Fosi M03 / Aiyima 3001 / Nobsound G2 Pro mono sub amp review

[mcdn]

This is a review and measurements of the Fosi Audio M03 mono TPA3255 amplifier. It's designed to drive passive subwoofers which is why I bought it as I have an old sub I want to put to use in my games room. It appears to be identical to the Aiyima 3001 and Nobsound G2 Pro amps, and costs $90 on the Fosi website. It's a dedicated mono amp so only has one set of speaker outputs, and comes with an adjustable low pass filter in "sub mode". It ships with a 36V 5A supply brick, so we can expect max power to be 150W or less. It does claim it can be used with a bigger power supply, so I also tested that with a 48V/350W supply - see the results below.

Build and components

Before we get to the measurements let's talk about build and component quality. For a $90 amp the case, switches and knobs are all solid, with positive switch movement and good soldering. The power supply cabling on the DC side is marked as 20ga/0.5mm^2, but I checked and it's more like 22ga/0.3mm^2. Either way it is marginal for the rated 5A. The internal cabling from the amp outputs to the speaker terminals is also too thin, but I didn't take it apart to measure it. The board itself seems reasonably well put together. To increase power you can use up to a 48V supply according to the manufacturer. But the power supply capacitors are only 50V rated, which leaves no margin and could decrease the lifetime of the unit if operated at high power outputs. The heatsink is also only connected to the amp chip not the case, and in soak testing it hits thermal cutout after 60 seconds 5 minutes at 100W continuous output, or 1 minute at 150W (with a 48V supply). The good news is that the built in protection works and after power cycling it works fine again.

Inputs, switches, gain and filters

I need to explain the inputs and switches because the manual isn't easy to interpret. There are two RCA inputs on the back, one set of speaker outputs, and a switch on the front marked "PBTL/SUB". Then there's a "VOL" knob and a "SUB FREQ" knob. In either "SUB" or "PBTL" modes both inputs are summed. Gain in PBTL mode with both channels driven and the VOL knob at its midpoint is 32dB. Switch to SUB mode and gain doubles to 38dB. So one input driven in sub mode, or two inputs driven in PBTL mode give you the same overall gain. I don't show it below but distortion is the basically the same with the VOL knob at midpoint or max, so if you have a low output source you can use the volume knob with confidence.

In PBTL mode the "SUB FREQ" control does nothing. In SUB mode it gives an approximately second order rolloff depending on the setting. -6dB points are 45Hz at the lowest setting and 200Hz for the highest setting. All measurements are made with sub mode off.

Let's go to the measurements.

Unless otherwise stated, these measurements are taken at 5W into a 4ohm load using the stock 36V/5A supply in PBTL mode and an AES17 40KHz filter as is usual for class D amp measurements. The software is REW, the DAC is a Topping D10 balanced, and the ADC is an E1DA Cosmos ADCiso.

A frequency sweep looks good although for a sub amp the drop at 20Hz isn't ideal - maybe it uses PFFB as there seems to be no hump after 20KHz?

The 1KHz 5W FFT is good for noise at -110dB, but not great for distortion for a TPA3255 implementation, where we would expect closer to a 94dB SNR. I suspect the issue is with the front end where the filtering is taking place. Even though filtering is not enabled we see a bit of excess 3rd harmonic. This will be irrelevant for the intended use as a sub amp though.

19&20KHz dual tone is what you would expect given the above performance, with IM peaks around -50dB vs best-in-class TPA3255 at -70dB. Not audible or important for the usage as a sub amp, but worth noting.

The 32 tone multitone test isn't very pretty either at 12.8 bits, but is consistent with similar amps and fine for a sub amp - we are paying less than $100 here!

Now let's talk about power. Using the stock supply into a typical 4ohm load at 20Hz we only get about 100W at 1% THD (the result is the same at 20Hz or 1KHz):

If we switch to a bigger supply (Mean-Well 48V 350W) we can get to 230W into 4ohms:

But if we use a 2ohm load, even with the bigger power supply we only get to 100W. I suspect the inductors are saturating, or some other internal limiter is kicking in. [EDIT: doh! The bigger supply is 350W = 48V*7.3A, so I^2*R = 7.3^2*2 = 106W. Lots of these power results are simply current limited by the power supply].

