Still rocking my DSP box on my speakers. Still sounding very good. I am sure it could be better though. Time to warm up the snake oil!
Lots of thoughts on this project, such as how could I sell something people might want to buy. Likely not for profit (just to cover my expenses from doing it), at least at first until I can create a range of boards with ultra low jitter clocks and coax cable interconnects for all the signals, then we will be into the realm of proper hi-fi.... Some ideas I have going forward for now: Just fit another flash chip to the XMOS board. I STILL cannot get the stupid thing to work with the current flash chip. I think changing to another device is possibly the easier option, unless that doesn't work as well. I did get some more though. Or just build a board with the XEF device on, as that has the flash built in. I/O boards. At the moment all the designs are for the signal processing side, there is no easy option for terminating to Phono/RCA sockets or XLR/balanced jacks. With the smaller uDSP and uCodec boards would be easy to make a motherboard for these with connectors on. But it would limit the flexibility of the system going that route. Still thinking on the digital video option, would be nice if there was an easy way to get digital audio in from a royalty free interface like DisplayPort and use this to create a preamp/processor solution. More of a dream still, but there must be a way... On a side note had a job interview with a very interesting company recently. Good to see a company that not only makes interesting products, but wants to deliver the best they can and keep improving as technology gets better.
0 Comments
Still enjoying my active setup, when it is plugged in.
Had electrical work done so felt like I have moved house several times and not gone anywhere. What I have learned is that an active 4 way speaker is more complex to plug back in than a passive setup... My conclusion is that active is still better than passive. Even just at a basic level, if you crank it right up, the amps hardly get warm. The AK4458 has more noise on the output, but I am still using the ADAU1966 filter board. I have mostly forgotten what I was going to do next, it's all in a box somewhere around here... Oh that's right, the XMOS board wouldn't see the flash chip. You can change the port used for the data lines, and on any port the data pins 1 and 2 didn't seem to work (0 and 3 did). Not really sure on that, unless I damaged the chip when soldering. Unlikely to kill 5 sets of output pins though! Looking online other people have had issues getting flash chips to work. I did have some alternatives to try. Which must also be in a box somewhere........ I don't normally "blow my own trumpet"..... But this is the most incredible, loudest, and just everything under the sun awesome thing every I have heard.
I have been in the SRE listening room at B&W when they were based at Steyning, listening to some of the most incredible speakers on the planet, in the room they were designed and tuned in. And I have every respect under the sun for the incredible sounding products they produced. However, me, a single person (not quite so amazing in lockdown), spent many years building some modified ZDT 3.5 speakers (or that should be "ZDT 3.5" speakers). They are close but the cabinets are different. I have built both the original passive crossovers (tuned by my own tone deaf ears), and also my custom designed and built 192Khz/24bit digital crossovers. These speakers are awesome with the passive crossovers. But I DARE anyone to try connecting them, through custom crossover hardware, through 8 channels of decent amplifiers you can find (original or fully refurbished commercial amps like my ones), and then compare to the 2 channel passive setup. I am no true audiophile. I could not care less if you use the cheapest aluminum electrolytic capacitors everywhere, or use polyester capacitors instead. I will not hear the difference (and if you can, "hear a difference", maybe you should just go outside for a while and get some fresh air...). But the difference between a passive 2 channel active setup, and switching to fully active 8 channel amplification, there is no contest. 8 channels of amplification, dedicated to the required frequency range of each driver, you will NOT match this with a passive setup. I have a pair of Bowers & Wilkins CM10s here, in fact, whilst they have the "wrong" bass drivers (the spiders on the bass drivers are tighter than the actual commercial versions), they are an amazing pair of speakers, But the fact I can put a pair of homebuilt 3.5 way speakers. with 8 channels of commercial amplification, and they sound as good, or better in many cases, just shows what you can do for the the same price if you put a little effort in yourself. So the noise is the attenuator box I built. I mean it is sat right next to the TV.
Well that was a simple fix! (Removing it that is). Well this thing does sound damned good, but I hadn't done my "can I hear it" test...
How close can you get to the speaker before you hear anything. Well on the mid-range there is a bit of noise up close (1/2 meter or closer). Only when my plasma TV is on though. Interestingly it's exactly the same noise my you can hear the TV making anyway. The output buffer board is the one I designed for the ADAU1966. The AK4468 recommends different filters, and much faster op-amps. The ADAU1966 app note has 10MHz parts, I am using 8MHz. The AK4468 recommends 80MHz parts from Texas Instruments! So I may just design a new buffer board. I should have done it anyway. There are no words or pictures to describe this.
