Hi, I've had this project on the cards for a long time in my spare time. I'm a professional engineer working in Industrial and Defense electronics in my day job, but I've always wanted to make hifi amplifiers. I'm from somewhere between Gen Y and X, so I've been on a digital only diet my whole life and have never really been interested or convinced by escoteric analog arguments. And I'm usually pretty broke. My budget is actually quite pathetic.
I'm calling it the Kraken Amp for now. The 8 armed tentacle monster. It's nearly done.
So my goal here is to:
1. Make a 600W amp that can drive 8 single ended loudspeakers, or 4 in BTL, or a combination of both.
2. Make it digital only with really high PSRR, snr and no loss in analog stages
3. Make it a hifi in a box style of product, so that it can be used in a lot of situations like home entertainment, not a hifi only setup
4. Make it stream. UPNP, Spotify, Internet Radio, etc
5. Make it inexpensive and compact, leverage technology, find 'sweet-spots' and use tricks to achieve cost goals.
I hope to Kickstart later this year, but can you please give me your opinions. Hit or sh!t? How do I make it more appealing?
600W Class D Amp
2x TAS5624A's driven by a TAS5558 Sigma Delta modulator. The latter is important because the algorithms therein are magic. The 5558 has 8 outputs for two 5624A's. Each of these quad H bridge parts drive an 8ohm BTL load 150W @ 10% THD, 70W @ 0.1% THD, 60W @ 0.02% THD. Passive parts and design are taken from a reference design.
Power Supply
Class D Amps have great power supply noise rejection. Its a huge advantage over traditional designs, but to make it better Kraken has it's power split into four channels, two for each 5624 chip. Each has it's own path, so there is less interference between channels. Each path has it's own terminals, reverse polarity protection, over voltage protection to 70V (20~42V is nominal input), inrush protection and over current protection. An LDO in the path regulates a dirty input providing +60dB ripple rejection, and passives including a large common mode choke provide another 10~40dB depending on the frequency of interest. Combined with the PSRR of the 5624's @ >64dB power quality should be superb.
What's the point of so much ripple rejection? That lets you use a low cost supply like this one: LS150-36 for $37 per channel, or $148 for a 600W total solution. And it should be completely silent.
Digital Only
Bluetooth audio from Andriod, (iHope) iPhones, and other BT sources, USB Audio, HDMI Audio instead of analog inputs (perhaps pass through but that's a bit harder,) PCM/I2S through a discreet connector, and... a Raspberry Pi. The latter has I2S output and out-of-the-box projects like Rune Audio give you a web interface from your Smart Phone. You can use your browser to any stream any internet or network source to the system - like Spotify which supports HD Audio streaming. I've tested the system with Rune and Volumio and the I2S playback works well. With some software work it could be seamless but I have to convince those guys to support me too.
No SPDIF? HDMI seemed more of a future proof choice and has superior audio bandwidth.
No Analog? Again HDMI can be found on every CD, DVD, Blu-Ray source and our lives are so Smart Phone oriented now that these seemed appropriate. Using a laptop or PC as a HD Audio source is possible through USB Audio and everything is asynchronously resampled. I think that people who love their analog are not going to buy this system, so I'm not going to slow the project trying to support everything/everyone.
More than one Kraken can be daisy chained, so if you want more power, you can use one for each cabinet. Phased array systems are possible this way too, but the system doesn't have the DSP capability to delay/store a signal. Because it accepts a discreet 3.3V or 5V I2S/PCM input, there is no reason why you couldn't connect it to any FPGA or DSP system for doing this sort of fun work.
Inexpensive
The goal is $400 for the base mode on Kickstarter.
The class D components are affordable, and only demand reasonable sized passive parts. The design is uncomplicated and made from commodity parts on a modest 4 layer circuit board. The design will have a few surprises to drive down cost in other ways, but I won't mention them until I've finished testing and know they are production ready.
It's a cheap amp, but the real cost saving in your system is:
- a really low cost power supply solution
- allows active cross overs - no expensive cross over components
- no pre-amp, it's a single box solution
- good use of your phone and internet streaming of HD Audio, reducing source cost
Because you can have 4, 6 or 8 outputs in this system, and because each can have its own biquad filter, active cross overs can be achieved. This can all be achieved through software setup either through USB or (I hope) through a web interface when a RasPi is fitted.