Conclusions:

If used as intended with the stock power supply it will give 100W of clean power to a 4ohm passive sub, while summing both the inputs and providing some useful low pass filtering options. That's what it's meant for and I'll be keeping it in my games room system for that purpose. I'm not so impressed with some of the other claims made by Fosi regarding upgrading the power supply and so on, but for the money it's hard to complain.

 

[amirm]

Nicely done. Promoted to home page.

 

[theREALdotnet]

It appears to be identical to the Aiyima 3001 and Nobsound G2 Pro amps, and costs $90 on the Fosi website.

I have a Nobsound G2 Pro, it has gain selector DIP switches underneath. Does the Fosi have those, too?

I'm using the amp to drive a 12" Velodyne sub with dead plate amp (6 ohm driver). Works wonderfully.

 

[Jeromeof]

these measurements are taken at 5W into a 4ohm load using the stock 36V/5A supply in PBTL mode and an AES17 40KHz filter as is usual for class D amp measurements. The software is REW, the DAC is a Topping D10 balanced, and the ADC is an E1DA Cosmos ADCiso.

By the way, did you use a Low pass filter when measuring this Class D amp? As per the Texas Instruments guide: https://www.ti.com/lit/an/sloa107/sloa107.pdf

and nicely described here:

How to measure amplifiers with Cosmos ADC

In the past, entry-level audio interfaces were sufficient to measure the THD+N of amplifiers. However, recent improvements in "chip amp" performance have made it possible for the DIYers to build amplifiers that exceed SINAD 100. So there is a desire to measure amplifiers with high performance...

www.openaudiolab.com

 

[mcdn]

I have a Nobsound G2 Pro, it has gain selector DIP switches underneath. Does the Fosi have those, too?

I'm using the amp to drive a 12" Velodyne sub with dead plate amp (6 ohm driver). Works wonderfully.

It does have them on the board, but not exposed through the case like the Nobsound does. I'm sure it does work well with your Velodyne, as the measurements suggest it should!

 

[mcdn]

By the way, did you use a Low pass filter when measuring this Class D amp? As per the Texas Instruments guide: https://www.ti.com/lit/an/sloa107/sloa107.pdf

and nicely described here:

How to measure amplifiers with Cosmos ADC

In the past, entry-level audio interfaces were sufficient to measure the THD+N of amplifiers. However, recent improvements in "chip amp" performance have made it possible for the DIYers to build amplifiers that exceed SINAD 100. So there is a desire to measure amplifiers with high performance...

www.openaudiolab.com

Yes, I mentioned it in my remarks. Specifically this one which I designed with the help of the ASR community: https://audiosciencereview.com/forum/index.php?threads/diy-amp-testing-with-aes17-filter.46684/

 

[mcdn]

I cannot believe I left capacitor bait out for you @restorer-john and you didn’t take it!

 

[peniku8]

Now let's talk about power. Using the stock supply into a typical 4ohm load at 20Hz we only get about 100W at 1% THD (the result is the same at 20Hz or 1KHz):

View attachment 404219

What was the duration of the stimulus for this test? Cooldown/recharge period betewen the signals? And how does REW interpret the result if the RMS level drops over the duration of the test? Does it take the average level? Maximum level? Minimum Level?
All that doesn't matter for amplifiers without dynamic headroom (which probably includes this Fosi amp), but in any other case, I think REW is not suitable to test amplifiers for power output capabilities. I've raised similar concerns with John, but he said the to-do list was too long to implement a proper amplifier (and CEA-2010 routine I asked for) test routine at the current time.

I wish somebody would publish proper output over time graphs in their reviews such as this:

 

[mcdn]

@peniku8 I am not really sure I understand the question properly. The REW stepped sine test at each level takes as long as necessary to gather the FFT data required, which depends on the number of averages, the FFT length and the sampling frequency. In this case that is about 4 seconds per level, with a 100ms gap between levels.

