Rob's Amp tests

[peniku8]

I built a loadbank to bench-test audio amplifiers on a resistive load.

The loadbank consists of 16 3KW 220V heating elements, which allows me to test any combination of 8 8Ohm clusters for up to 4 channels.
More on the loadbank in this post.

 

Results:

• Sanway FP-13000 (MKI)
• t.amp TSA-1400
• t.amp TSA4-300

 

Testing methodology:

1. Calibrating the signal chain without an amp in the loop
2. Getting a frequency response measurement of the amp
3. Determining the amp's maximum unclipped output voltage (Vmax@1khz) with no load connected
4. Adjusting my audio interface's input gain to scale the input signal at Vmax to -12dbFS (optional, just for convenience and consistency)
5. Doing a reference recording at Vmax@1khz (any other known uncompressed output voltage would work as well) with no load connected
6. Feeding the amp various sine waves and recording the maximum signal levels generated into various loads
7. If clipping (>0.5%THD) was observed during step 6: Scale back the signal level in 0.1db steps until the THD drops below 0.5%
8. Calculating the power into a given load with the level difference between the calibration recording at Vmax and the under-load recording Vmax'

 

To 3:
I used REW to do this. I sent the amp a 1khz sine wave and monitored the signal return with REW's RTA. I increased the level until THD exceeded 0.5%, then scaled it back by 0.1db. For amps with built-in soft limiters I approached the maximum possible outut voltage for the minimum possible input voltage in 0.1db steps.
The generator in REW fades in, which results in 3.5 cycles (@1khz) being between -6db and 0db. That is equivalent in power to 1.75 cycles at 0db, which is why I generally add 2ms to the duration charts for the results. Not that it really makes a difference, but hey, I'm doing it.

 

To 5-7:
I record all the signals with Cubase 10 Pro. Vmax is determined through the Pmillet's voltage meter readout. I compared it against my HoldPeak TrueRMS meter and both were always within 1%. You can see my test environment in Cubase below. I checked THD of the recordings with MathAudio's THD Meter and noticed that as soon as the sine wave started looking clipped, it approached 0.5% THD. A very convenient way to quickly see if I need to reduce the output for a certain load (I did so in 0.1db steps if needed). I check all results with the THD meter notheless.

To 8:
I export the recorded files and open them in Adobe Audition. Audition can display peak levels under the cursor, which makes my life much easier.
I use Microsoft Excel to do the calculations for me and plot the charts with. I feed it the reference dbFS(RMS) value and its corresponding RMS voltage, the recorded db value and the load it was connected to. Excel calculates the generated power and plots it over the user defined time axis.
This is what my excel table looks like, with the Sanway FP-13000 into 1x4Ohm and 1x2Ohm as an example:

Everything in green is user input for easy navigation.
Power numbers posted will be the 100ms value. If any value within the 10s window deviates from that value by more than 5%, the power over time graph will be posted as well.
Default time stamps are used when the changes are linear (left example above). I will adjust the time stamps when the default time stamps wouldn't represent the actual power over time plot accurately (right example above).

The whole setup looks like below. The supply line is 6mm² (9AWG) 3x230V 32A. I log the voltage and current readouts of my power distributor. Speakon cables are 4mm² (11AWG). Even after several back-to-back full duration tests (2 minutes each) with ~2KW dumped into the loadbank, it was still completely cold to the touch from the outside. The tank holds about 15l of water.

 

[peniku8]

The Loadbank

While a bucket might've been the easiest solution, I wanted something rack mounted (for easy storage) that at least somewhat resembled something store-bought.
I figured I could fit everything into 3 rack spaces, so I bought a 3U case from Amazon.

As with everything I do, I first made a 3D model in Inventor. That always proved to leave very little room for errors. This is the front plate I made:

After I was happy with the design, I generated some G-Code with Inventor HSM and made the front plate on my cnc router.

I used a 2mm end mill to cut the holes. I have since bought a 3mm end mill specifically for sheet metal, which is much better suited for the task obviously, but the results were decent nontheless.

