PDA

View Full Version : Experiments of the DAC kind



NRG
25-02-2009, 23:59
Back in 1987 I think it was I purchased my first ‘good’ CD player an original Arcam Alpha. A few years later I upgraded it to ‘Plus’ status which involved the swap out of the DAC board for the Alpha Plus version.

Since then the original board has sat in a draw for over ten years, as it’s based on the TDA1541A I thought it was about time I did something with it! As I’ve been modifying Stan’s excellent 7510 over the last few weeks I thought that the Arcam board would make an interesting comparison and project…if it still worked.

http://i94.photobucket.com/albums/l82/_NRG_/other/Alpha_DAC/DSCF4019.jpg

With the help of Leo (without his valuable advice I’d still be hunting around for solutions) and a rummage through my spares box I drew up an idea to get the thing running. I needed a power supply and a SPDIF receiver chip. The Arcam board accepts I2S directly in the correct format so Leo recommended a CS8412, one of which I already had for an earlier DAC project that never got of the ground.

I knocked up a stripboard based cct just to see if I could get it to work, sourced a transformer from Farnell and a case for mounting everything in. As the Arcam board had its own PSU I just stole volts of it to power the CS8412.
You could have knocked me down with a feather. It worked! Music from both channels and nothing obviously wrong, I let it run for about 24hrs before taking a critical listen. It brought back memories, like wearing an old familiar pair of shoes. The sound was warm and cuddly, lacking any real HF extension and having a deep but wooly bass. So I set out on an upgrade program.

First I upgraded the TDA decoupling caps junking the old Rubicon caps and ceramics for some OSCONs. I replaced the opamp output caps with some FK BlackGates and also replaced the I/V ‘lytics with some Nichicon FG caps. This lifted the sound a little adding some clarity and separation.

Next I replaced the NE5534 opamps with some LME49710 opamps and the sound took on a whole new dimension, I was getting somewhere. Next up I replaced all the regulator output bypass caps with Starget ROD caps and took a leaf out of Guido Tents decoupling article and used a ferrite bead in the -15v supply line replacing the installed inductor.

The -5v reg for the TDA had no decoupling cap and yet the ‘Plus’ board did so I added one. All this helped but I still felt the top end was somewhat veiled and I struggled to make out some low level detail that Stan’s DAC clearly highlighted.

The next change involved supplying the CS8412 with its own regulated digital and analogue +5v supplies. At the same time I changed the diodes on the Alpha board for some Schottkys and replaced the old Rubicon smoothing caps with Panasonic NHG’s. Bingo! These last changes lifted the veil and changed the depth and texture of the sound for the better. The sound has become much more detailed and clear, little things are now easily made out and overall it’s much more engaging.

Comparing to the modified 7510 is an interesting exercise, the 7510 has a lighter, leaner and more airy balance, the TDA1541 is richer and more textured in its presentation, the bass is a tad deeper as well, the overall impression is of a darker sound and yet with all the detail still apparent.

It’s really enjoyable to listen to over a long period of time and makes me want to see how much more I can get from it. I can really appreciate the high praise the TDA1541A commands from its aficionados. The next stage would involve high quality super regulators for the TDA1541 or maybe pre-tracking regs ahead of the LM317/337 which would come in at a lower cost. For now I’m just going to enjoy it!

Pics:
http://i94.photobucket.com/albums/l82/_NRG_/other/Alpha_DAC/DSCF4046.jpg

CS8412 board:
http://i94.photobucket.com/albums/l82/_NRG_/other/Alpha_DAC/DSCF4047.jpg

Modified Arcam board:
http://i94.photobucket.com/albums/l82/_NRG_/other/Alpha_DAC/DSCF4048.jpg

http://i94.photobucket.com/albums/l82/_NRG_/other/Alpha_DAC/DSCF4049.jpg

leo
26-02-2009, 00:11
Good fun isn't it:lolsign: nicework though Neal, I'm impressed

Hey your not tempted to try passive I/V into a tubed output stage next then :sofa:

jonners
26-02-2009, 08:58
Good fun isn't it:lolsign: nicework though Neal, I'm impressed

Hey your not tempted to try passive I/V into a tubed output stage next then :sofa:

Or a discrete solid state active I/V?
Btw, is this running in nos mode?

NRG
26-02-2009, 10:45
Yes NOS...

Don't start me off Leo! I'm happy as it is.....for now. :lolsign:

NRG
26-02-2009, 10:48
I just checked the pics of the 'plus' board currently installed in the player and can see where I made mistakes in modding it many years ago...think I'll apply some of the mods I've recently learnt to it and see how it sounds.

Tony Moore
26-02-2009, 11:19
Nice work Neal and a good write-up! :smoking:

NRG
04-03-2009, 17:41
I've been tweaking and experimenting and made some improvements to the regulator ccts for the CS8412 and also improved the isolation between supplys for the TDA. I went backwards at one stage but retraced my steps and went in a better direction. The sound has moved ahead, the top end is clearer and more open than previously and the midrange has also improved... the sound is very analogue like and has moved ahead of my modified B7510.

I'll live with this for a while (like a day or two!) and then start to implement my plan of pretracking regulators for all supplys.

leo
04-03-2009, 21:31
Not bad eh Neal, amazing how good commercial units can be improved with a bit of work isn't it;)

Anyway I'm pleased to hear yor getting such good results!
I've still yet to hear ANY Delta Sigma based dac beat a well thought out classic multibit based one like the TDA1541A

NRG
12-03-2009, 21:10
Sound continues to improved. Latest additions are a lower noise configuration for the CS8412 regs with the use of two standard green LEDs for the set resistor (R2). This gives a more presice voltage with lower noise and lower o/p impedence.

