ASR dummy load configuration

[Armand]

I am interested in the spesifications of your dummy load amirm. You said you had to rebuild it for the AHB2 test but I cannot find any information about it in the thread (or other threads).

Also, why does distortion of the dummy load matter if you measure using kelvin connection?
When I measure low distortion amplifiers I have sometimes connected the measurement cable directly on the amplifiers PCB to avoid the extra distortion from cables and connectors.

Your thoughts regarding this would be much appreciated.

 

[amirm]

I always had the Kelvin connection so that didn't need to be redone. What had to be redone was the load resistors. You can buy 300 watt load resistors for $30 or whatever. Turns out these have high VCR: voltage coefficient of resistance. In simple terms, as the waveform changes in amplitude, the resistor modulates its resistance, presenging varying and dynamic load to the amplifier. This becomes dominant, causing third harmonic distortion.

A great paper on this is from 1966 with the title: Harmonic testing pinpoints passive component flaws. So the effect has been known for decades. The mechanism that causes it is complex though. Of note, it has nothing to do with heating of the load resistor as that happens over time. It has to do with micro-effect inside the load resistor at such high currents such as impurity of connections. Here is a starting quote:

Unfortunately, VCR is not a documented spec for load resistors even from high-end resistor suppliers. It is through experimentation the Dale power resistors have been found to have low-enough VCR. Alas, the 250 watt version is expensive at $180 each and you need four of them for 4 and 8 ohm stereo testing! Worse yet, the parts are simply not available. See this for example: https://www.mouser.com/ProductDetail/Vishay-Dale/NH250-400-1?qs=81LrLu1ylWBveqWvVGPieQ==

I lucked out finding an alternative from a UK brand, Ohmite/ARCOL: https://www.mouser.com/ProductDetail/ohmite/nhs300-8r0-1/?qs=T7nPz1OZfaCYYWtzcKdZcQ==&countrycode=US&currencycode=USD

I was worried they would not perform the same but turned out they did. Benchmark was kind enough to send me their Dale dummy load and mine and theirs perform the same. The difference in cost is that the Dale resistors are much beefier with a larger heatsink. So they can dissipate power more easily. I chassis mounted mine anyway so that acts as a heatsink.

4 OHm Dale resistors are available but they are even more expensive at $213: https://www.mouser.com/ProductDetail/Vishay-Dale/NH2504R000HJ01?qs=sGAEpiMZZMtbXrIkmrvidIpgiW9xqmWBjhwJWAFua5A=

Note that the VCR gets reduced by the damping factor of the amplifier. To the extent you have a high damping factor in the amplifier, you may get away with less. Or alternatively suffer from worse results with low-damping factor amplifiers.

As it turns out, I had used very high quality/precision Dale resistors in my headphone amplifier load. I thought at the time it was overkill as one does not need precision there but I spent the money on them anyway. That likely enabled me to measure headphone amplifiers with such low distortions.

The other contributing factor was the speaker connectors. Bare wire outperformed my non-name banana plugs. So I purchased locking, solid brass ones from parts express: https://smile.amazon.com/gp/product/B00O3NTB28/ref=ppx_yo_dt_b_search_asin_title?ie=UTF8&psc=1

The knobs on these a bit hard on hand but otherwise, they do a fantastic job of clinging to the speaker sockets. And are easily terminated with large hole for speaker wire and dual screws to hold it. It can accept wires as thick as 8 gauge.

I had used silicon wire with thousands of strands for my speaker wires. I need high flexibility in wires due to countless times I connect and disconnect the wires. John Siau from Benchmark suggested star quad cables. I swapped those for mine and found no difference. But there may be an issue with other wires.

The key here is that I had a platform to test all of these effects: Audio Precision Analyzer's ultra-low-distortion analog sine wave generator and Benchmark AHB2 amplifier with similarly low distortion. Without these, you wouldn't know if what you have is good enough or not.

The old setup by the way, would cap out at 105 dB SINAD and third-harmonic would not go lower than -110 dB. That is more than 20 dB shy of the final measurements. Here are sample measurements of AHB2 with my old load (ignore the low frequency components):

Compared to what it did with the final load:

Notice the difference in third-harmonic. Overall SINAD does not improve as much because it also has noise contribution.

 

[amirm]

By the way, these resistors come in inductive and non-inductive versions. The inductive version is much cheaper but I worry about its impact on measurements so I opted for non-inductive.

 

[Armand]

Thank you for an elaborate answer.
The measurement of AHC2 with the old dummy load seems to be missing?

I am aware of the VCR component of resistors as it comes into play when I am designing low noise buffers. Todays high current capable opamps allow us to use low value resistors in the feedback loop to reduce noise but at one point the VCR starts to add distortion. I prefer the MELF type as these seems to suffer less from the problem.

As you point out, the VCR problem of the dummy load depends on the utput impedance of the amplifier. If you measure for instance the 1ET400A modules directly at the amplifiers feedback point the voltage coefficient of the load would really not matter since the output impedance in that point is in the micro-ohm area. Don't you agree?

I built a 1600W 8/4 ohm dummy load using 32 cheap nomame 100W resisors from ebay mounted on a big fan cooled heatsink. While the power handling is superb I have also noticed increased THD if not measuring close to the amplifier output.

The reason I am having increased consern about this now is that I am evaluating the new 1ET400A modules. I am testing these in bridged mode and have so far been able to push them to 900W into 4 ohms with 0,0003% THD+N. While this is impressive I cannot free myself from the though that the load might add some distortion even if I measure at the feedback point.
I will try to measure closer to the load next time and see what happens.

 

[amirm]

Also, why does distortion of the dummy load matter if you measure using kelvin connection?