For the hardware I used regular Neutrik connectors, heavy duty (30A) switches and busbars for easy wiring. I glued the busbars to the front plate.

The heavy gauge wires were a pain to solder with my tiny soldering iron... I have 12 heating elements with the blue cap on the back and 4 without (went out of stock). The elements without the cap have a slightly lower impedance (~0.4 Ohm less).

I cut a hole into the rear case wall and made a mounting bracket for the heating elements with some acrylic. The plan was to screw the elements into the acrylic (has been tested with scrap pieces before) but that didn't work out and I didn't want to try myself out on thread milling just yet. I used my favourite stepped drill bit to make the holes a little larger and glued in the elements that wouldn't fit with silicone. Everything else came together nicely.

A quick sanity-check revealed that everything works as intended, which was a small surprise to me because I did all the wiring from the back of my head without a wiring diagram.
With the current setup I can do 4x8, 4x4, 2x2 and 1x1. And of course various useless impedance combinations.

The device is quite heavy fully assembled. I haven't weighed it yet, but it's probably around 40lbs. And it's sturdy.
This is what the back looks like now:

At first I thought I could get away with air cooling, but after one of those elements almost exploded into my face after plugging it into the wall in free air, I decided to go with a basic water bin. I still have lots of spare acrylic, which came in handy here.

I trimmed the edges with a flush trim bit and even made a roundover. I never planned on doing any experiments with hot-forming or welding. The glue I used basically welds plastics anyways. The container was made to slide into the case. You basically fill it up with water (to the point I marked to avoid overflow) and put the loadbank over it. The loadbank then rests on the container, with the heating elements being surrounded by at least an inch of water at all sides.

The case has enough room to accomodate the container as well. Without water preferably when stowing it...

That about wraps up my build, which I'm pretty satisfied with. I tested one amp already and it worked like a charm. The banana sockets are connections for my Pmillet Soundcard Interface (or basically any oscilloscope).

The font I used for the text on the front plate is called Elixia!

[klipsch]

Fun. Seen quite a few of these builds in the wild now. Looking forward to seeing you test. 

[Kyle]

Cool stuff!

[kipman725]

This is the best design I have seen so far.  Especially for the compact size and ability to have such a wide selection of loads.  I attach an image of my load bank which sits at the other end of the quality spectrum and is just a fixed 4ohms.  I tested a couple of amps with this load here:
https://www.diyaudio.com/forums/pa-systems/339848-quiet-pa-amps-5.html

 

 

[Ricci]

Excellent work. I look forward to your amp testing!

How do you feel about the Pmillet unit so far? 

Also about how much money do you have into the test load unit? 

[peniku8]

5 hours ago, kipman725 said:

This is the best design I have seen so far.  Especially for the compact size and ability to have such a wide selection of loads.  I attach an image of my load bank which sits at the other end of the quality spectrum and is just a fixed 4ohms.  I tested a couple of amps with this load here:
https://www.diyaudio.com/forums/pa-systems/339848-quiet-pa-amps-5.html

Hey, cool topic and cool loadbank. As long as it works, it works! I have a hard time settling for anything other than perfection so it had to be something like this at least haha!
As for quiet amps I can definitely recommend the t.amp TSA4-300 I'm currently testing. More power than what it's specced for and really quiet fans (temperature controlled). I also heard the lab gruppen C series seems to be passive when not under heavy load. I'm looking at the C20:8X for my theater plans.

[peniku8]

2 minutes ago, Ricci said:

Excellent work. I look forward to your amp testing!

How do you feel about the Pmillet unit so far? 

Also about how much money do you have into the test load unit? 

The Voltage meter on the Pmillet is really accurate down to maybe 15Hz. I've compared it to my HoldPeak true RMS meter and they're always within 1%. I did measure it but forgot to save and post the data here, but I can do it right now. I'm typing this in the studio sitting here with all the equipment. Distortion was 0.002% loopback of my audiointerface and 0.003 with the pmillet I think in the loop. But I have a wierd issue where distortion increases up to like 1% at 19dbU. It's 0.002 when I measure the loopback at -5dbU, idk why that is. But I have a plethora of devices connected to the interface and can't take it out.