I especially want this product to appeal to guys making their own cabinets and OB designs, and intend it to be inexpensive enough to experiment with, and embed in a cabinet design.
Please let me know your thoughts and suggestions. Prototyping in August at this rate. thanks, Brent.
I'm calling it the Kraken Amp for now. The 8 armed tentacle monster. It's nearly done.
So my goal here is to:
1. Make a 600W amp that can drive 8 single ended loudspeakers, or 4 in BTL, or a combination of both.
2. Make it digital only with really high PSRR, snr and no loss in analog stages
3. Make it a hifi in a box style of product, so that it can be used in a lot of situations like home entertainment, not a hifi only setup
4. Make it stream. UPNP, Spotify, Internet Radio, etc
5. Make it inexpensive and compact, leverage technology, find 'sweet-spots' and use tricks to achieve cost goals.
I hope to Kickstart later this year, but can you please give me your opinions. Hit or sh!t? How do I make it more appealing?
600W Class D Amp
2x TAS5624A's driven by a TAS5558 Sigma Delta modulator. The latter is important because the algorithms therein are magic. The 5558 has 8 outputs for two 5624A's. Each of these quad H bridge parts drive an 8ohm BTL load 150W @ 10% THD, 70W @ 0.1% THD, 60W @ 0.02% THD. Passive parts and design are taken from a reference design.
Power Supply
Class D Amps have great power supply noise rejection. Its a huge advantage over traditional designs, but to make it better Kraken has it's power split into four channels, two for each 5624 chip. Each has it's own path, so there is less interference between channels. Each path has it's own terminals, reverse polarity protection, over voltage protection to 70V (20~42V is nominal input), inrush protection and over current protection. An LDO in the path regulates a dirty input providing +60dB ripple rejection, and passives including a large common mode choke provide another 10~40dB depending on the frequency of interest. Combined with the PSRR of the 5624's @ >64dB power quality should be superb.
What's the point of so much ripple rejection? That lets you use a low cost supply like this one: LS150-36 for $37 per channel, or $148 for a 600W total solution. And it should be completely silent.
Digital Only
Bluetooth audio from Andriod, (iHope) iPhones, and other BT sources, USB Audio, HDMI Audio instead of analog inputs (perhaps pass through but that's a bit harder,) PCM/I2S through a discreet connector, and... a Raspberry Pi. The latter has I2S output and out-of-the-box projects like Rune Audio give you a web interface from your Smart Phone. You can use your browser to any stream any internet or network source to the system - like Spotify which supports HD Audio streaming. I've tested the system with Rune and Volumio and the I2S playback works well. With some software work it could be seamless but I have to convince those guys to support me too.
No SPDIF? HDMI seemed more of a future proof choice and has superior audio bandwidth.
No Analog? Again HDMI can be found on every CD, DVD, Blu-Ray source and our lives are so Smart Phone oriented now that these seemed appropriate. Using a laptop or PC as a HD Audio source is possible through USB Audio and everything is asynchronously resampled. I think that people who love their analog are not going to buy this system, so I'm not going to slow the project trying to support everything/everyone.
More than one Kraken can be daisy chained, so if you want more power, you can use one for each cabinet. Phased array systems are possible this way too, but the system doesn't have the DSP capability to delay/store a signal. Because it accepts a discreet 3.3V or 5V I2S/PCM input, there is no reason why you couldn't connect it to any FPGA or DSP system for doing this sort of fun work.
Inexpensive
The goal is $400 for the base mode on Kickstarter.
The class D components are affordable, and only demand reasonable sized passive parts. The design is uncomplicated and made from commodity parts on a modest 4 layer circuit board. The design will have a few surprises to drive down cost in other ways, but I won't mention them until I've finished testing and know they are production ready.
It's a cheap amp, but the real cost saving in your system is:
- a really low cost power supply solution
- allows active cross overs - no expensive cross over components
- no pre-amp, it's a single box solution
- good use of your phone and internet streaming of HD Audio, reducing source cost
Because you can have 4, 6 or 8 outputs in this system, and because each can have its own biquad filter, active cross overs can be achieved. This can all be achieved through software setup either through USB or (I hope) through a web interface when a RasPi is fitted.
