Very impressive planning and build process, I'm really impressed. But I can't understand how you are managing with one 500VA transformer for 2 big and 4 small channels, when some of us Modulus 686 builders are using much bigger transformers for just 2 channels of the 686. I am sure you have thought this through and Tom has helped, but I can't seem to figure it out. I'd have used a 1KVA transformer. At least.
Given that this build is in a 3U chassis, I'm thinking it's a Safe-n-Sane build. There are quite a few options for power supplies with the Modulus-686. Two that I recommend are:
Note that my recommendations assume music reproduction into a 4 Ω load. If you're running 8 Ω speakers (which I think OP did) and want to optimize the amp for that, you can get away with a smaller power transformer. Some choose to do that to save a few bucks on the chassis and power transformer.
Tom
- ±27-30 V provided either by a pair of Mean Well RPS-400-27 or a Power-686 and a 2x22 VAC, 400-500 VA power transformer. I call this the Safe-n-Sane build because it's relatively low stress on the Modulus-686 and can go in a 3U x 300 mm ModuShop Mini Dissipante chassis.
- ±35-36 V provided either by an SMPS or a Power-686 with a 2x25 VAC, 600-800 VA power transformer. This is the Full Power build. The Modulus-686 dissipates more power with the higher supply voltage, so this build needs a 4U x 300 mm chassis minimum. Many use a 4U x 400 mm to be able to fit everything in the chassis.
Note that my recommendations assume music reproduction into a 4 Ω load. If you're running 8 Ω speakers (which I think OP did) and want to optimize the amp for that, you can get away with a smaller power transformer. Some choose to do that to save a few bucks on the chassis and power transformer.
Tom
I now understand that you mapped the 3U height to the power rating. Makes sense.
But if that wasn't the case, then your specs say that a Modulus 686 is rated 360W into 4 Ohms, and a Modulus 86 is rated 65W similarly. This adds up to (360x2) + (65x4) = 980W, which makes a 1KVA transformer pretty much a bare minimum, wouldn't it? After all, your power ratings are for power pumped into the load, not power dissipated by the amp+speaker combo.
But if that wasn't the case, then your specs say that a Modulus 686 is rated 360W into 4 Ohms, and a Modulus 86 is rated 65W similarly. This adds up to (360x2) + (65x4) = 980W, which makes a 1KVA transformer pretty much a bare minimum, wouldn't it? After all, your power ratings are for power pumped into the load, not power dissipated by the amp+speaker combo.
Math says the power consumption of each amp is:
Modulus-86: 99.68 W (sine wave); 30.68 W (music, 14 dB CF)
Modulus-686: 398.7 W (sine wave); 122.7 W (music, 14 dB CF)
Numbers assume ±36 V operation and 8 Ω load which is what that multi-channel amp is optimized for. You can find the math here: https://neurochrome.com/pages/power-supply-design (and in any engineering text that deals with Class AB output stages).
4*30.68 + 2*122.7 = 368.1 W. A toroidal transformer around 400-500 VA would be plenty.
Tom
Modulus-86: 99.68 W (sine wave); 30.68 W (music, 14 dB CF)
Modulus-686: 398.7 W (sine wave); 122.7 W (music, 14 dB CF)
Numbers assume ±36 V operation and 8 Ω load which is what that multi-channel amp is optimized for. You can find the math here: https://neurochrome.com/pages/power-supply-design (and in any engineering text that deals with Class AB output stages).
4*30.68 + 2*122.7 = 368.1 W. A toroidal transformer around 400-500 VA would be plenty.
Tom
The assumption of 8 Ohms load in this case makes complete sense, if the designer has decided on his drivers. But I found your analysis of derating the power supply based on crest factor quite interesting. I used to believe that crescendos in music can last even a few seconds, not just one or two wavelengths, and if the power supply is rated for 14dB lower power, then those loud peaks will sound muddy, and subjective evaluation of your amp will take a hit, through no fault of the amp module per se.
You refer to "any engineering text that deals with Class AB output stages", saying that they all do similar power supply calculation, but I haven't seen this kind of derating for power amps in Randy Slone's books. Did he miss this out? I don't remember what Doug Self or Bob Cordell's books say.
The equation for the power dissipation in a Class AB output stage is available in many texts. I like Sedra/Smith. You don't need the latest edition. I used the 3rd edition in college. That has the math. You can also derive the equation.
I'm not aware of an engineering text that covers the math around compensating for the crest factor, but there are a few app notes on the topic, including one from National Semiconductor if I recall correctly.
I suspect whether an amp sounds muddy on the peaks has more to do with the amount of power supply capacitance.
The VA rating of a power transformer should not be interpreted as the destruction limit. If you draw 500.001 VA from a 500 VA rated transformer for a brief moment, it won't explode. I promise. Not because you overloaded it anyway. The transformer will fail with sustained overload. And by "sustained" I mean "for hours".
If you want to build for continuous sine wave operation, that's your choice. You will end up with a very large and very expensive amp.
Note that the 3U (±27-30 V) and 4U (±35-36 V) heat sink sizes I mention assume the same 14 dB CF. If you want to design for sine wave operation, you'll have to invest in larger heat sinks.
