to use the 403’s built in attenuator. That’s what I did for the 100 W test point.

Even though we are using the attenuator, THD+N is about 10 dB less than it was when we ran through the 451B. The next test point of approximately 175W will be interesting because when I ran that with the 451B in the loop, we were up around 10% THD+N.

Now that we’re not running through the QA451B, the amp is measuring as expected. Less than 0.03% THD or THD+N. About 18 VRMS is going into the QA403 input. The next step, of course, was to increase the input further.

Even less relative distortion and noise at 320W.

At 500W the distortion is finally setting in, but that’s about what one would expect given the 500W power rating into 8 ohms. Just under 32 VRMS is running into the QA403. We probably wouldn’t want to give it much more. I used the Oscilloscope visualizer to get a little more insight.

This looks like standard clipping at about 87V peak. That calculates out to about 473 W RMS into 8 ohms as the clipping point.

So, one would conclude from this that the amplifier is performing within expectations and we can see that when running an external load directly into the QA403. So the real question is what is going on when the QA451B is part of the measurement circuit? There is clearly some limitation (at least with my unit or test setup) that is preventing high power testing with this amp (and the similarly spec’ed QSC amp).

[matt]

So the real question is what is going on when the QA451B is part of the measurement circuit?

The answer is simply that the QA451 doesn’t have an infinite input voltage range. It runs from +/-15V rails. Now, your are pushing 63.25V and with an 18 dB attenuation factor that is 0.125, and so the 63.25 * 0.125 = 8Vrms, which is 11.22V peak. So, at 63Vrms you are starting to bump up against the 15V rails when the OPA1612 input voltage range is taken into account. Also, there are two important notes in the product specification. The first is the maximum input, which is 40Vrms, and the second is the absolute max input, which is 60Vrms. At some point, the input attenuation resistors will have their power handling exceeded and could open. I think the 60V limit reflects that.

Also, something to be aware of (though you might not be facing it here) is that any time an attenuator is used, you need to be aware if the topology is balanced or unbalanced. See the note in the section titled “Balanced versus Unbalanced (single-ended)” at the link HERE. This explains how an attenuator that provides a lot of attenuation can still output balanced signals that will exceed your expectations (though the signals are still correct). Just something to be aware of.

PS. I don’t think I’ve yet seen a Class D amp that actually needs the QA451 filtering. I’m sure they exist, but the QA403 antialiasing is very good–typically > 100 dB. Much of the discussion on the need for brickwall filters in Class D measurements came from 25+ years ago when data converters were much more primitive and relied heavily on external anti-aliasing circuits.

[Mudjock]

Hi Matt,

I’m not sure I understand your point here. Footnote 7 says that if I exceed 32 dBV (40 Vrms) input, then I will see increased distortion due to input overload. I started to see increased distortion when the Samson amp was putting out approximately 28 Vrms and half of that (-6 dB attenuation) or 14 Vrms was sent from my external load to the 451B. Even at 500W, I am sending a little less than 32 Vrms to the 451B. When one couples that with your linked blog post on the QA451 product page that you published in 2020

Using the QA451 with External Loads

your current characterization of the situation seems to conflict.

I can appreciate that you are taking the effort to continue to learn and share your learning, but I would recommend that you clean up some of this documentation before releasing the successor to the QA451B.

With respect to my situation, it appears I will need to find 12-18 additional dB of attenuation or maybe just skip the center tap between my 4 ohm resistors and attenuate 18-24 dB across the entire 8 ohm modules. The easiest and most cost effective way to do that will be to put together a balanced/unbalanced voltage divider (from what I understand, that will need to consist of 3 resistors that won’t see a lot of power, but will be able to handle some voltage drop without introducing measurable distortion. Any recommendations for resistor options that will serve that purpose well?

[matt]

Hi @Mudjock,

I started to see increased distortion when the Samson amp was putting out approximately 28 Vrms and half of that (-6 dB attenuation) or 14 Vrms was sent from my external load to the 451B.

The SX2400 has three output modes of operation: Bridge, Parallel and Stereo. What mode are you using? All of the stuff below assumes you are using the SX2400 in Stereo mode (unbalanced). Ignore if not…

The link above to Measuring Power Amps shows you how you can run out of headroom quickly if you are measuring an unbalanced amp with a balanced attenuator (as in the QA451). Look at the spice simulations to see how a 24 dB attenuator on a 100 Vrms source can result in the post-attenuator inputs swinging at unexpectedly high levels (26.48Vrms and 23.4Vrms) even though the true diff measurement is 24 dB (~3V).

I can appreciate that you are taking the effort to continue to learn and share your learning, but I would recommend that you clean up some of this documentation before releasing the successor to the QA451B.

But I think the link that showed the NX1000 measurement was correct, and walks through the issues of overloading the QA451…What is in error or confusing?

Any recommendations for resistor options that will serve that purpose well?