So the PCM4202 and AK4468 are running at 192kHz/24bits. This gives me goosebumps. I am happy. So I pulled out my crossover box.
I also built up an AK4458 DAC board with an AK4468 fitted to add some VELVET SOUND (they are pin compatible). This is now fitted, but not quite working yet! Need to get all the clock rates right, but it should run at 96kHz and possibly 192kHz all in hardware (pin control) mode. Unlike the ADAU1966 which was stuck at 44.1/48kHz in hardware control. I have tuned up the termination resistors a little. Output pins from the ADAU1452 look fine at 75Ohms over the ribbon cables. The NB3L553 clock fan out seem better with 47Ohms. However the signal at the output end is not perfect as you can see some reflections interacting with the signal. That seems to be feeding visible jitter onto the clocks. At the receiver end, where the signal "slopes off" a little at the top as in the picture above, you can see the signal moving around on the oscilloscope. It's different for each board so the PCM4202 looks different compared with the AK4468. So it's all down to compromise (unless you want to tune everything perfectly, which I might do sometime...) It's giving me some more ideas for other designs. Would be interested to play with some designs with U.FL connectors/cables instead of ribbon cables. Try and reduce any interaction between signals. Plus U.FL has a small footprint so might work quite well. Also look at adding some better driver ICs in (or better clock fan outs). I did have some nice ones I had found that are designed to reduce jitter. Have used them before and they are very very good. Although it will all add complexity and cost, and I really doubt I would hear a difference, and they are a PITA strange PCB footprint... Standard power up tests:
Any loud noises on power up? Nope. Any smoke? Nope. Then we are GOLDEN. xTIMEcomposer studio see's something, but also throws up errors about connecting to xmos.com every time I try to do something useful. There is always something that just gets in the damned way of something just working! uDSP and uCODEC above.
Both power up without smoke. uDSP talks to SigmaStudio, I have fitted an ADAU1450, but the crossover program for my ADAU1452 loaded onto it (I assume the unused registers are just ignored), it was putting out the expected clocks. It's missing the EEPROM - ordered from both Digikey and Farnell and forgot to get a SOIC-8 EEPROM. Need to test uCODEC with clocks. But I need to get some tall 2.54mm headers, then I can stack them as intended. Some people use programmable electronic loads, I use a resistor and a burnt out mains switch. Well it does the job.
That's my new CoreOne-xCORE200 board, only had it for over a year waiting to build. Less cramped and components on both sides, easier to build, and has the thermal reliefs I need on the ground plane. So far I am just testing the PSU to death (literally). Blew up the first two RT7285C switch mode regulators. Had the bootstrap diodes on, and then turned the voltage up too high. Ooops! Diodes of death,... Even starts up at 4.5V without them fitted, at least at room temperature, any load. Would recommend leaving them off unless you know better or you may fry the whole board. The Failure mode is to short input to output, dumping full input supply onto your 1V supply rail, the diodes survived though!. Luckily I only put 15V on it. All the ceramic capacitors are rated 16V or higher. The diodes were removed, RT7285C replaced. The power LED blew up as well, removing lifted a pad off the PCB, did the same just soldering a component on a uCODEC board. Not impressed with the JLC PCB 4 layer quality compared with DirtyPCB, but they are cheaper, 2 layer seems fine. JLC missed a routed slot off one set of boards, except the same foot print is on several boards and the others are fine. Nothing wrong in the data I sent either from what I can see. Might do 4 layer from DirtyPCB and 2 layer from JLC if the cost is okay with shipping the next time I order anything. One other issue, the RT7285C is really not stable with low impedance aluminium Electrolytic capacitors. Even with just one on the output the start up behaviour is horrible and not monotonic as the XMOS chip needs. So they will all be not fitted on this board - oh no, think of the reduction in audio quality on my digital signals... With just ceramic capacitors the performance is remarkably good. Startup behaviour is clean. The 1V rail "glitches" around 50mV max with a 0A to 1.5A load switched on, before settling within 10 to 20mV of it's no load voltage, 3.3V with 1A load is similar. Very impressed with performance, but concerned over the failure mode of input to output. Maybe need to add some polyfuses, and TVS diodes, or some post linear regulators to be safe on the next version. But you don't know this stuff until you know. Best way to find out is to put your money where your mouth is, or into the burnt out components that need replacing. Can't beat real world testing. |
Paul JanickiAn electronics engineer and a long term electronics hobbyist. I like tinkering with stuff and making things. Archives
December 2020
Categories |