ASR dummy load configuration

[Langston Holland]

Hofer's two-part article was wonderful, thank you. I knew about it years ago but never bothered to read it because the title sounded like an AP brag to me and I was already a fanboy. : )

In my case, I don't think it's possible that better resistors than the ARCOL's I have would be helpful. This is only because my setup already has lower distortion levels of any type than my analyzer can detect. If I had a APx555 that would be different. Hint, hint! : )

Interestingly, Hofer's VCR math confirms the 1966 paper that Amir referenced in that the 3rd harmonic is the tell-tale for the non-linear distortion I experienced, Hofer adds (b) and (c):

(a) Resistor voltage coefficient
(b) Resistor power coefficient
(c) Resistor thermal modulation under 200Hz.

Each of these occur with the voltage modulation, thus dynamically, thus violating the LTI (linear, time-invariant) requirement for measuring anything.

In my case, I have more than dealt with longer-term temperature coefficient problems through overkill heat transfer. It's of little use to buy resistors with lower TC's if they aren't allowed to heat up in the first place. Thus these kinds of thermal affects aren't an issue in my application, what IS important is that the dynamic distortions are kept well below levels present in the DUT's and analyzers.

This is happy news in that one apparently can just keep an eye on the 3rd harmonic and keep the resistors cool.

A final observation is that even the best audio amplifiers show orders of magnitude more distortion than the worst dummy load resistors I've used when approaching max output.* This whole VCR thing came to me like it apparently came to Amir, by accident. In my case I bought the larger 200W versions of the otherwise identical 100W resistors from Parts Express and immediately saw much higher THD (noise not included, it was a continuous sweep) at very low voltages. I reconnected the 100W version and the distortion was back down to "normal". Then I used a smaller high quality resistor at the same 4Ω and saw additional reduction in THD. That's when I started searching for answers and Amir provided the solution - and you helped a lot as well with the Hofer article. : )

PS: Another VCR paper I found helpful.

God bless you and your precious family - Langston

* Edit (23 Sept 2021): this isn't true anymore. I just measured a 4Ω dummy load by paralleling (3) 12Ω water heater elements and it didn't go well. The elements showed increased 3rd harmonic distortion, particularly at high frequencies. See this post and the next for details.

---

Edit: Below is something I made years ago out of (4) 8Ω 300W air cooled wire wound resistors. It allowed 2-channel 4Ω testing which was all I needed to deal with at the time. I threw it away a couple of weeks ago because it is strongly affected by VCR even at 1V drive levels. That's 1/8th watt per resistor over a 2 second sweep. This takes Hofer's (b) and (c) non-linear distortion types out of the running. There's just no power or heat happening.

 

[Langston Holland]

A final thought:

Those of you interested in a switched dummy load using 8Ω resistors, (which is a bit more logical because it doubles the number of resistors as you halve the load), see the following schematic for an idea. It uses a much easier to source On-Off-On DPDT switch for each channel. Just like the resistors, high current and quality is a must for switch selection. You can of course use series and/or parallel arrays of resistors to achieve the 8Ω's, just be sure to plan adequate heat sinking for your application.

God bless you and your precious family - Langston

 

[audio2design]

Here is an experiment if you are willing. Go to Mouser.com and purchase 4 Vishay 16 Ohm 50 Watt wire wound resistors and connect them in parallel for a test load of 4 Ohms and a TCR of 20 PPM / C. (That is maybe $30.00 in parts) 71-RH50-16

I suspect that you will see an improvement over your current measured 4 ohm 100 PPM / C load resistor.

Thanks DT

I will put money on it will not make much if any difference at all. I will even go with the current setup of 100ppm resistors will be better.

The Arcol used by the OP is rated for 1200 watts and have a commensurate thermal mass. That compares to a 200W combination with commensurate lower thermal mass. 1200/200 * 20/100 = 1.2.

Now obviously that is a simplistic analysis, but less simplistic than the claim the 20ppm will be better.

The 20ppm may be better or it may be worse. It all depends on the design and thermal mass of the actual resistive material, and the time constant for transport into the surrounding medium. There is a temperature and a time based factor here. 200W of 20ppm may be better at low frequencies, but worse at high. You just do not know.

 

[audio2design]

And this is all moot since any real world speaker is orders of magnitude less linear than a resistor bank

 

[DualTriode]

Hello,

Someone finally gets it.

