Some of you may recall that I used to recommend the Mean Well RPS-400-36(-C) for use with the Modulus-686. A few builders had issues with audible whine from these supplies under light load, so I stopped recommending that model. I still stand by my recommendation of the 27 V version (RPS-400-27-C). At least the ten RPS-400-27(-C) I've used so far to power Modulus-686 amps have all worked flawlessly.
Anyway. Thanks to efforts behind the scenes by members BrianL and SLEdwards12375 (thanks guys!), I can now provide a fix for those affected by the supply whine: Turn up the voltage on the RPS-400-36(-C) to 40 V by turning the adjustment pot on the RPS-400. You will definitely need large heat sinks if you go this route. The ModuShop Dissipante 4U x 400 mm would be a good fit.
I generally don't recommend going beyond ±36 V with the LM3886 (hence, Modulus-686) as doing so really pushes the thermal dissipation in the IC. However, with a large enough heat sink and a regulated supply (to prevent over-voltage on the chip) it is possible. At the very least, it's a possible solution for those wishing to use the RPS-400-36(-C) with the Modulus-686.
Tom
Anyway. Thanks to efforts behind the scenes by members BrianL and SLEdwards12375 (thanks guys!), I can now provide a fix for those affected by the supply whine: Turn up the voltage on the RPS-400-36(-C) to 40 V by turning the adjustment pot on the RPS-400. You will definitely need large heat sinks if you go this route. The ModuShop Dissipante 4U x 400 mm would be a good fit.
I generally don't recommend going beyond ±36 V with the LM3886 (hence, Modulus-686) as doing so really pushes the thermal dissipation in the IC. However, with a large enough heat sink and a regulated supply (to prevent over-voltage on the chip) it is possible. At the very least, it's a possible solution for those wishing to use the RPS-400-36(-C) with the Modulus-686.
Tom
Last edited:
I can report now that after continuous usage at very aggressive output levels, the heatsinks are barely warm to the touch on my 4U Dissipante chassis with the supplies adjusted to 40 volts (39.85 with adjustment pots full cw). No whine from my speakers or power supplies themselves, just dead silence.
Steve
Steve
NICE!!! Thanks for following up.
Re. lukewarm heat sinks: I'll be the first to admit that I'm a bit more conservative than most when it comes to thermal design. So far, lab measurements using AP's 32-tone signal (music approximation) played at levels just below clipping have matched math pretty well, so I have no plans to change that.
My ±27 V powered (RPS-400-27-C) 3U Mini Dissipante build of the MOD686 only reaches lukewarm with aggressive output levels too. I've had it up to 65 ºC in the lab on more than one occasion with the 32-tone signal, though. I suspect the difference is caused by actual music being well below clipping for the vast majority of the time. Only on loud passages will the signal approach clipping. So, in addition to the crest factor, there's effectively a low duty cycle at play too.
Tom
Re. lukewarm heat sinks: I'll be the first to admit that I'm a bit more conservative than most when it comes to thermal design. So far, lab measurements using AP's 32-tone signal (music approximation) played at levels just below clipping have matched math pretty well, so I have no plans to change that.
My ±27 V powered (RPS-400-27-C) 3U Mini Dissipante build of the MOD686 only reaches lukewarm with aggressive output levels too. I've had it up to 65 ºC in the lab on more than one occasion with the 32-tone signal, though. I suspect the difference is caused by actual music being well below clipping for the vast majority of the time. Only on loud passages will the signal approach clipping. So, in addition to the crest factor, there's effectively a low duty cycle at play too.
Tom
As promised. You can find my thoughts on snubbers as part of my Taming the LM3886 article series: Power Supply Design: Rectification & Snubbers.
To nobody's surprise, the addition of snubbers to the power supply makes no difference on the output of the power supply or on the output of the power amp connected to the power supply. Thus, I will continue to view claims of audible superiority of snubbers with the same skepticism as I view claims of audible superiority of speaker cable stands, at least in the cases of solid state audio power amplifiers and their power supplies.
That said, a properly designed snubber does reduce the RF ringing that can occur when the rectifier diodes turn off. Thus, I encourage those who are concerned about RF emissions, EMI, and such to use snubbers. 100 nF + 10 Ω will work well. For those who wish to optimize the snubber further, I give a simple iterative procedure. The method I use accounts for the parasitics in the circuit when operated at the target working voltage, thus arrive at a snubber which is much closer to optimum than you'd find with other methods. This method requires no special tools. An oscilloscope and a simple RC filter is all you need.