The case was 40€, the heating elements 9€ each inlcuding shipping from China (like wtf). I got them from Ebay here. Switches, connectors, cables, busbars maybe 60 total. 250€ total what is like 280$ or so. Acrylic you can get scraps at shops for like 5$ a pound.

[peniku8]

That's my small test setup. Pretty simple, I'm doing a reference measurement without a load and then compare that to the measurements with different loads. Reference voltage is logged to calculate power from the db loss. More on the whole process later. I also log mains voltage and current. I have a 230V 32A 9 AWG supply line (3 Phase). Highest voltage sag observed was 1% (2V) so far at 14A current draw. I doubt the readout on the power distributor is very accurate, but it gives a general idea of long term efficiency. Speakon cables are also 11AWG/4mm² 5ft/1.5m cables to keep other error factors low. Btw the amp on the test bench there made me fear it'd blow up into my face.

[peniku8]

My signal chain consists of an audio interface (RME Fireface UC) and a sophisticated voltage divider network (shielded Pmillet Soundcard Interface).

All measurements have been taken through 15m (50ft) of balanced cables, as the interface is wired to the Neutrik patchbay that is installed in my outboard rack.
All measurements are take are taken at 96khz sample rate.
The 2V, 20V and 200V settings resulted in exactly the same FR.

Edit: I since revised the measurements below and split up the calibration files (one of the interface and one for the pmillett for an easier workflow in REW). The RTA pictures are new measurements, the ones below are older.
Ignore the dbV unit, that's not correct, but normalized to have a reference.

More measurements of the pmillett SCI at different voltages can be found here.

 

Note: the generator passthrough of the SCI doubles the output voltage, which means the Fireface generated the same voltage in both measurements.

 

Old measurements (not reliable) here:
Interface frequency response:

Interface Distortion:

 

Pmillet SCI frequency response:

Pmillet SCI distortion (high noise to due incorrect wiring):

 

A full loopback calibration file through both devices was active during amp measurements taken in REW.

[kipman725]

6 hours ago, peniku8 said:

Hey, cool topic and cool loadbank. As long as it works, it works! I have a hard time settling for anything other than perfection so it had to be something like this at least haha!
As for quiet amps I can definitely recommend the t.amp TSA4-300 I'm currently testing. More power than what it's specced for and really quiet fans (temperature controlled). I also heard the lab gruppen C series seems to be passive when not under heavy load. I'm looking at the C20:8X for my theater plans.

Ah great I have added them to the list, I started the thread after getting a load of Cloud VTXs home and finding they where so loud that I could hear them clearly several rooms away from where they where on with a closed door! subsequently I fixed a few of the amps and found they had failed because the airflow over the circuit board in one particular location was so strong it had worn away the track that switched the output relay.

[Ricci]

On 4/13/2020 at 12:45 PM, peniku8 said:

The Voltage meter on the Pmillet is really accurate down to maybe 15Hz. I've compared it to my HoldPeak true RMS meter and they're always within 1%. I did measure it but forgot to save and post the data here, but I can do it right now. I'm typing this in the studio sitting here with all the equipment. Distortion was 0.002% loopback of my audiointerface and 0.003 with the pmillet I think in the loop. But I have a wierd issue where distortion increases up to like 1% at 19dbU. It's 0.002 when I measure the loopback at -5dbU, idk why that is. But I have a plethora of devices connected to the interface and can't take it out.

The case was 40€, the heating elements 9€ each inlcuding shipping from China (like wtf). I got them from Ebay here. Switches, connectors, cables, busbars maybe 60 total. 250€ total what is like 280$ or so. Acrylic you can get scraps at shops for like 5$ a pound.

Good stuff and thanks. The load bank investment wasn't bad at all. 

Usually when you see distortion creep up abruptly in electronics something in the chain is at the onset of clipping. 

Measurements for the Pmillet look quite good. 

I look forward to your amp tests!

