Thanks!
Thank you DPA,
This is helpful and provides some reassurance that I was on the right tracks.....As I am looking at a commercial application which performs multiple functions within the FPGA....Crossovers, Eq, limiters, room / speaker measure and correction, delay plus others, and I need up to 16 channels, so its well worth the serious investment in R&D....In the long run way cheaper than buying 16 channels of all these different functions and attempting to assemble them into an active studio system....
Thanks again and all the best
Derek.
Thank you DPA,
This is helpful and provides some reassurance that I was on the right tracks.....As I am looking at a commercial application which performs multiple functions within the FPGA....Crossovers, Eq, limiters, room / speaker measure and correction, delay plus others, and I need up to 16 channels, so its well worth the serious investment in R&D....In the long run way cheaper than buying 16 channels of all these different functions and attempting to assemble them into an active studio system....
Thanks again and all the best
Derek.
Conceptually I like the idea however I don't think this is the sort of thing that happens over night. That's not to discourage one from doing such a thing.
The ideal DIY scenario would be someone who has some knowledge of programming a chip like a Spartan6, sell boards with a flashed chip designed to perform filter and waveshaping functions. Allowing the end builder incorporate additional components - power, output state, etc, however they like.
The main goal here is to replace a conventional DAC chip with an FPGA.
The ideal DIY scenario would be someone who has some knowledge of programming a chip like a Spartan6, sell boards with a flashed chip designed to perform filter and waveshaping functions. Allowing the end builder incorporate additional components - power, output state, etc, however they like.
The main goal here is to replace a conventional DAC chip with an FPGA.
Last edited:
If you really want to get the absolute most out of the implementation (and if you're going to go through the trouble of a bespoke solution, I'd hope that's the ultimate goal!), then you really need to have absolute control over the entire design. Leaving parts of it up to the builder seems counter that goal.
Overkill--I don't think your stated aims really necessitate a custom DAC. The advantage in the latter is the control of the implementation of the DAC itself, rather than large-scale integration. I'd be wont to spend my money on a modest, albeit quiet desktop computer with a pro-level soundcard and pay someone to configure it appropriately for your workflow.
Overkill--I don't think your stated aims really necessitate a custom DAC. The advantage in the latter is the control of the implementation of the DAC itself, rather than large-scale integration. I'd be wont to spend my money on a modest, albeit quiet desktop computer with a pro-level soundcard and pay someone to configure it appropriately for your workflow.
Thanks again DPA,
I'm happy the bespoke FPGA route is the best choice.
The end product will not be for my own private use or DIY, pure commercial ROI at the end of the day and when combined with my active loudspeakers it makes for a very compelling high performance package.
Looks like its a good 12 month R&D....Even working with a top flight designer who has track record here....He has completed a very ambitious 64 channel DAW incorporating FPGA's and whole bunch of other goodies!
Anyway time will tell and thanks again.
Derek.
I'm happy the bespoke FPGA route is the best choice.
The end product will not be for my own private use or DIY, pure commercial ROI at the end of the day and when combined with my active loudspeakers it makes for a very compelling high performance package.
Looks like its a good 12 month R&D....Even working with a top flight designer who has track record here....He has completed a very ambitious 64 channel DAW incorporating FPGA's and whole bunch of other goodies!
Anyway time will tell and thanks again.
Derek.
Derek,
A lot of the functions you have listed there are better served with DSP. You can brute force them with loads of logic gates, but you start needing big chips and big R&D budget for no real gain when grunt is cheap. Of course to confuse things you can put DSP cores in a lot of FPGAs these days so the line between hardware and software gets blurred very quickly. But I would seperate processing and conversion into two problems and pick the right tech for each.
And if you are making less than 10,000 of them a year license as much as possible otherwise you burn 100k in R&D for something that's available for pennies or open source.
A lot of the functions you have listed there are better served with DSP. You can brute force them with loads of logic gates, but you start needing big chips and big R&D budget for no real gain when grunt is cheap. Of course to confuse things you can put DSP cores in a lot of FPGAs these days so the line between hardware and software gets blurred very quickly. But I would seperate processing and conversion into two problems and pick the right tech for each.
And if you are making less than 10,000 of them a year license as much as possible otherwise you burn 100k in R&D for something that's available for pennies or open source.
It's all about the right tool for the job, and then limiting using too many different tools for a design....
FPGA's are great for bit processing and fixed filter functions, perfect for a DAC like the dam1021. But if you want to implement more complex processing, like a crossover, use a DSP as they much cheaper than a large FPGA.
FPGA's are great for bit processing and fixed filter functions, perfect for a DAC like the dam1021. But if you want to implement more complex processing, like a crossover, use a DSP as they much cheaper than a large FPGA.
Agreed. The simplicity factor is what makes such a design so appealing. Start adding too much and things get weird and bloated. Soren, I own a 1021 which is up for sale currently. I simply don't have time/confidence to dedicate to completing it even though it's 90% there. It's a design I'd like to come back to at some point. Alternatively, I hope you continue exploring this realm as your work is really a gift to the community.
Commercial DAC chip converter versus an FPGA based converter mostly boils down to a choice between converter linearity versus converter quantization noise. An digital FPGA based converter will necessarily be 1-bit. Which means that it can natively produce only two quantization levels. As such, 1-bit converters produce a great deal of quantization noise. Sigma-delta modulation is a DSP technique which can greatly reduce the in-band quantization noise by relocating it out of band. What 1-bit converters offer is perfect integral linearity, as two points can only define a perfectly straight line. 1-bit converters, of course, are nothing new.
Native multibit DAC chips feature more than two quantization levels. To be multibit, such converters need 2 or more bits of native resolution. The more native bits, the less can be the quantization noise. However, having more than two levels means it's possible for non-linearity to result. What multi-level converters offer is a greatly reduced quantization noise. Most audio DAC chips today combine multiple levels with sigma-delta modulation, in an attempt to realize the best of both worlds.
There seems to be two motivations for FPGA based 1-bit converters. The first reason is that FPGAs enable a massively parallel 1-bit converter implementation. Very basically, many 1-bit converters output the same bitstream slightly time skewed from each other. The other reason, I suspect, is that the inclusion of an FPGA has become a marketing bullet with audio consumers. Consumers are constantly looking out for who may have the secret technical sauce to delivering musical satisfaction. FPGAs are the latest ingredient in that technical sauce for some designers.
Last edited:
I agree too.
Basically ring dac uses many 1-bit converters in a random way to distribute noise and to decorellate it from input signal.
Then using an fpga is economically more convenient because usually:
1) you have a great idea for a new tipe of dac
2) you simulate in your lab on a fpga
3) you build a custom chip and mass produce it
On point 3) because hi-end producers build few pieces it is not conveniente to build a custom chip so (incredibly) it becames more convenient to put the fpga directly on final board. Then marketing department tells that fpga "is the best"
The issue driving FPGA use in production is much more economic than it is convenient. The problem with an custom ASIC is it's multi-million dollar fixed cost (wafer mask sets, etc). FPGA doesn't entail those huge fixed costs, however, it does entail a higher per unit cost. The deciding factor is the expected total production unit volume. At some high unit volume, the per unit cost savings of an ASIC justify it's higher fixed cost. Which is why, as you indicated, smaller volume products utilize FPGA. As far as product marketing departments are concerned, having their own custom ASIC in their product carries more market cache than having an FPGA. The huge fixed cost difference for low volume products simply overrides any such marketing luxury.
Member
Joined 2009
Paid Member
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.