I really like the Arcol planar resistors. A very economical and forgiving load that can handle both balanced and unbalanced amps is covered at the end of the Measuring Power Amps link. Schematics, part sources, gerbers, adjustable attenuation, temperature considerations…the attenuator was pushed up to almost 2kw on the Behringer NX3000D, and the same setup achieved credible THD measurements on the Hypex NCx500 at 10W, 100W and 300W, with best-case THD approaching -120 dB.

And you can see the NX3000D hitting 1kw at -72 dB THD, and 1.7kw at -20 dB THD (10%).

The gerbers and dxf files are all there. I really think a load like this is a very good place to start if you wan to push power. The QA451 can do it, but note the NX1000 in the link was in balanced (bridged) mode. If you want to drive it from an unbalanced source, a lot changes.

[Mudjock]

@matt

Any recommendations for resistor options that will serve that purpose well

Hi Matt,

I was asking more about the attenuation resistors than the load resistors. These are the load resistors I am using. I can use the center point between the resistors to get 6 dB of attenuation if I need to. I use one of these modules per channel for 8 ohm testing and two in parallel for each channel to form a 4 ohm load.

The Arcol HS100 4R f has 1% tolerance and impressive power handling when mounted to a heat sink with thermal compound, which I did. I was a little concerned about inductance (i.e. should I go with the NHS series). The inductance I actually measured was very low for the standard HS100. It helps that the resistance values are very low. All 4 modules measured very close to this.

I would choose something more like your planar resistors if I was building the successor to the QA451, but without that space constraint this setup, per the data sheets, should be able to take full power from amps like the Samson for close to a minute.

For attenuation, I quickly picked up an assortment of 2W metal film resistors and chose 4.7k and 680 ohm values to try out a single ended voltage divider that provides approximately 18 dB of attenuation. I am running the Samson in stereo mode since I use the amp as a primary source for DIY speaker gatherings, so that’s how I want to test and report it. I’ll share some results from that in a separate post. I may eventually mount them to something more permanent and add jumpers or separate connection points to switch between balanced and single ended configurations.

[Mudjock]

Now, for the amp testing results to date. Here is the setup - no fancy hardwired or board mounted test fixtures at this point. Still in the trial and error stage.

My first standard test is a “dashboard” that compares to how Amir from Audio Science Review tests. Not looking to debate the merits of this, but he’s tested a lot of equipment using this methodology, so I find it useful to do a more-or-less direct comparison. Quite a bit of noise relative to some amps, but THD and noise are relatively well balanced at this 5W into 4 ohm power output.

Next, I measure frequency response. No real surprises here. The data sheet indicates -3 dB points of 5 Hz on the low end and 60 kHz on the high end. My measurements wound up at 4 Hz and 50 kHz respectively. FR doesn’t vary significantly with load, which is what one would expect for a Class H amp.

Next are the THD+N sweeps at 4 and 8 ohms.

There are a few things to discuss here.

• The stairsteps up at around 100W for the 8 ohm curves and 150W for the 4 ohm curves happened about the same time a relay clicked on in the QA403.
• The spec sheet indicated 0.1% THD at 500W into 8 ohms and 750W into 4 ohms. We were more like 400W into 8 ohms and 600W into 4 ohms for this test. I did investigate that further as I suspect this test was running into limits of available AC power at 115 V in a normal residential circuit.
• The other spec sheet info was THD of 0.03% at 10 dB below rated power. That would correspond to around 50W into 8 ohms and 75W into 4 ohms. In that range, we are around 0.02% or less for THD+N, which looks good.

When I only test the left channel, we get more power.

Now we start to see the primary distortion rise at 450W into 8 ohms and 700W into 4 ohms. The spec sheet 0.1% THD limits of 500W and 750W respectively now look very close to what I measure here.

Next, I do THD+N sweeps into 4 ohms at different frequencies.

There shouldn’t be too much of an increase in distortion at higher frequencies for a class H amp, and we see pretty consistent performance here. The 30 Hz power output is silightly reduced and the amp made some funny noises during that sweep.

Another way to look at frequency dependence of amp performance is multitone testing. Here is an attempt at that.

At low frequencies, the noise floor is about 95-100 dB down from the test tones, which can be interpreted as about 16 bits of resolution (redbook CD quality). At higher frequencies, the noise floor stays around 85 dB down, which corresponds to about 14 bits - still better than many class D amps in that regard. One can also see some of the discrete noise bands sticking up as peaks above the noise floor.

I think this is about what I would have expected to see from the SX 2400. It isn’t the last word in performance, but is a workhorse that can deliver as much power as one can expect out of a single 115V circuit. I’ve included an excerpt from the data sheet below to illustrate the point.

The next lower powered model is the SX 1800, which can draw a max current of 19.5A and is rated at 400W into 8 ohms and 600W into 4 ohms. Since I’m running off a 20A circuit and couldn’t get more output than that when running both channels, it seems to make sense.

I’m still curious about the stair steps in the THD+n vs. power curves. I did a run in which I changed the larger attenuator resistor to 10k ohms to give approximately 24 dB attenuation and that did not help.