[peniku8]

16 minutes ago, Ricci said:

I look forward to your amp tests!

 So far one amp specified to 2x800W into 2 Ohm output 2x1200W instead and started smoking. Another amp started smelling very bad at 2 Ohm tests. I have 3 amps tested now and all the results saved, just need to get it up here. I know a few people who own amps which I don't own which they'd probably lend me for free, which is great. Talking Crown and Behringer amps here. I currently own 6 different of which 5 will be tested. One is a guitar power amp, which I don't feel like testing.

 

8 minutes ago, Ricci said:

Usually when you see distortion creep up abruptly in electronics something in the chain is at the onset of clipping.

That's the issue. When I see amps clipping they go from like 0.1% distortion to 5%+ within like 3 db. The thing with the loopback on the audiointerface is that I get those clean measurements you can see above at around -10dbU output and it only rises up (but steadily) to like 1.5% THD at +4dbU in that setting. It's a 1k audio interface. I wouldn't complain about something like that with my Behringer device because that was only 60$, but the Fireface should not output a 1.5% THD signal unless you really drive it into clipping. I even soldered new cables with high quality connectors, which did improve THD measurements from 0.002% to 0.000% from the clean measurements, but still. I did an amp sweep at 70V and distortion of the whole chain was 0.005% at 1khz. I also have a hard time finding out if my input or my output is causing the distortion. Or both or if it's internal crosstalk or something. I guess I will try the signal generator on my X32 as Input device and see if I can get clean input levels at +19dbU.
 

[kipman725]

with the +19dBu issue quite a few amps are designed to give their maximum output at 18dBu input.  So the input of something connected to your interface output could have some input voltage protection circuit that's causing distortion

[peniku8]

  Sad times, my clone amp exploded while testing. It tripped my C16 breaker on 230V after 10 seconds into any load except single channel 8Ohm, so I hooked it up to a 32A connection. Didn't like that 😪

But I can already tell you that it (Sanway Fp13000) does 7.5KW into 2Ohm for 80ms (one channel driven). It sustained 22A power draw but I haven't evaluated all results yet. Seems like just the main relais is broken, but I exploded the entire circuit board during troubleshooting, so I need to get that replaced. A coil, a Thermistor, two Resistors, one fuse and the PCB itself blew up. The amp is 7 years old.

 

9 hours ago, kipman725 said:

with the +19dBu issue quite a few amps are designed to give their maximum output at 18dBu input.  So the input of something connected to your interface output could have some input voltage protection circuit that's causing distortion

This happened during a loopback measurement of the audio interface and THD gradually rises over a span of ~15db.

[SME]

Only 80 ms?  That's not long at all considering that people are using those Sanways for ULF HTs.

[peniku8]

3 hours ago, SME said:

Only 80 ms?  That's not long at all considering that people are using those Sanways for ULF HTs.

Well those amps were not made to deliver maximum ouput for anything longer than you'd see with typical live music. I'll check sustained ouput later, but it's above 4KW from the PSU for sure. Well until it explodes.

[Ricci]

Yep...Based on previous tests on various clones the burst power is huge and should meet spec but true sustained power is about 4kw give or take which will eventually throttle back even more. There really isn't enough in the AC line to support more sustained power in most cases unless on 240v with large breaker so most of the big "tour" grade amps cap out at about the same sustained power. Even multiple speakers really cannot handle that kind of sustained power anyway. Some of the old "iron" amps would do sustained power for at least a few seconds but the power ratings were lower and again you better bring a stout AC line.

IMHO amp power ratings should be based on duration of a 1 cycle minimum for burst, 6.5 cycle shaped CEA-2010 signal as a short term rating and minimum 3 seconds sustained for long term. 

That's one thing that the name brand tour amps seem to do better than the clones is protections from blowing into tough loads. 