5 watts produces right at 17BTU’s. Depending on thermal mass a larger heavier resistor will have a smaller increase in temperature. But it is not only that simple. What matters is the mass of the resistive element and how fast the heat transfers to the ceramic core and aluminum fined body of the resistor.

I have distortion tested many resistors, among the worst are Vishay Naked Foil. Despite perhaps the lowest VCR and TCR they are very light weight. Naked Foil resistors heat and cool very rapidly resulting is very high thermal modulation and distortion.

Back to 300 Watt resistors with 100 PPM / C and 50 Watt 20 PPM / C resistors. If you place multiple 50 Watt 20 PPM / C in series and in parallel the 20 PPM / C resistors will outperform the larger heavier single 100 PPM / C.

For grins I have resistors to test the above being delivered today; several 50 Watt 16R 20 PPM / C resistors to wire up in parallel and a single 300 Watt 100 PPM / C resistor.

I will post 5 Watt 4 Ohm test plots later.

Thanks DT

Perhaps I have lost the bet already, I paid for the resistors.

HaHa

 

[DualTriode]

And this is all moot since any real world speaker is orders of magnitude less linear than a resistor bank

In the real world of listening to music, yes the speaker drivers produce more distortion.

Around here we are required to keep mother's milk pure. We test amplifier noise and distortion with resistors.

Thanks DT

 

[DualTriode]

Hello,

Friday afternoon there was a box of resistors from Mouser.com on my door step.

I ran some test plots with the 300R 4 Ohm resistor. The thing weighs a pound and is bigger than my fist.

The amplifier is the same AHB2. The test speaker cable is the 6S11 Canare with the SpeakON plug.

I am surprised by the high frequency FR droop. most likely due to inductance.

There in an increase in 2nd HD 3rd HD, 5th HD and 7th HD with this 300R resistor over the last test with 8 Ohms.

The noise looks much the same as the last 8 Ohm test.

The SINAD is reduced.

See the attached plots.

Thanks DT

It will take awhile to solder the test leads on the new 16 Ohm test resistors in the morning.

 

[Langston Holland]

I am surprised by the high frequency FR droop.

I was too until I thought about the DF plot in the AHB2 manual:

Your resistor is innocent - the "problem" is the amp's increasing output impedance at high frequencies. Though I have the non-inductive "N" series and you may not, the droop is due to voltage division (the way passive attenuators work).

The following measurements were made at the drive level required to dissipate 5W into 4Ω. 5 Second continuous sweeps, 192k 24bits, 2Hz-80.1kHz.

Benchmark AHB2 No Load vs. 4Ω Load (2dB Vertical Scale)

Modified Hypex NC400 No Load vs. 4Ω Load (2dB Vertical Scale)

All (4) Measurements Normalized to 0dB at 1kHz (10dB Vertical Scale)

If you're interested in the modifications I made to my NC400's, see the following links. The mods made the NC400 an ideal amp for the bench, while I prefer the AHB2 for how it sounds (shoot me now!). : )

NC400 Circuit Mod
NC400 NL4 Mod

God bless you and your precious family - Langston

 

[DualTriode]

I was too until I thought about the DF plot in the AHB2 manual:

View attachment 154113

Your resistor is innocent - the "problem" is the amp's decreasing output impedance at high frequencies. Though I have the non-inductive "N" series and you may not, the droop is due to voltage division (the way passive attenuators work).

The following measurements were made at the drive level required to dissipate 5W into 4Ω. 5 Second continuous sweeps, 192k 24bits, 2Hz-80.1kHz.

Benchmark AHB2 No Load vs. 4Ω Load (2dB Vertical Scale)
View attachment 154128

Modified Hypex NC400 No Load vs. 4Ω Load (2dB Vertical Scale)
View attachment 154129

All (4) Measurements Normalized to 0dB at 1kHz (10dB Vertical Scale)
View attachment 154131

If you're interested in the modifications I made to my NC400's, see the following links. The mods made the NC400 an ideal amp for the bench, while I prefer the AHB2 for how it sounds (shoot me now!). : )

NC400 Circuit Mod
NC400 NL4 Mod

God bless you and your precious family - Langston

Hello,

Thanks for the insight and post.

Not finished yet.

Not sure yet if the resistor is guilty, benign or innocent. Big picture, a voice coil is likely far worse.

Low resistance wire wound resistors have lower inductance than higher resistance wire wound resistors.