As it happens, this specific amp runs at full power (100W) for about 60 seconds 5 minutes before shutting itself down. But it doesn't drop in power output over time, it just gets too hot and stops.

 

[peniku8]

@peniku8 I am not really sure I understand the question properly. The REW stepped sine test at each level takes as long as necessary to gather the FFT data required, which depends on the number of averages, the FFT length and the sampling frequency. In this case that is about 4 seconds per level, with a 100ms gap between levels.

As it happens, this specific amp runs at full power (100W) for about 60 seconds before shutting itself down. But it doesn't drop in power output over time, it just gets too hot and stops.

Well as music is not sine waves, amplifiers typically don't need to be capable of holding a 3dB crest factor signal (sine wave) for an indefinite amount of time.
And why make an amplifier that can hold a sine wave at max power for minutes, when the loudspeaker would've long burned by that time and also music is not a single sine wave at maximum amplitude either. That's why most modern PA amplifiers for example can output say 4KW per channel for a fraction of a second and then taper down to a quarter of that output level with longer signals (capacitors in the PSU drained). Your test (and any test in REW for that matter) would fail to identify such characteristics and eventually lead to false data/conclusions.
It's probably accurate for most low-power home audio products, but I wanted to raise general awareness that using stepped level sines in REW is not an appropriate test for amplifiers with any kind of dynamic headroom. If you want to measure dynamic headroom, you'll have to record the signal the amplifier is generating and prod the recording manually (or use an AP/Klippel routine; I'm not sure if either of those has a method to output power over time graphs).

Here is an example of my data-logging of an amplifier test in Audition:

I eyeball the 1% THD and then calculate RMS level based on peak level that Audition displays (that way clipping can't artificially boost output power levels).
It's annoying to do all this manually, but it's what you have to put up with if you're not in the game to drop AP kind of money.

 

[chych7]

I see these used often for driving tactile transducers, which can create usable output well below 20 Hz. Any other measurements or considerations that could be beneficial, to characterize the amp for this type of use case?

 

[Timbo2]

Thanks!

I just wound up picking up the smaller Fosi Audio M04 Amplifier and found it worked really well for my PC and an old repurposed subwoofer. My trial and error with a test tones and the subwoofer crossover knob suggest it has the same curves as the one you measured here. I was pleasantly surprised by the quality of the case and feel of the knobs and switches.

 

[restorer-john]

@mcdn I like your review style.

You may want to mark the frequencies used on plot 6&7.

Based on your tests, what is the actual continuous (5 minutes) maximum power this amplifier can do with the cover on, at 1kHz? If it's shutting down after 60 seconds at 100W@4R it's not going to be impressive...

 

[mcdn]

@mcdn I like your review style.

I copied another bloke on here

You may want to mark the frequencies used on plot 6&7.

It is a bit buried in the text - 20Hz and 1KHz were used, but only the 1KHz graphs are shown as the result is the same for both.

Based on your tests, what is the actual continuous (5 minutes) maximum power this amplifier can do with the cover on, at 1kHz? If it's shutting down after 60 seconds at 100W@4R it's not going to be impressive...

Oops, I double checked and I'd misremembered the 60 second power. That's actually at 160W or so with the bigger supply - review edited.

Retesting just now with the stock supply it gets to 6 minutes at 100W/4 ohms before shutting down. That's at 21C ambient, and the top of the case was at 26C when it shut down. I'd guess somewhere around 50W as the true continuous power, especially if living in a cabinet.

@peniku8, as you'll see from other reviews by Amir, Class D amps don't often have much if any increase in instantaneous power, as the sources of distortion aren't time variant. So if you want 1% distortion you get 100W with this amp, whether it's for 50 milliseconds or 5 seconds. 10% arrives around 110W from memory. With a bigger supply it's 230W for 1% with thermal shutdown arriving in under a minute. But as you say on music material rather than test tones that's probably plenty. In that case you could consider 230W/1%/4ohms as the "instantaneous peak power".