[kipman725]

My commiserations on blowing the amp I have also had an amp blow during testing (DAP Audio Qi-4600), nothing aggressive just sustained 4ohm load on one channel.  I then proceeded to totaly blow it trying to fix it.  Forgetting that it had a bipolar supply I hooked the scope probe up to the negative lead of a decoupling capacitor, shorting the power supply and unleashing a spectacular fireworks show.  My view is that if an amp fails like this its better than failing in a more critical situation and/or taking out speakers (the DC protect did not work on the Qi-4600, relay jamed shut).

My main lock down activity has been working on a new subwoofer amp that addresses the stout AC line issue by using a current limited supply and large capacitors.  The idea is that it never pulls a line current over a preset maximum but has huge transient energy reserves.  The attached PDF shows the basic idea.

Bass_hit_analysis.pdf

[radulescu_paul_mircea]

5 hours ago, kipman725 said:

My commiserations on blowing the amp I have also had an amp blow during testing (DAP Audio Qi-4600), nothing aggressive just sustained 4ohm load on one channel.  I then proceeded to totaly blow it trying to fix it.  Forgetting that it had a bipolar supply I hooked the scope probe up to the negative lead of a decoupling capacitor, shorting the power supply and unleashing a spectacular fireworks show.  My view is that if an amp fails like this its better than failing in a more critical situation and/or taking out speakers (the DC protect did not work on the Qi-4600, relay jamed shut).

My main lock down activity has been working on a new subwoofer amp that addresses the stout AC line issue by using a current limited supply and large capacitors.  The idea is that it never pulls a line current over a preset maximum but has huge transient energy reserves.  The attached PDF shows the basic idea.

Bass_hit_analysis.pdf 383.3 kB · 2 downloads

Extremely interesting idea about this amp? Are you limiting it to 110V-16A it will you go to 230 V and even more, like 32 A ?

[klipsch]

Fp22000q

Another's test of the largest/most powerful clone.

[kipman725]

52 minutes ago, radulescu_paul_mircea said:

Extremely interesting idea about this amp? Are you limiting it to 110V-16A it will you go to 230 V and even more, like 32 A ?

I'm in the UK where the most common power is 240V, 13A so that's my initial target*, but if I can complete the project I will try and do universal PFC.  Basic topology is:

240VAC -> 3.5 kW boost PFC (390VDC output) -> three phase buck converter operating in average current mode control (this is where the maximum current regulation occurs which in turn will regulate the line current) (300VDC bus, that drops under load) -> Full bridge IGBT based, low switching frequency class D amp 

Input is digitized and a look ahead limiter with trajectory estimation on the rail voltage is used to avoid clipping.  The buck converter is using microprocessor control so the maximum current can be set based on the available AC power and there is the possibility of lowering the 300VDC bus for IPAL compatibility.  Another feature is no mains isolation to increase efficiency.  Its early days at the moment I am working getting the buck converter stage test boards produced.  One issue is actually testing this properly is going to require a really big electronic load!

*I have made power converters before up to 90kW continuous output so I'm OK with going to higher currents but I don't have the facilities or test equipment at the moment and things get very expensive fast.

[peniku8]

@kipman725 Interesting concept, I guess you're going into a similar direction as what current high power PFC amps are doing?
I like the trend of new amps to come with a universal power supply. The new Powersoft amps work with any input voltage from 70-270V iirc, single bi and triple phase. Fully equipped with PFC, relatively high efficiency and the option to set a breaker in the dsp. As I see it, only increasing efficiency and stored energy will allow for more output with high crest factor content.
And yea on the amp, it was my main HT amp. I swapped it for one of my spare amps and lost ~8db headroom. The SKHorn is efficient in the upper register and room gain of my small room helps with the lower end of the spectrum, so the single sub I'm running will probable be fine. Gramps is currently investigating the remains of the amp to check for bad parts. I'll swap out the entire broke board and see if I get it running again. If that doesn't do it I'll see what if would cost to swap the entire PSU. I suspect the mains relais to be the point of failure, which would be an easy fix.