First, I will plot the 50 Watt resistors in parallel.

There is not a "low" inductance or "N" version resistor in the box.

Thanks DT

 

[Langston Holland]

Just for (more) fun:

Amp output impedance from damping factor = Ω/DF = 8/40 (at 20kHz per the manual's plot) = 0.2Ω
Voltage division in dB = 20log(Rl/(Rl+Rs))
Rl = Load impedance
Rs = Source (amp output) impedance
20log(4/(4+0.2)) = -0.4dB, which is pretty close to the -0.3dB I measured (see the AHB2 plot "cursors" box on the bottom right).

The difference might be the additional feedback available with the 8Ω resistance John Siau used to load the amp, or the newer amps have a little better HF damping factor, or my measurement may be wr.., I mean wrooo.., wrooonn... The new amps must be better! : )

 

[DualTriode]

Hello,

This is with 4, 16 Ohm 50 watt wire wound resistors in parallel.

Between the 300 Watt resistor and the 4 16 Ohm 50 watt resistors in parallel I call it a draw.

Next impedance and phase plots.

Interesting at least to me the 300 watt resistor weighs 1 pound. The 4, 50 watt resistors weigh 1/4 pound total.

Thanks DT

 

[Langston Holland]

Interesting at least to me the 300 watt resistor weighs 1 pound. The 4, 50 watt resistors weigh 1/4 pound total.

Thermal inertia is a good thing here unless you compensate with heat sinks. And big and heavy impresses the girls! Well, maybe not..

More fun (can't tell I'm retired, can you?):

Coefficient of thermal conductivity

1. Aluminum 239
2. Copper 386
3. Silver! 419
4. Water 0.58
5. Air 0.024

Looking forward to your measurements. : )

 

[audio2design]

Thermal inertia is a good thing here unless you compensate with heat sinks. And big and heavy impresses the girls! Well, maybe not..

More fun (can't tell I'm retired, can you?):

Coefficient of thermal conductivity

1. Aluminum 239
2. Copper 386
3. Silver! 419
4. Water 0.58
5. Air 0.024

Looking forward to your measurements. : )

Now move the air or water

For distortion most of this is meaningless. It's more the thermal mass of the direct resistive element not all the things around it. A small mass thermal element couple to a high transfer system may allow high dissipation but be prone for comparatively high thermal modulation.

 

[DualTriode]

Hello All,

I came back to take another look see.

For the comparison between the 4 Ohm 300 Watt 100 PPM / C resistor and the 4 parallel 16 Ohm 20 PPM / C resistors it is still a draw. Not because they are equal but because we reached the limits of the AHB2 amplifier.

I ran the same plots with the Audio Precision APx1701 transducer test interface (amplifier, test microphone power supply).

The Frequency Response is very much improved, inductance questions disappeared.

SINAD is 117dB's plus.

Another day I will test the one pound wire wound resistor with the AP amplifier.

Thanks DT

 

[Langston Holland]

For the comparison between the 4 Ohm 300 Watt 100 PPM / C resistor and the 4 parallel 16 Ohm 20 PPM / C resistors it is still a draw. Not because they are equal but because we reached the limits of the AHB2 amplifier.

Nicely put. I thought I was jealous enough due to your 555. You have their 1701 as well. That'll eliminate challenges, if any, to your impedance measurements. It's possible, but tough to achieve equivalent impedance measurement accuracy without it. The 1701 uses a 0.1Ω current sense resistor on the amp's negative output(s) to make the measurement and still manages a low wide bandwidth output Z of 0.13Ω vs. the AHB2's approx. 0.20Ω at 20kHz. My guess is that they're paralleling the output of (2) amps to achieve this, which would explain the higher 16Ω minimum spec for two channel measurements. Of course AP lies about their specs - they are always better than published. : )

The 1701 is DC coupled, which means flat phase and group delay to 0Hz. The Hypex NC400 is also DC coupled, which is very unusual for an audio amp these days. DC coupling allows some interesting measurements, such as testing rub & buzz near a driver's Xmax with little acoustic output, which may also improve detection sensitivity. You just offset the driver with +DC steps then, then run the rub & buzz measurement, then repeat with -DC steps.

God bless you and your precious family - Langston

 

[DualTriode]

Hello All,

For the comparison between the 4 Ohm 300 Watt 100 PPM / C resistor and the 4 parallel 16 Ohm 20 PPM / C resistors it is no longer a draw.