Enjoy.
Tom
Quite a few of the vacuum tube folks who install 1R-1C snubbers across the secondary(s) of their power transformer, insist upon fitting a flameproof type of resistor for the snubber. Here's one example at DigiKey link.
Their reasoning is:
Their reasoning is:
- A voltage spike or "surge" on the AC mains, will create a large delta-V across the secondary
- Thanks to the series capacitor in the 1R-1C snubber , all of this delta-V appears across the snubber resistor
- Suddenly you've got half a kilovolt across a 10 ohm resistor, making it go kaboom very quickly
- When flameproof resistors go kaboom, it is less dangerous than the kaboom from other types of resistors
Yeah. There are many design practices that apply to vacuum tube circuits that don't apply at voltage levels relevant to the Modulus-686. I wouldn't recommend coating the supply connections in a solid state power amp with corona dope, for example. Even if the added cost is minimal.
I will always encourage circuit designers and builders to apply science and engineering in their component choices. That's what I do.
Tom
I will always encourage circuit designers and builders to apply science and engineering in their component choices. That's what I do.
Tom
Back to the Mean Well supply topic for one additional diversion of the thread:
I’ll add a few comments on my investigation into the Mean Wells’ performance for those who stumble across this thread but have a different end-use in mind. As Tom indicated, the ‘fix’ for the Mod686 design is to use the output trim pots on the power supplies to increase their Vout to the max, which is right at 40V. While this works for the Mod686, it is a serendipity that may not occur for other applications of this particular supply. Here is a brief summary of what’s going on:
- Many switch-mode supplies, when outputting small amounts of power relative to their max rating, work in a ‘burst’ mode, essentially turning themselves on and off to inject small bursts of power into their outputs to keep the output voltage within tolerance. This switching on/off of the supply’s switcher occurs at frequencies in the audio band (varying with load conditions) and somehow gets into the audio amplifier’s output, revealing itself as audible noise. Still to be determined in this particular application is how this happens, but it does not appear to be a simple case of power supply noise getting through the amplifier’s power supply rejection (PSRR).
- What makes this ‘fix’ work is that, oddly, this Mean Well supply exits it’s low-power burst mode of operation when the output voltage is cranked up to 40V _AND_ the load is drawing more than approximately 350mA.
- Second, the quiescent (no-signal) current consumption of the Mod686 is just a bit more than this minimum current. Thus this fix works just fine for this target amplifier.
- It’s not clear why the Mean Well behaves in this manner. You can adjust the output voltage to a lower value, draw quite a bit more output current/power, but the supply will operate in the noisier burst mode.
The bottom line is that if you’re using this model of Mean Well power supply in a slightly different application that has less quiescent current consumption, you likely will need to add a bit of resistive loading to the supply’s output to keep the minimum current consumption at least approximately 350 mA.
Different switch-mode supplies likely will work differently – there are a number of different common circuit topologies - and perhaps even different Mean Well models will behave differently. You can’t assume that this fix will work for a different supply without extensive characterization.
All this said, I’m generally positively impressed with the Mean Well products. In my investigation I’ve accidentally tried to destroy it quite a few times; all it did was shut itself down and then after pulling the power cord for about 10 seconds, came right back to life.
I’ll add a few comments on my investigation into the Mean Wells’ performance for those who stumble across this thread but have a different end-use in mind. As Tom indicated, the ‘fix’ for the Mod686 design is to use the output trim pots on the power supplies to increase their Vout to the max, which is right at 40V. While this works for the Mod686, it is a serendipity that may not occur for other applications of this particular supply. Here is a brief summary of what’s going on:
- Many switch-mode supplies, when outputting small amounts of power relative to their max rating, work in a ‘burst’ mode, essentially turning themselves on and off to inject small bursts of power into their outputs to keep the output voltage within tolerance. This switching on/off of the supply’s switcher occurs at frequencies in the audio band (varying with load conditions) and somehow gets into the audio amplifier’s output, revealing itself as audible noise. Still to be determined in this particular application is how this happens, but it does not appear to be a simple case of power supply noise getting through the amplifier’s power supply rejection (PSRR).
- What makes this ‘fix’ work is that, oddly, this Mean Well supply exits it’s low-power burst mode of operation when the output voltage is cranked up to 40V _AND_ the load is drawing more than approximately 350mA.