I forgot to answer this part. When measuring power amplifiers, there is a lot of current involved and as such, there is a voltage drop across even speaker wires. A kelvin connection samples the amplifier output at the terminal, rather than load resistors and as such, would be immune to this voltage drop. This will certainly impact measurements like output impedance. How much it would impact distortion measurements, I don't know because I have always tested this way.

Here is a sample measurement from AP showing the difference:

Connection at the amplifier is the Kelvin method. As seen, it shows higher damping factor by excluding the effect of the speaker wire.

 

[amirm]

As you point out, the VCR problem of the dummy load depends on the utput impedance of the amplifier. If you measure for instance the 1ET400A modules directly at the amplifiers feedback point the voltage coefficient of the load would really not matter since the output impedance in that point is in the micro-ohm area. Don't you agree?

I would guess so. Indeed with Hypex NC400, I did not run into this problem either.

 

[amirm]

The reason I am having increased consern about this now is that I am evaluating the new 1ET400A modules. I am testing these in bridged mode and have so far been able to push them to 900W into 4 ohms with 0,0003% THD+N. While this is impressive I cannot free myself from the though that the load might add some distortion even if I measure at the feedback point.

I had a box of these no-name resistors that I tested. Some worked better than the others and indeed that was the clue I needed to realize the load matters.

From my point of view, my measurements have to be solid and without questions so I had no choice to go with lesser options.

Overall THD+N is dominated by noise in 1ET400A so I don't think it will matter as much as it did with Benchmark AHB2 that was so quiet that distortion was dominant.

 

[amirm]

One clue that you may have an issue is if you don't get the same results out of both channels:

With my old dummy load, I would get a different distortion curves between channels. Those differences have gone away with the new load (for high-performance amps).

 

[amirm]

Thank you for an elaborate answer.
The measurement of AHC2 with the old dummy load seems to be missing?

Oh, let me fix that.

 

[HammerSandwich]

The other contributing factor was the speaker connectors. Bare wire outperformed my non-name banana plugs.

FWIW, Brent Butterworth tested a bunch of bananas - including measurements! - at Wirecutter.

 

[amirm]

FWIW, Brent Butterworth tested a bunch of bananas - including measurements! - at Wirecutter.

Thanks but he is just measuring resistance. That kind of testing is very unreliable anyway.

The issues at hand here are distortions caused by the connector! Very different thing that he did not measure.

In addition, locking plugs is absolutely necessary as there are variations and some are barely making connections.

 

[Armand]

I have stopped using bananas when testing amplifiers because I have had too many problems with them. In my opinion the very best connection is fresh copper screwed directly to the terminal. When testing high power amplifiers the current can easily pass 20A and I trust no bananas with those currents. I have however not tested the best locking types, and I agree the locking must be a minimum.
I have tested terminals also and found that rhodium plated copper terminals gave the best results. I switched 300A through them with a MOSFET and measured the voltage drop. We are talking micro ohms here but still...

 

[Rja4000]

I have stopped using bananas when testing amplifiers because I have had too many problems with them.

What about Speakon, by the way?

 

[Armand]

Speakon is rated at 40A and has acccording to the spesifications less than 2mOhm resistance so it is a reliable and good connector.
I have not measured speakon, but my test of copper cable directly to the terminal shows less than 0,1mOhm resistance.
This is Rhodium. The resistance from 8mm^2 copper cable to the terminal was 65uOhm and the resistance through the terminal itself was 9uOhm. A total of 139uOhm from copper in to copper out.

The test setup was like this:

 

[amirm]

What about Speakon, by the way?

On Benchmark AHB2, I could not match the performance of Speakon with the banana jack. John told me the speakon are soldered to the amps so has a shorter path and less moving parts compared to the banana jack.

 

[restorer-john]

This will certainly impact measurements like output impedance. How much it would impact distortion measurements, I don't know because I have always tested this way.

When measuring output impedance for a damping factor calculation, the currents are low. Damping factor was usually measured/specified at 1W, not sure what you use. (5W maybe?) Trouble is, at 1W, the voltage differences are so small between the various loads, that the DF or Rout numbers calculated are inaccurate.

At high powers, the non-linearities of the load and/or speaker wire connectors will come into play and also skew the results.

The AP method using loaded and unloaded voltages vs two distinct valid loads (8/4) can yield quite different DF numbers again.

 

[March Audio]

@amirm seeing this thread reminded me I was meant to dig out my Blue Jeans locking banana plugs to show you.

They are ultrasonically welded. The collet screws down splaying the banana pins for tight connection.

 

[amirm]

Since I have to make my own cables for the Kelvin connection, I can't use pre-made cables. I could buy them and cut them up but it seemed like a waste of money to do that.

 

[March Audio]

Since I have to make my own cables for the Kelvin connection, I can't use pre-made cables. I could buy them and cut them up but it seemed like a waste of money to do that.

Blue Jeans will make custom cables I believe. I am particularly impressed with their ultrasonic welding of the cable. I would guess thats as low resistance as can be obtained.

 

[amirm]

Blue Jeans will make custom cables I believe. I am particularly impressed with their ultrasonic welding of the cable. I would guess thats as low resistance as can be obtained.

Again, the resistance in a dummy load doesn't matter. We don't care of the load is 4 ohm or 4.01 ohm. What we care about is whether there is a characteristic that changes with voltage. Maybe electronic welding is good. Maybe it is not.

Also, I connect and disconnect these cables hundreds of times. I routinely break cables and have to buy new, or make new. With a welded connection I can't fix it if it breaks.

So for everyday use they may be excellent. Testing though is a different matter.