Mass wins.

Using the APx 1701 at 4 Watts the 300 Ohm 1 pound resistor scores at -141ish dB's for 3rd HD and -139ish dB's at 2nd HD.

Thanks DT

see the attached plots.

 

[Langston Holland]

Mass wins.

Fascinating. It's amazing that at only 4W there is any kind of thermal modulation going on in (4) paralleled 50W resistors. It's also amazing that the 555 has the ability to detect such tiny harmonic voltages. Thanks for the input. : )

Meaningless trivia, but I think the 300W resistor is spec'd at 700g (1.54 lb.).

 

[DualTriode]

One part per million is -120dB’s. We speak of -120dB’s all the time, it kind of just rolls off the lips pretty easy. However, come to think of it -120dB’s is slicing the pepperoni pretty thin on the pizza. Your APx 515 will do just that. If you want to see resistor distortion with your 515 break out some carbon compound resistors.

The APx500 software will do things your ear/brain will not do.

Noise is random. The APx 500 software will go into the frequency domain and with 10 or 20 samples the noise can be averaged or removed from our FFT plots. What remains, where we see it, are the distortion peaks. The distortion peaks are not random; they will are prominent where they were previously lost in the real time domain noise.

Some engineers, Douglas Self being one, think of distortion in terms of mathematical formulas. @amirm mistakenly speaks of voltage caused distortion. For wire wound resistors I think of voltage squared divided by R * in my mathematical model of distortion.

Think of perfect sine waves or perfect cosine waves. Where the voltage, plus or minus, is peaking the resistive element is heating. Where the voltage is crossing zero the resistive element is cooling. The resistive element does not cool instantly. Think of it, at 1khZ the resistive element is going through 2000 heating cycles per second. That is the definition of thermal modulation. Yes this is what shows up as distortion on the FFT.

The higher the frequency the shorter the heating and cooling cycles become. At “high” frequency and short thermal cycles distortion levels attenuate.

Thanks DT

* Voltage squared divided by R is Watts, Watts is heat.

 

[amirm]

Some engineers, Douglas Self being one, think of distortion in terms of mathematical formulas. @amirm mistakenly speaks of voltage caused distortion.

I don't know who that is more insulting to, Doug Self or me. Either way, it seems you have lost the plot and keep thinking tempco is the same as VCR. Here is the AP article you linked to:

See the voltage dependency?

So leave the lecturing out until you understand the topic well.

 

[DualTriode]

I don't know who that is more insulting to, Doug Self or me. Either way, it seems you have lost the plot and keep thinking tempco is the same as VCR. Here is the AP article you linked to:

View attachment 154468

See the voltage dependency?

So leave the lecturing out until you understand the topic well.

Hello @amirm,

Not so fast.

Voltage Coefficient and Thermal Coefficient are different models for resistor caused distortion. No one is saying that they are the same.

We are speaking of wire wound load resistors here. Resistor Thermal or Power Coefficient is the much larger of the two.

From the left hand column of page 20 of the Hofer article you posted the formula for the Resistor Voltage Coefficient. You neglected to mention or quote, “In most common electronics, the VCr is safely assumed to be negligible”. Still keep your eye on it for ultra-low THD + N applications.

On the right hand column of the same page Hofer explains the Thermal or Power Coefficient. This formula is provided. Note that (V squared/Ro) or Watts is included. I did not make it up.

R(Vs) = Ro x [1 + TCr x Z (w) x (V squared/Ro)] *

* Voltage squared divided by R is Watts, Watts is heat.

As good as the amplifiers are that we are discussing they have too much distortion to test these models of resistor distortion.

So what I did was use resistors of higher value (1k Ohm) and used the APx555 generator for a voltage and current source. I used Mouser part number https://www.mouser.com/ProductDetail/Vishay-Dale/PTF651K0000FZEK?qs=QWYorlrD8IUh24FyOrKTdA== as the reference resistor. I used 3 resistors in series and 3 groups of 3 in parallel as my reference 1K Ohm reference resistor. This Vishay PTF series 5PPM / C resistor was the best that I found.

I put this 9 resistor network into a Wheatstone bridge with the other Resistors Under Test. The bridge circuit nulled the noise making it possible to measure very low levels of distortion.

No.

I do not think that Voltage Coefficient and Thermal Coefficient are the same.

Thanks DT