- Second, the quiescent (no-signal) current consumption of the Mod686 is just a bit more than this minimum current. Thus this fix works just fine for this target amplifier.
- It’s not clear why the Mean Well behaves in this manner. You can adjust the output voltage to a lower value, draw quite a bit more output current/power, but the supply will operate in the noisier burst mode.
The bottom line is that if you’re using this model of Mean Well power supply in a slightly different application that has less quiescent current consumption, you likely will need to add a bit of resistive loading to the supply’s output to keep the minimum current consumption at least approximately 350 mA.
Different switch-mode supplies likely will work differently – there are a number of different common circuit topologies - and perhaps even different Mean Well models will behave differently. You can’t assume that this fix will work for a different supply without extensive characterization.
All this said, I’m generally positively impressed with the Mean Well products. In my investigation I’ve accidentally tried to destroy it quite a few times; all it did was shut itself down and then after pulling the power cord for about 10 seconds, came right back to life.
I was under the impression that CRC filtering in a linear supply is not meant to smoothen the 120Hz primary ripple, but filter the noise which happens at much higher frequencies because of diode switching. I believe most amps have impressive PSRR at low frequencies, but the PSRR graph slopes downwards with frequency. So, HF noise on rails can get through to the output. And CRC for the rails, even with 0.5Ohm R and each C as big as what a normal pure-C filter would do, can give you a knee somewhere in the 100Hz area. This means that with a knee that early, the attenuation of HF noise can be very effective by the time one reaches 1K and beyond, even at 6dB cut/octave. You get something like 18-20dB cut at 1KHz or so, very roughly, and further up the frequency axis it only gets better.
Is my reasoning making sense?
Basic question: why do amp designers not add the speaker protection to the basic amp PCB itself? Is there any good reason to omit speaker protection (specially in big, high quality, high power amps) or keep them separate? I'd really love lowering of PCB count.
As far as I understand it, CRC filtering is an idea borrowed from tube amps. CRC or CLC filtering is used to get the supply ripple down. It works in a tube amp because the supply current is rather low (a few hundred mA maybe), so you can use a significant resistance without incurring a huge voltage drop.
Using a CLC filter in a solid state amp (lower supply voltage, higher supply current) would require an inductor with a saturation current in the tens of ampere. You can find inductors like that, but they tend to be in the uH range. So people use CRC instead. As far as I can tell, they just seem to forget to check the math to see if they actually get any meaningful attenuation at frequencies relevant to mains rectification.
You are correct that you'll get some reduction of the switching hash. But you will also get increased signal-dependent ripple with the CRC filter, which could lead to an increase in THD of the connected amp. Exactly where that tradeoff lies could be the topic of another article. However, the mains-related switching hash is below the noise floor in my amps, so I don't see any reason to attenuate it and risk increasing the THD.
It adds cost and PCB area. Some builders prefer to be without it.
Tom
I agree about your amps -- the SNR measured figures are adequate proof. I suspect not all amps built by DIYers have equally good HF PSRR. In particular, I see a leaning towards the minimalist approach in a lot of DIY designs, where the amp circuit topology itself (things like ultra simple CCS, etc) makes me worry about their PSRR.
The ISS uses a combination of a "solid state relay" (TRIAC) and a mechanical relay. The TRIAC makes the initial connection/disconnection, which drastically reduces the wear on the relay contacts. The relay in turn saves on the "wear" experienced by the TRIAC. Solid state relays tend to "wear" from the current flowing through.
I take it a step further and reduce the power dissipated in the relay coil once the contact has been made. This increases the lifespan of the relay coil.
Tom
While looking through the Doug Self preamp schematic which has been published in Linear Audio, I came across this way of wiring up a balanced and unbalanced input socket without a switch:
- Pin 1 of XLR --> shield of RCA --> signal ground
- Pin 2 of XLR --> signal pin of RCA --> one of the differential inputs
- Pin 3 of XLR --> no connection with RCA --> the second differential input
Which volume of Linear Audio?
I don't like the idea of leaving XLR Pin 3 floating. That means the inverting input of the differential receiver is floating.
I prefer to connect XLR Pin 3 to the RCA shell (along with XLR Pin 1). Unfortunately, it isn't possible to have both RCA and XLR connectors hooked up without using a switch to select between the two, though.
There may have been something inherent in Self's design that allowed such a setup to work, but I would not recommend it for any of my amps.
Tom