I especially want this product to appeal to guys making their own cabinets and OB designs, and intend it to be inexpensive enough to experiment with, and embed in a cabinet design.
An externally hosted image should be here but it was not working when we last tested it.
Please let me know your thoughts and suggestions. Prototyping in August at this rate. thanks, Brent.
Hi,
I've seen your design elsewhere and i'm pretty interested in doing the same kind of thing. Here's what I see, though...
First: I want analog input. I understand most sources are digital but the only output is analog for all sort of things that I may want to hook up. For example, I've got a media room with a 7 channel receiver and 7 speakers that I can't connect to it?
Second: this is way too much money. I can go to Amazon and buy the equivalent pieces for under $200 so paying $400 for a single board with less versatility is a non-starter.
Third: most speaker systems (mine included) use tweeters, mids, and woofers - each with their own power requirements. Using a high powered amp for a tweeter seems like a bad thing - your noise floor will be higher than it needs to be and the sweet spot for distortion is too loud.
Fourth: if you're running Bluetooth etc into the I2S then there's no control. How do I remotely set the EQ for this system? One of the huge pluses of a complex electronic multiamp is real-time control over the crossover and eq curves. I think this needs a processor.
Just my first thoughts.
Mark
I've seen your design elsewhere and i'm pretty interested in doing the same kind of thing. Here's what I see, though...
First: I want analog input. I understand most sources are digital but the only output is analog for all sort of things that I may want to hook up. For example, I've got a media room with a 7 channel receiver and 7 speakers that I can't connect to it?
Second: this is way too much money. I can go to Amazon and buy the equivalent pieces for under $200 so paying $400 for a single board with less versatility is a non-starter.
Third: most speaker systems (mine included) use tweeters, mids, and woofers - each with their own power requirements. Using a high powered amp for a tweeter seems like a bad thing - your noise floor will be higher than it needs to be and the sweet spot for distortion is too loud.
Fourth: if you're running Bluetooth etc into the I2S then there's no control. How do I remotely set the EQ for this system? One of the huge pluses of a complex electronic multiamp is real-time control over the crossover and eq curves. I think this needs a processor.
Just my first thoughts.
Mark
A pure digital amp might be the cheapest solution, theoretically. Practically digitally PWM is multiplied with power supply voltage as the "DAC reference". Thus you need a very clean power supply to obtain noise margins of 100dB or more.
Last edited:
Hey Speaky, thanks for your reply. All good points.
There are two PCM/I2C ports to external equipment, for expansion or daisy chaining amps. One is marked Input and the other Output, but they have both channels on 3.3V signals (not ethernet) over an ethernet connector, for standard cat5 eth cable. This is because everyone has them in their draw, so I'll use them off label for expansion.
The great thing here is it lets you design an analog board that plugs in. I was thinking a Wolfson WM8731 stereo analog to I2S would be my first trial. I was also going to use these ports to add a control panel with leds, dials, buttons, etc, but anyone can make a quick board.
Another important thing - this is going to be open source. I'll release the schematics and circuit board in Altium format, as well as the software for the PSOC processor. The RasPi project i talk of: rune and others, are all open source and well supported.
You are right, but there are many surface mount parts to achieve a design like this. A lot of the cost is in the 4 layer PCB and the assembly, but also the cooling design, etc, that allows it to be very compact. $400 was including all assembly of all parts + testing, but I'm wondering if I can offer the raw PCB for sale too... I would not want to support it though, as you'd have to be a soldering ninja to put it together without failure. It's very integrated and busy.
The model for pricing is cost +70%. So I'm expecting my finished board (parts, PCB, placement, packaging, & don't forget testing) to cost $235. The 70% is insurance in case things go wrong, cost of money, some of my time as I'd have to take time off work to fulfill the orders, etc.
Being open source, you can absolutely go and make the design yourself from my files, but if I can Kickstart a 100 or so boards together, there is bulk advantage.
Not sure I accept this one. Remember that the sources are all digital, so there is optimal dynamic range and (should be) no noise floor from analog sources. Equalisation and low sensitivity tweets might be necessary - it's going to take a listening test I think. Running these class Ds at lower power means very low THD - 0.02% for this part.