Tom
I'm not aware of an engineering text that covers the math around compensating for the crest factor, but there are a few app notes on the topic, including one from National Semiconductor if I recall correctly.
I suspect whether an amp sounds muddy on the peaks has more to do with the amount of power supply capacitance.
The VA rating of a power transformer should not be interpreted as the destruction limit. If you draw 500.001 VA from a 500 VA rated transformer for a brief moment, it won't explode. I promise. Not because you overloaded it anyway. The transformer will fail with sustained overload. And by "sustained" I mean "for hours".
If you want to build for continuous sine wave operation, that's your choice. You will end up with a very large and very expensive amp.
Note that the 3U (±27-30 V) and 4U (±35-36 V) heat sink sizes I mention assume the same 14 dB CF. If you want to design for sine wave operation, you'll have to invest in larger heat sinks.
Tom
That's a different issue altogether. Power dissipation inside a Class B OPS is quite different from sizing the power supply transformer for a particular amp power rating. For instance, the power dissipation inside the amp actually reduces at peak power output, because of the power is dissipated in the speaker load. Therefore, the power dissipation in a Class B OPS becomes relevant for sizing heatsinks, but a different tack is taken for sizing the power supply.
Yes, this is too obvious for us to debate. We both understand the meaning of "VA rating of the transformer" similarly, and voltage droop due to load etc come into this sizing.
I think the point of discussion between you and me right now is about how much distortion will be introduced into the amp output due to the voltage droop in the transformer secondaries due to typical transient peaks in the music programme. Neither of us is debating whether a peak of 2 secs will result in the transformer getting destroyed. My question becomes more pertinent for certain music industry practices like "hot mixes" where the crest factor shrinks. (Again, I'm not arguing about what the typical crest factor today is. I'm just trying to understand whether peaks trigger distortion if the xfmr is not rated at the amp's rated dissipation.)
From the last few messages we've exchanged, the impression I am getting is that no one seems to have hard data about this.
Thanks. Neither the size difference nor the price difference between a 500VA vs a 1 KVA xfmr is a very large percentage change if we take the overall size, weight and cost of a finished amplifier as the baseline. So, if that was the only issue, we can live with it. In particular, I would expect that any builder who has chosen the Modulus 686 module is shooting for uncompromising SOTA objective performance, and the small extra expense and weight of a larger xfmr will hardly be a factor in decision making if it means that we can keep subjective SQ at peak levels during music crescendos.
In the meantime, I looked at books on amplifier design (the usual suspects) and it appears that their authors admit that the VA rating of the transformer which is needed for a specific amp is not easy to arrive at based on just the published VA rating alone -- it requires experimentation with samples from the transformer manufacturer and measurements of the finished amp with different transformer models. Doug Self brings up the crest factor angle in the following way: "Since most amplifiers are intended to reproduce music and speech, with high peak-to-average power ratios, they will operate satisfactorily with transformers rated to supply only 70% of the current required for extended sinewave operation, and in a competitive market the cost savings are significant."
He also says: "...the voltage losses in the power amplifier itself are not that easy to predict, some of the clipping mechanisms being quite complicated in detail. The inevitable conclusion is that the fastest way to reach a satisfactory transformer design is to make only approximate calculations, order a prototype as soon possible, and fine-tune the required voltage from there."
He also explains how the data given by transformer manufacturers about their products is not sufficient to take accurate decisions based on the paper spec alone, and he also highlights how the variations in the specs of the smoothing capacitors too impact the final amp behaviour.
So, it appears that the entire issue of VA rating for a specific power delivery with real music is a matter of debate and experimentation.
Thanks for your patience, Tom.
I honestly wasn't sure what you were debating until now, so thank you for clarifying.
The Neurochrome Modulus amps (thread topic) are not sensitive to any perturbations on the power supply. So if the power supply droops by 2 V under load, the amp would simply clip 2 V earlier. Other amps may behave differently.
And, of course, much of this is moot when a regulated SMPS is used.
Tom
The Neurochrome Modulus amps (thread topic) are not sensitive to any perturbations on the power supply. So if the power supply droops by 2 V under load, the amp would simply clip 2 V earlier. Other amps may behave differently.
Sounds like you've found an opportunity to go and generate such hard data. 🙂 That seems like a topic for a thread of its own. If you start the thread I'm happy to participate. I may even be able to contribute some data.
It seems like you're making it out to more than it needs to be. Yes, transformers are not linear devices. Neither are rectifiers. But with sufficient PSRR of the amplifier circuit, neither really matters. There's also a tradeoff around the amount of power supply capacitance in the power supply.
And, of course, much of this is moot when a regulated SMPS is used.
Tom
I have no clue how to even start. I can see that there are two technical questions whose answers we'll need to find out
- How does the waveform of the amp output change (change in harmonics, etc) if the supply rail voltages drop for a few seconds.
- How does the VA rating of the PSU transformer correlate with the change in supply rail voltage if the signal level increases for a few seconds, "all else remaining the same"
As long as the amp is operating below clipping its output voltage will be unaffected by anything that happens at the power supply. Whatever perturbation that occurs at the power supply will be attenuated by the PSRR.
Higher VA rating -> beefier wire -> lower loss -> less voltage droop. That's about it.
Why?
Tom