 

[restorer-john]

Retesting just now with the stock supply it gets to 6 minutes at 100W/4 ohms before shutting down. That's at 21C ambient, and the top of the case was at 26C when it shut down. I'd guess somewhere around 50W as the true continuous power, especially if living in a cabinet.

That's actually really impressive. I highly doubt normal* usage would cause thermal shutdown.

*excluding teenagers, parties, extreme subwoofering etc.

 

[mcdn]

That's actually really impressive. I highly doubt normal* usage would cause thermal shutdown.

*excluding teenagers, parties, extreme subwoofering etc.

I totally agree with you. The headline numbers might make people think it's "not so great", but on real music, even highly compressed rock/pop/hip-hop, crest factors are around 9dB, which is a factor of 3. So this little thing can pump out party music at 30W into 4 ohms forever, which is way better than my little old Cyrus 1 back in the day! With 86dB floorstanders like the Polk XT60 at $249 that's an ear-bending 96dBSPL/1m (103dBSPL/1m for 2 speakers, though you'll need 2 amps). Party on!

 

[mcdn]

Subjective note: now that it's off the test bench and into a real system it's performing very well. It's driving an Artcoustic Diablo Sub Panel, which is 2x10" Peerless drivers in a wall mounted shallow sealed cabinet. The aim is to supplement the LSR305P active speakers that sit on the wall, and for that the filter and gain settings work well without any more DSP required. I thought I was going to have to press an old MiniDSP 2x4 into use and create even more cable mess, but it's really not needed. Games room movies and Xbox bass sorted.

 

[peniku8]

@peniku8, as you'll see from other reviews by Amir, Class D amps don't often have much if any increase in instantaneous power

This is a false conclusion from the sort of products tested. Like I mentioned earlier, low-power home audio amplifiers typically don't have much dynamic headroom because it's easy enough nowadays to make a 500W SMPS small, light and efficient enough for the task at hand, while multi-KW pro audio amplifiers are almost exclusively class D and they all offer some sort of dynamic headroom, typically multiples of what they can sustain indefinitely (otherwise amplifiers with more than 3.6KW of output power wouldn't make sense on a normal 230V 16A socket and quite a number of amplifiers offer more than 10KW of output power nowadays).
Don't get me wrong, I'm not trying to say your review is wrong, I just want to highlight what drawbacks this kind of testing has in general. Also I would be quite interested to see *if* there are some low power products on the market that behave differently over time. Note that this behavior is also not linked to distortion. The amps' built-in limiters reduce output power dynamically and cleanly (most of the time) and it's even pretty much impossible to clip the outputs of some amplifiers.

 

[mcdn]

This is a false conclusion from the sort of products tested. Like I mentioned earlier, low-power home audio amplifiers typically don't have much dynamic headroom because it's easy enough nowadays to make a 500W SMPS small, light and efficient enough for the task at hand, while multi-KW pro audio amplifiers are almost exclusively class D and they all offer some sort of dynamic headroom, typically multiples of what they can sustain indefinitely (otherwise amplifiers with more than 3.6KW of output power wouldn't make sense on a normal 230V 16A socket and quite a number of amplifiers offer more than 10KW of output power nowadays).
Don't get me wrong, I'm not trying to say your review is wrong, I just want to highlight what drawbacks this kind of testing has in general. Also I would be quite interested to see *if* there are some low power products on the market that behave differently over time. Note that this behavior is also not linked to distortion. The amps' built-in limiters reduce output power dynamically and cleanly (most of the time) and it's even pretty much impossible to clip the outputs of some amplifiers.

I'm sure you're right, essentially at some point it becomes impractical to make the power supply bigger (especially in the 110V USA), so manufacturers add big enough banks of capacitors to make up the current shortfall in the short term. Is this correct? Warning, I am not great at the analogue stuff so the calculations below may be way off, please correct as needed!