I'm planning it to support two cabinets, each has a BTL + 2x SE. 150W for sub, 75W mid, 75W tweet. The latter is overkill, so volume control is necessary. Active cross over biquads on each channel, no passives.
That's all covered. It has a PSOC3 (or PSOC5LP if needed) on board. This part handles USB Audio, programming biquads/settings through USB, plus all the house keeping. Control can come through RasPi as well (but the system can work without RasPi) through web interface and I'm talking to some folks about support for Kraken. Either can be used to program the biquad filters in the TAS5558. These allow all active crossovers, volume control, mute, equalisation that you like. I have some plans for control, but will hold off announcing them in case they don't work. Needs testing.
The psoc has some light dsp capability even (beyond programming the TAS5558 that has extensive DSP), but what it won't allow is delay lines for making a phased array. I'd have to add SRAM here and I'm out of pins for parallel, and I've not done the math to see how hard a serial SRAM solution would be - I'm going to wait and see if there's a request/requirement before spending this time.
There are two PCM/I2C ports to external equipment, for expansion or daisy chaining amps. One is marked Input and the other Output, but they have both channels on 3.3V signals (not ethernet) over an ethernet connector, for standard cat5 eth cable. This is because everyone has them in their draw, so I'll use them off label for expansion.
The great thing here is it lets you design an analog board that plugs in. I was thinking a Wolfson WM8731 stereo analog to I2S would be my first trial. I was also going to use these ports to add a control panel with leds, dials, buttons, etc, but anyone can make a quick board.
Another important thing - this is going to be open source. I'll release the schematics and circuit board in Altium format, as well as the software for the PSOC processor. The RasPi project i talk of: rune and others, are all open source and well supported.
You are right, but there are many surface mount parts to achieve a design like this. A lot of the cost is in the 4 layer PCB and the assembly, but also the cooling design, etc, that allows it to be very compact. $400 was including all assembly of all parts + testing, but I'm wondering if I can offer the raw PCB for sale too... I would not want to support it though, as you'd have to be a soldering ninja to put it together without failure. It's very integrated and busy.
The model for pricing is cost +70%. So I'm expecting my finished board (parts, PCB, placement, packaging, & don't forget testing) to cost $235. The 70% is insurance in case things go wrong, cost of money, some of my time as I'd have to take time off work to fulfill the orders, etc.
Being open source, you can absolutely go and make the design yourself from my files, but if I can Kickstart a 100 or so boards together, there is bulk advantage.
Not sure I accept this one. Remember that the sources are all digital, so there is optimal dynamic range and (should be) no noise floor from analog sources. Equalisation and low sensitivity tweets might be necessary - it's going to take a listening test I think. Running these class Ds at lower power means very low THD - 0.02% for this part.
I'm planning it to support two cabinets, each has a BTL + 2x SE. 150W for sub, 75W mid, 75W tweet. The latter is overkill, so volume control is necessary. Active cross over biquads on each channel, no passives.
That's all covered. It has a PSOC3 (or PSOC5LP if needed) on board. This part handles USB Audio, programming biquads/settings through USB, plus all the house keeping. Control can come through RasPi as well (but the system can work without RasPi) through web interface and I'm talking to some folks about support for Kraken. Either can be used to program the biquad filters in the TAS5558. These allow all active crossovers, volume control, mute, equalisation that you like. I have some plans for control, but will hold off announcing them in case they don't work. Needs testing.
The psoc has some light dsp capability even (beyond programming the TAS5558 that has extensive DSP), but what it won't allow is delay lines for making a phased array. I'd have to add SRAM here and I'm out of pins for parallel, and I've not done the math to see how hard a serial SRAM solution would be - I'm going to wait and see if there's a request/requirement before spending this time.
Voltwide, that's not completely true. The Class D's get so much PSRR - power supply rejection ratio - because they switch around 400kHz. At this frequency the ear can't hear the quantisation noise from the sharp edges. That said, the power supply noise gets through the H bridge of the class D chip, but is reasonably squashed by the demodulation filters - ie. the big passive LC filter on the output - and the filtering you put before the class D chip.
That said, there's a lot of additional filtering and insertion loss in each power path to turn a bad supply into a good supply. This includes active regulation that has >60dB rejection of noise at all audible frequencies. 60db means 1 millionth of the noise gets through.
