Unofficial Mackie User Forums > DL1608/DL806/DL32R/ProDX Mixers

DL1608 Architecture Notes

(1/3) > >>

WK154:
A  word about the DL's architecture and possible improvements that would appear to be possible considering the components before I forget all this. I'll try not to be too techie but some can't be helped.
The diagram on pg. 162 of the manual is really just an old analog interpretation of the software that runs the DL. The only actual hardware shown on the diagram are the preamps in the lower left corner (16 channels worth). The others are the 10 output drivers on the top of the page (L/R , phones (stereo), aux's (1-6). The one D/A converter on the main board (more on the analog section) handles 2 out of 10 outputs and the two CS5368  handle the 16 inputs. All the rest of the block diagram is software logic to create the data flow. Since I didn't take the analog section appart I have no way of knowing what D/A are used for the rest but most likely it's the same part.  These are all 216 kHz parts quite capable of a 192 kHz audio stream.
Let's start with the A/D converters Cirrus Logic CS5368. A 8 channel A/D converter capable of 216 kHz of audio data per channel and there are two of them so the DL can process 16 channels of up to at least 192 kHz of 24 bit data with these chips. It's currently doing 48 kHz. There may be speed-bumps down the road to prevent this but we'll see.
 Since there are no software VC on the preamps a true DCA can not be obtained but a software implemented one can take the values off the A/D converters and based on DCA groups change the values to simulate a DCA. Time for digitally controlled Onyx pre's.
 Next stop the processing (Polarity, HPF, Gate, Compressor, EQ) computations are handled by the Sharc DSP from Analog Devices ADSP21488 a 400 MHz part. The bulk of the routines are most likely AD canned routines for this part. There is no reason why the processing order can't be changed or made more flexible.
 All the data flow is handled by the Blackfin ADSP BF516 ARM Processor the control and heart of the DL it is also a 400 MHz part. The Sharc is more of a coprocessor doing number crunching. The two are in fact purposely designed to play these rolls. Running them is most likely the ucosII or III operating system from Micrium . I can't imagine Mackie taking anything but the easy way out. Lets look at the possible speed bumps.
The Sharc and number cruncher has a tremendous processing capability with i/o that has yet to be used by the current DL code. It's a dual core DSP processor with no less than 65 DMA controllers, 28 Serial interfaces of various types, a 10/100 Ethernet controller (why Mackie included a separate controller SMCA 8710A for this remains a mystery) , MMC, SD and ATA disk controller, FFT, FIR and IIR accelerators. Hardware support for up to 32 ring buffers and stack support as well. Did I forget to mention that it can handle 32 bit floating point, extended 40 bit FP and fixed point 32 bit math and VISA and has support for S/PDIF and PCM. Some of these clearly can't be use by the existing hardware implementation. I don't see a SD slot or any other peripheral connections to take advantage of this or a S/PDIF connector. RTA's certainly a implementable function as is the well debated recording capability. Sending 16 channels of already existing data to the iPad over the pseudo USB via a DMA channel won't stress anything. So is the Sharc at it's limits? Not according to the European Loud GM a former techie who stated that the system is at about 25% of capacity. He has access to way more info than I have.
 Lets move on to the Blackfin. The heart of the DL and the main traffic cop. It handles allocation of resources, communications with the iPad, network and directs the data streams. Doesn't sound like much but its well equipped to handle the tasks. Is it at it's limits? Hardly. So what is keeping this hardware from providing all that it can give? Mackie marketing or code? You decide.

Update: Turns out that my first guess on the OS was right and the DL is run by uClinux. The cost of ucos must have been too much for Mackie.

RoadRanger:

--- Quote from: WK154 on January 12, 2014, 11:03:37 PM --- All the data flow is handled by the Blackfin ADSP BF516 ARM Processor the control and heart of the DL it is also a 400 MHz part. The Sharc is more of a coprocessor doing number crunching. The two are in fact purposely designed to play these rolls. Running them is most likely the ucosII or III operating system from Micrium.
--- End quote ---
Kinda funny, the product I'm working on at the moment has a 50MHz ARM M0 clocked at 8 MHz, that's plenty for it as the 6 channels of A/D are only being sampled at 100Hz and the DSP is just 6 or 12 dB/oct LP. It runs in "standby looking for an 'on' keypress" mode at 32 KHz on a few microamps, I think the battery pack is good for three months in standby. I also remember when uCos came out as a free OS published in "Embedded Systems" magazine if I remember correctly. My OS's are loosely based on it although I always skip the pre-emption stuff for efficiency.

RoadRanger:

--- Quote from: WK154 on January 12, 2014, 11:03:37 PM ---[...] Sharc DSP from Analog Devices ADSP21488 a 400 MHz part. The bulk of the routines are most likely AD canned routines for this part. There is no reason why the processing order can't be changed or made more flexible.
--- End quote ---
I suspect they either don't have an on-staff DSP guy or that he's busy with other products - they have made some suspect statements in the past like the inability to do multitrack with this hardware or that it would be hard to split up the DSP functional blocks. Child's play to any competent DSP guy. As you know any "off the street" programmer can do ARM and iPAD stuff but there is a shortage of DSP guys and they are expensive. They should hire me (or you?) and as a bonus get us to shut up LOL .

WK154:
RR I doubt they could afford either of us and I for one would turn the place upside down. >:D

WK154:
Just as an aside the X32 uses a prior generation Sharc running at 266 MHz two of them. You do have 40 inputs to process instead of 16. It also doesn't have an Ethernet controller and  can do all kinds of processing as demonstrated. The reason for the separate Ethernet  controller on the DL may have been for just the same reason that the new generation Sharc may not have been available when Mackie started but then they switched to the newer part as it became available and did not design out the separate controller. P.S. I'm surprised your 50 MHz ARM runs down at 8 Mhz without problems, a static design?

Navigation

[0] Message Index

[#] Next page