Hi Matt - thanks for your detailed suggestions. I might give it a try sometime but right now I'm doing it the old, traditional way of hand-soldering each pin. Due to my earlier experiences with multiple paralleled DACs where I got very noisy sound due to 'failed' chips I'm plotting a more conservative approach where I just solder the two power pins down on each chip first and run a power-on test on each individual DAC. This way if and when I encounter a 'dud', I can much more easily remove and replace it. So far using this method and more than 10 boards soldered up I think I've had to remove 3 chips. I also found one short on the digital side of the chip on one board.
Good news, I PM'd for one kit couple days before, so even my message was lost I can still get this one. Thanks.
Hi @Averagevalley - thanks for your interest. I am thinking opening a brand new thread to talk about my next generation parallel DAC is going to be best. Mainly because there's already so much confusion between the three DAC designs already on this thread (lingDAC, PhiDAC and PhiDAC SE) but also because the design's a bit different from the previous ones in its heavy reliance on transformers. The trafos will probably generate a whole new world of questions which will benefit from a clean sheet. When I get around to opening that thread - hopefully within the next day or two then I'll post a link here.
In the meantime if anyone wants to get building (a lot of soldering is involved with so many chips) here are the gerbers for one of the component PCBs which holds 36 TDA1387s. The first edition of the next gen DAC will most likely use 4 of these PCBs for a total of 144 DACs. Hence I was planning to call it the 'Gross DAC' but I notice there's already a DAC with that name even though it uses the German character for 'ss'.
I completed the first prototype of the NG-gross DAC last night and listening reveals it has a significantly lower noise floor than PhiDAC SE, more low frequency ambient information is apparent which has the effect of drawing me further in to the musical performance. But then again, its a much more complex design, likely needing six PCBs.
In the meantime if anyone wants to get building (a lot of soldering is involved with so many chips) here are the gerbers for one of the component PCBs which holds 36 TDA1387s. The first edition of the next gen DAC will most likely use 4 of these PCBs for a total of 144 DACs. Hence I was planning to call it the 'Gross DAC' but I notice there's already a DAC with that name even though it uses the German character for 'ss'.
I completed the first prototype of the NG-gross DAC last night and listening reveals it has a significantly lower noise floor than PhiDAC SE, more low frequency ambient information is apparent which has the effect of drawing me further in to the musical performance. But then again, its a much more complex design, likely needing six PCBs.
So ... clarify ....
HINT: You can create a TABLE inside a post.
So .... why not create thread (or devote a post or blog entry) into clarifying the difference?!!!!!
HINT: You can create a TABLE inside a post.
Striving for better sound from dacs always seems to lead to fairly complex designs, IME.
I've taken a leaf out of @matt_garman's book and updated the first post of this thread in a similar manner to how he amended his TDA1387 DAC thread. Hopefully it will help to clear up some of the confusion between the different designs.
@Markw4 - yes there are two complementary processes at work in my own designing. There's the continual pull towards complexity yet there's also what's perhaps best called the 'Muntzing' tendency (not as strong as the first one) to reduce a DAC to its bare essentials. Oftentimes those essentials aren't revealed until a DAC has grown to a scarcely manageable level of complexity.
Madman Muntz - Wikipedia
@Markw4 - yes there are two complementary processes at work in my own designing. There's the continual pull towards complexity yet there's also what's perhaps best called the 'Muntzing' tendency (not as strong as the first one) to reduce a DAC to its bare essentials. Oftentimes those essentials aren't revealed until a DAC has grown to a scarcely manageable level of complexity.
Madman Muntz - Wikipedia
Thanks, that's helpful.
I'm in line for a couple of your PhiDAC kits courtesy of Matt and will be interested to get a flavour of your approach.
You may recall that a friend is building a DIY pick and place machine and we plan to use the PhiDAC as a test piece - he's now got the movement control sorted and is now working to solve the suction mechanism so maybe not too far off being able to give it a try.
I agree, but I can't wait for discovering essentials, so subscribing to a "NG-DAC" thread immediately
(I picked-up a name from the message body): Grossly parallel multibit DAC adventuresI have a question. It will be a pure 44.1/48kHz NOS DAC, or allow variations, by example increasing maximum clock rate to let us chose between NOS or 4x oversampling? It could then give a possibility to play 196kHz source files directly. I know from your previous posts, clocking could reach a new level of complexity, I only want to know whether your view in this matter has changed over a time.
Cabbage Patch
Cabbage Patch riots - Wikipedia
If I could get Richard to autograph one for me, I'd fork out some serious loot. Kinda like the Cabbage Patch phenomenon decades ago ... hysteria and violence .... all over some toys.
Cabbage Patch riots - Wikipedia
Don't worry pal, PhiDAC was dirty cheap and you perhaps had a desire, but this time demand will be regulated by a price. Richard already said a word about. I quess, you cannot afford that.
I quess, you cannot afford that.
PhiDAC is clearly a fashion/speculative item in his eyes. How else to explain how its perceived value has skyrocketed since the summer? In July he wanted me to pay him $2 for the privilege of receiving one and now its worth 'serious loot'?
Oh on second thoughts maybe its the power of the autograph he's after. Do you think if I sign a DAC in green pen it'll correct all the errors in the data coming in?
Ding | Dong
That said, NO-ONE can tell a Phi from a Ling from a Ding from a Dong ....
Try this template:
Product: Phi | Ling | Ding | Dong
Price: $2 | $4 | $0.50 | $0.55
Chopsticks included: yes | yes | no | no
Soysauce included: yes | yes | no | no
Etc.
Rich ... sheeeeha man ... yo' stuff so hot, we'd pay top $$s fo' a 'graphed set o' yo' dirty drawers.
That said, NO-ONE can tell a Phi from a Ling from a Ding from a Dong ....
Try this template:
Product: Phi | Ling | Ding | Dong
Price: $2 | $4 | $0.50 | $0.55
Chopsticks included: yes | yes | no | no
Soysauce included: yes | yes | no | no
Etc.
Connoisseurs of extreme irony might wish to delight their eyes with this post : Denafrips Ares
Back to talking about circuits for a moment....
In one sense its a 'pure 44k1 NOS DAC' - that being the frequency response is designed to be as close to flat (at least up to 17kHz) as possible with a 44k1 input clock rate. At any other input rate the frequency response won't be quite flat - the error at 48kHz though will probably be inconsequential. At 88k2 there will be a slight (1-2dB) HF lift, most likely audible as more 'air'. At 176k4 the HF lift will be more pronounced. The DAC chips themselves can run up to 384k though I would expect a worsening of subjective dynamics from such a high rate.
In one sense its a 'pure 44k1 NOS DAC' - that being the frequency response is designed to be as close to flat (at least up to 17kHz) as possible with a 44k1 input clock rate. At any other input rate the frequency response won't be quite flat - the error at 48kHz though will probably be inconsequential. At 88k2 there will be a slight (1-2dB) HF lift, most likely audible as more 'air'. At 176k4 the HF lift will be more pronounced. The DAC chips themselves can run up to 384k though I would expect a worsening of subjective dynamics from such a high rate.
Well, that's really good news. I personally don't see any benefit of going above 192kHz. On my Topping D30 I see no benefits comparing to the 96kHz source files. My interest in 192kHz clock is purely in getting rid of conversion images, but 96kHz (88k2 for a CD sample rate) combined with a quality output filter should give excellent results.
It will be interesting to switch in a Foobar between NOS and 2xOS/4xOS and compare a difference. We don't have such feature in commercial DACs!