Hi Brent,
I've been going through a build-buy cycle and I'd love a single board solution but here's what it has to compete with. This is one of my possible builds:
WiFi/Ble USB - $15
Cirrus logic Audio Card -$40
Rasp Pi 2 - $40
4 x irs2092 250w - $120. At $30 each aliexpress
This gives me 24 bit 48,96,192 analog input
Spdif I/o
5 band eq crossover in chip
Plenty of HP for the pi2 to do audio processing
Ble 4 input
WiFi direct
Web server
Spare USB port
Total $215
Many of these features I find compelling. Linux on the crossover... sure.
Mark
I've been going through a build-buy cycle and I'd love a single board solution but here's what it has to compete with. This is one of my possible builds:
WiFi/Ble USB - $15
Cirrus logic Audio Card -$40
Rasp Pi 2 - $40
4 x irs2092 250w - $120. At $30 each aliexpress
This gives me 24 bit 48,96,192 analog input
Spdif I/o
5 band eq crossover in chip
Plenty of HP for the pi2 to do audio processing
Ble 4 input
WiFi direct
Web server
Spare USB port
Total $215
Many of these features I find compelling. Linux on the crossover... sure.
Mark
Pardon me, you are correct. 1:1000000 would be for power. 1:1000 for voltage attenuation is correct. I still think that's pretty ok 🙂
Mark, you are right of course. You can do the whole project cheaper, but you are probably not my target market. The guys I want are building cabinets and playing with active crossovers, etc. You sound capable of putting the electronics together as well as I am, and clearly you know your purchasing choices really well - but will you get any time left over to design a cabinet? I'm guessing you already have the speakers but are looking to improve your system through a better amp.
I looked at the Cirrus card for integration and inspiration, but thought it was a tool for DSP and recording on the Pi so left it out. The Pi in my system is a streaming source, an accessory, and not the centerpiece. It might end up being the best place for control software.
I looked at the Cirrus card for integration and inspiration, but thought it was a tool for DSP and recording on the Pi so left it out. The Pi in my system is a streaming source, an accessory, and not the centerpiece. It might end up being the best place for control software.
you might want to look at the beaglebone black. BeagleBoard.org - black
It has i2s, more GPIO, onboard flash (this is actually a pretty big deal - you don't want to be dealing with SD cards)
Oh and this is pretty cool! https://upverter.com/Beagle/afdfe0be7c0bcec5/BeagleBoneBlack/ they have the whole schematic!
I'm a little unclear however if your plan is to have the raspberry pi inside the main amp, or as a external piece. I would discourage you from having the raspberry pi tied to your device. we WILL get a new model, pinouts will change, connectors most likely will change.
Getting support in the various audio player distros will just be a matter of a pull request (providing you stick with opensource stuff)
It has i2s, more GPIO, onboard flash (this is actually a pretty big deal - you don't want to be dealing with SD cards)
Oh and this is pretty cool! https://upverter.com/Beagle/afdfe0be7c0bcec5/BeagleBoneBlack/ they have the whole schematic!
I'm a little unclear however if your plan is to have the raspberry pi inside the main amp, or as a external piece. I would discourage you from having the raspberry pi tied to your device. we WILL get a new model, pinouts will change, connectors most likely will change.
Getting support in the various audio player distros will just be a matter of a pull request (providing you stick with opensource stuff)
I've got a BB and do really like it much more than my pi, but am constrained by the audio card. I'm still waiting to test the CLAC and the pi2 - my pi2 shows up tomorrow. If I love the CLAC but can't figure out the i2c I'll just leave the pi in the crossover.
Brent, well, good luck with your project. If you can end up in the <= $250 bracket I'd be interested. In any case I'm curious to see the actual performance of the chips.
Brent, well, good luck with your project. If you can end up in the <= $250 bracket I'd be interested. In any case I'm curious to see the actual performance of the chips.
Beagle board is great, but Pi just has the numbers behind it. The difference in processing is huge, but doesn't seem to matter - Pi has the support and the momentum. Now that the Pi2 is out it's less of a performance gap too but I haven't really checked.
Yes Beagle is open source and that is commendable.
On the Kraken the PI is a bolt on, and I would kick start it on a 'bring your own PI' basis. Or bundle with as a perk.