Take the TPA3255 in use with this amp. Checking the datasheet, max current per channel at 1% THD is 5A (255W/51V into 4ohms in BTL config). So 10A max current required from the power supply. Given that, and allowing for 90% efficiency, a 51V x 11A = 550W regulated supply is all you would ever be able to use with this chip. Adding more capacitors won't do anything except cost money.

Alternatively you could use a lower power supply and add lots of storage capacitors for peak current delivery. To maintain 550W for 50ms requires 550 * 0.05 = 27.5 watt-seconds (joules). Which at 51V is the energy in 2*27.5/51^2 = 21milliFarads. But voltage drops quickly, so you'd probably need 4x that, making ~80mF, or 80000uF, which at these power levels would likely be a poor tradeoff because that's probably $40 even in bulk, and you can just get a bigger power supply for that money.

For this strategy to deliver 10kW for 50ms at 51V would need an astonishing 1.6 Farads of capacitance, but I believe these pro amps would use much higher voltages in order to drive strings of speakers in series? At 200V we'd be back down to ~100mF which seems reasonable for the application.

 

[peniku8]

I'm sure you're right, essentially at some point it becomes impractical to make the power supply bigger (especially in the 110V USA), so manufacturers add big enough banks of capacitors to make up the current shortfall in the short term. Is this correct?

That's how it is, indeed. With cheaper and lower power models, manufacturers are of course most interested in keeping costs down, so the power supply is designed with whatever makes the most sense economically. Even with amps with less than "breaker-tripping" power, a bank of caps (and a PFC module) would make sense to reduce current draw spikes on the AC input side.

I'm sure you're right, essentially at some point it becomes impractical to make the power supply bigger (especially in the 110V USA), so manufacturers add big enough banks of capacitors to make up the current shortfall in the short term. Is this correct? Warning, I am not great at the analogue stuff so the calculations below may be way off, please correct as needed!

Take the TPA3255 in use with this amp. Checking the datasheet, max current per channel at 1% THD is 5A (255W/51V into 4ohms in BTL config). So 10A max current required from the power supply. Given that, and allowing for 90% efficiency, a 51V x 11A = 550W regulated supply is all you would ever be able to use with this chip. Adding more capacitors won't do anything except cost money.

Alternatively you could use a lower power supply and add lots of storage capacitors for peak current delivery. To maintain 550W for 50ms requires 550 * 0.05 = 27.5 watt-seconds (joules). Which at 51V is the energy in 2*27.5/51^2 = 21milliFarads. But voltage drops quickly, so you'd probably need 4x that, making ~80mF, or 80000uF, which at these power levels would likely be a poor tradeoff because that's probably $40 even in bulk, and you can just get a bigger power supply for that money.

For this strategy to deliver 10kW for 50ms at 51V would need an astonishing 1.6 Farads of capacitance, but I believe these pro amps would use much higher voltages in order to drive strings of speakers in series? At 200V we'd be back down to ~100mF which seems reasonable for the application.

As you describe it, I think it's pretty safe to assume that the solution with the capacitor bank is too costly for home audio amplifiers, hence there being no dynamic headroom, as the power supplies are able to provide full power all the time (the only limitation would then be efficiency aka heat build up).
Speakers are typically not driven in series in pro audio and it's not uncommon to actually have a bunch of subs in parallel down to like 2 Ohm. One amplifier I recently bought states 750 Joule of energy storage for example. It has a Vpk of 185V, which means it should have +-185VDC from the PSU (rails) if I understand that correctly (am not an amplifier designer but maybe @Mad_Economist can chime in because I couldn't find @Ratio here). The rails should be stable at that voltage, no matter what the amp is outputting, so the caps can run at that voltage accordingly (and not need to be able to supply insane currents). If the rails collapse, it's because the caps are drained and the PSU can't keep up with the power demands anymore (that's based on my limited understanding on how those amplifiers work).
I'd be happy if someone with a deeper insight in the topic could chime in, as this is out of my comfort zone.

Edit: I saw Ratio's ping worked despite the forum not showing him as a registered user, maybe he hasn't logged in for a while. I also don't even know if that's the same person, sorry if not!