@klipsch I've read about it blowing up. notnyt is actually the guy who sold me the Pmillet Soundcard Interface, which I'm using for my tests 🙂

@Ricci I had it on a C16 breaker, which tripped after 14s. When I hooked it up for 32A230V it blew up...
The clone muted its output (DC protection) a few ms into the 1Ohm test. Maximum it did was 3.5KW for half a cycle. It also muted the output when I set my test frequency to 10khz instead of 1khz accidentally. I do all tests with a 1khz sine wave. "If the amp can do 10 cycles at 1khz it can do 1 cycle at 100Hz" is my logic, so the 1khz test allowed for a 1ms "resolution". I will publish the results as seen below. The graph below is the first I've finished. @SME I evaluated the results of the Sanway amp and noticed that the 2Ohm power sustained for 260ms, not 80. 260ms is very nice imo. Idk what it would look like with both channels driven, since I couldn't test that anymore.

The graph stops at 13.3s, because the breaker tripped. I won't test the amp again on a 32A circuit, even if I get it working again.

I tried to do all tests at a total duration of 2 minutes, but out of the 4 amps I tested this way one was fine, one started smoking, one started smelling and the clone blew up.
What would you guys suggest? Is 2 minutes too long? Not useful info anyways? Not all amps have superior safety mechanisms built in like maybe a Powersoft amp does...
I had a look at Martin Garrix's "Animals", which is a pretty bass intensive song. The longest duration of more or less uninterrupted bass was 28 seconds with a pretty low crest factor. Even most heavy bass dubstep songs have something that resembles a chorus and quieter sections in between, so I think a 30s sine wave test should cover most of it. I'm not in the situation yet where I can afford to blow up half of my amps because I want to test them at their absolute limits.

I'm hoping to have a write up on my testing methodology posted after the weekend and maybe some final results already.

[kipman725]

Two minutes of full output it too long in my opinion, sure the amp should survive but it doesn't reflect real world operating conditions and as such the 'thermal reserve' (to use the MA5002VZ terminology) is typically insufficent.  When testing the Crown CTS4200 which is an SMPS class AB the amp started to thermally limit at 1:30 of full output on a single channel.  "The longest duration of more or less uninterrupted bass was 28 seconds with a pretty low crest factor." I feel this is towards the upper end of what is asked of amps.  Contrast is needed in the music so going on much longer than 10s has diminishing returns.

Yeah my amp looks quite close to what Powersoft is dong, much less sophisticated and not full range but similar results.  I would love a couple of X8 but I don't think that's happening any time soon.

 

 

[SME]

17 hours ago, Ricci said:

Yep...Based on previous tests on various clones the burst power is huge and should meet spec but true sustained power is about 4kw give or take which will eventually throttle back even more. There really isn't enough in the AC line to support more sustained power in most cases unless on 240v with large breaker so most of the big "tour" grade amps cap out at about the same sustained power. Even multiple speakers really cannot handle that kind of sustained power anyway. Some of the old "iron" amps would do sustained power for at least a few seconds but the power ratings were lower and again you better bring a stout AC line.

IMHO amp power ratings should be based on duration of a 1 cycle minimum for burst, 6.5 cycle shaped CEA-2010 signal as a short term rating and minimum 3 seconds sustained for long term. 

That's one thing that the name brand tour amps seem to do better than the clones is protections from blowing into tough loads. 

Lots of people are using that for home theater subs that are meant to go to 10 Hz or below, so I think it's an important detail that you don't get peak power long enough to complete a full cycle at 10 Hz.  I know this kind of short duration peak power output is typical for amps which are designed to give their best for kick drum peaks and not a moment longer.  In a HT setting though, this amps doesn't look like it's going to hold up to one that can sustain the output for several hundred milliseconds at least.

If we're talking about highly dynamic medium duration "bursts" (e.g. on the order of 500 ms long), the "AC line" should handle it just fine.  I mean I'm sure there are exceptions.  The length of the run is important, and especially how the amp handles voltage droop.  The better amps don't care much about voltage droop and will suck every extra bit of current they need to keep the output going full-tilt for a medium to long "burst" duration (i.e. up to a few seconds with the SP amps).

I find it kind of lame that so many amps on the market now don't have decent protection circuits.