I'll finish the job and post the results. No doubt Rev 1 will have warts. But by Rev 3 I could have a nice system. Getting to $250? Depends on how many supporters I get. Making 1000 boards? I'm confident I could get close. Making 100? Probably not.
Yes Beagle is open source and that is commendable.
On the Kraken the PI is a bolt on, and I would kick start it on a 'bring your own PI' basis. Or bundle with as a perk.
I'll finish the job and post the results. No doubt Rev 1 will have warts. But by Rev 3 I could have a nice system. Getting to $250? Depends on how many supporters I get. Making 1000 boards? I'm confident I could get close. Making 100? Probably not.
Brent,
This looks like a very interesting project. It has many of the '2015 features' (digital) I like to have. And it sounds like you're talented and confident enough in your design. While I'm a huge supporter of KickStarter projects, I've found that the real end cost is often tied to issues both with the engineering and production of the final design. Consequence: usually a higher, not lower cost - at least initially. I'm still waiting for my Soap router! 😀
But that fact doesn't always put me off. If I'm convinced a class-act product will arrive and be a device I'd like to showcase, I'll stick with it.
Please keep us update with your project and when the project goes live on Kickstarter.
Rick
This looks like a very interesting project. It has many of the '2015 features' (digital) I like to have. And it sounds like you're talented and confident enough in your design. While I'm a huge supporter of KickStarter projects, I've found that the real end cost is often tied to issues both with the engineering and production of the final design. Consequence: usually a higher, not lower cost - at least initially. I'm still waiting for my Soap router! 😀
But that fact doesn't always put me off. If I'm convinced a class-act product will arrive and be a device I'd like to showcase, I'll stick with it.
Please keep us update with your project and when the project goes live on Kickstarter.
Rick
Hello
maybe you interested in a free GPL software solution for a 4 way - crossover.
You can see my setup here:
diyAudio server HTTPS page
HIFI-FORUM » Do it yourself » Lautsprecher » Aktive Frequenzweiche per Standard Software mit GUI
Sorry it is in German
You have
4 x 2 crossover
many parametic equalizer
limiter
....
much more
HDMI 8 Channel output
8 Channel analog output
For activating for example 2, 3 or 4 way loudspeaker
So to the question of the first post. Put a MINI-ITX board with HDMI and onboard sound in a 7.1 HDMI Receiver +SSd + 8 GB memory. Thats it.
Regards
Loafmeat
maybe you interested in a free GPL software solution for a 4 way - crossover.
You can see my setup here:
diyAudio server HTTPS page
HIFI-FORUM » Do it yourself » Lautsprecher » Aktive Frequenzweiche per Standard Software mit GUI
Sorry it is in German
You have
4 x 2 crossover
many parametic equalizer
limiter
....
much more
HDMI 8 Channel output
8 Channel analog output
For activating for example 2, 3 or 4 way loudspeaker
So to the question of the first post. Put a MINI-ITX board with HDMI and onboard sound in a 7.1 HDMI Receiver +SSd + 8 GB memory. Thats it.
Regards
Loafmeat
The PI certainly has the numbers and the mind share. My main reason to caution against the PI is mainly that we can say with certainty that a new one WILL come out, maybe with a different form factor, with more or less ports, and most importantly that you have no assurance that the old one will stay in production
Contrast that to the beaglebone where their first model is STILL in production. BeagleBoard.org - beagleboard (Sadly the price hasn't changed in the last 7 years)
I know folks who have deployed something like 1000+ rpi's in the field. Many people are choosing the RPI as their primary platform and taking the risk of a model going out of production (maybe the rpi folks will see this and keep their older boards in production to appease the industrial/commercial users it would be foolish to leave that money on the table, these guys are happy to pay the premium)
It's of course easy to arm chair when you're not involved. Either is a solid choice - it's never good to bike shed too much 😉 . Good luck!
PSRR in the audio band is unrelated to the switching frequency.
The output devices directly couple the load to the PSU when they are in the "on" state.
The LC filter on the output does nothing to filter out power supply noise in the audio band.
Because of this, PSRR in a class D amp is inherently low.
It is often addressed by feedback, which is what TI have implemented.
Cheers,
Mike
- Status
- Not open for further replies.