I've also place a simple gyrator or capacitior multiplier ahead of the CS regs, borrowed the idea from the PFM flea cct...not sure about this one though, could be me but I think it has taken some speed and impact from the sound...needs more listening...as I intened to use one ahead of each regulator stage.

Also added a rough pretracking regulator cct with 'remote sense' for the -15v TDA supply and this looks good on the 'scope compared to before. No gyrator at this time.

Next will be pretracking regs for the -5 and +5 TDA feeds and also pre regs for the CS supply. This will leave the on-board + and -15v regs to supply the opamp and IV cct only.


Pics:

http://i94.photobucket.com/albums/l82/_NRG_/other/Alpha_DAC/DSCF4053.jpg

http://i94.photobucket.com/albums/l82/_NRG_/other/Alpha_DAC/DSCF4052.jpg

freddiecas
16-03-2009, 00:00
have you any links to the circuits for these regulators please?

I'm proposing using LM317/337 for +/-15V to power 18 OPA134s in an active crossover so any improvements would be interesting, thanks :)

NRG
16-03-2009, 15:42
The pre-tracking reg cct is on page 20 of the Nat Semi data sheet for the LM317. You need about 5v of headroom IE 12v out means min of 17v in.

The cct omits the set resistor bypass caps and o/p caps but just add those in. Remote sense of the load is detailed on p9 under "Load Regulation", essentially take the earth connection of the set resitor and bypass cap to the load.

From what I've learnt by first hand experience and also read to get the best from these devices they need to supply something like 30mA minimum (more is better), make the o/p cap a good quality 'lytic but not one that is low ESR, I use Starget ROD here, value 100~220uF. The set resistor bypass cap can be low ESR and value 47~100uF but I also use the Stargets here. Heatsink them well.

freddiecas
17-03-2009, 10:02
thanks, I'm confused about where the second set of leds go, the first replaces R4 in the page 20 diagram? does the second replace the R2 720 ohm?
and then can you tweak the output voltage with the 120 R3?
cheers.

NRG
17-03-2009, 10:51
Ah! my pictures probably caused the confusion.

The LED's are used with two single regulator set for 5v operation feeding the analogue and digital side of the CS8412 but the same principle would apply if it was a pretracking configuration. The diodes replace the normal set resistor (R4 in the pretracking diag). Voltage drop on a std green LED is 1.9v or thereabouts, two in series makes it 3.8v plus the 317 1.25 ref voltage gets you to 5.05v more or less. Lower noise, more accurate voltage setting and lower o/p impedance.

The pretracking reg I pictured is set for -15v, it does not use LED's...for this I used a 13v Zener and 1N4001 diode to set o/p voltage. 13+0.6+1.25 = ~ 14.85v The Zener again replaces R4 in the NatSemi diagram.

Looking at it again I have set R1=R2 (240R is fine), bypass R1, bypass R4 or Zener if you want to try it and place a cap on the o/p of the second 317.

Take the earth for R4/Zener and bypass cap to the load, don't connect to the same earth as C2.

--Neal

freddiecas
17-03-2009, 12:12
ah! I see. I've seen led references for the noise performance before so was intrigued by the pics :)

sorry to bypass your thread away from your dac mods Neal :sorry:

It looks like I'll have to buy another 317/337 and maybe up the voltage on the torroid secondary. A 15-0-15 at 30VA should have given me 4V across a single 317/337 for +/-15V out. Might not be enough for the double version because approaching dropout?

18 x OPA134 at quiescent of 5mA each in a relatively high impedance active crossover should be less than 200mA for the supply I'm guessing.

thanks again, you've helped a lot to get a really good supply design, I had not considered the pre-tracking idea. My "ground plane" will be a thick copper wire across the vero ground outputs of 6 parallel sallen key circuits, so I can sense the ground from the centre point there.

cheers, Fred :)

NRG
17-03-2009, 14:06
No problem Fred, 15-0-15 may just be enough.

Full bridge rec would yield 21.2v, let’s say 20v with 10% regulation. You need 5v head room for the pretracking cct.

But you also need to consider the ripple voltage, so a pretty large cap would be needed as ripple voltage always reduces the Vdc by half of the ripple. At 50Hz and 200mA a nice big 4700uF cap on the bridge o/p would reduce Vdc by about 0.2v.

Your current draw may be a little higher as there is also a constant current flowing through each ‘program resistor’ between the o/p and Vadj, if you go for 240R then you will have 5.2mA flowing for each regulator in addition to the cct draw. So maybe another 22mA on top.

Also take into account ripple current, although not large in this case, as being 4~6 times larger than Adc at approx 1.2A, rate the diodes and ripple capacity of the cap above this to be safe.

Edit. If you look at the dropout V chart on page 6 you can see at 200mA and a case temp between 25 and 50C the droput V is about ~1.6 so I think you will be OK.

Also forgot you will drop about 1.2v across the bridge rec diodes so maybe cutting it a bit close with standard diodes but as its for audio put in some nice Schottky diodes to regain the dropped volt and improve the sound quality ;)

--Neal

freddiecas
17-03-2009, 16:44
thank you Sir! :) , the circuit I was sort of basing it on was for a Kef Kube from the 80s.

This had 2R 2W resistors from the sec to bridge, then 1000uF after the bridge, then another 2R 2W then another 1000uF then 4.7uF around the input of the 317/337 then more 4.7uF around the reference and output.

I had upped the 1000uF to 2x 2200uF keeping the RC filtering, and 3x10uF plus 3x0.1 ceramic around each of the regs. Then each parallel leg of the actual circuit board would have 10uF (total 60uF) with 36x0.1 ceramic around each individual op-amp supply. Or so I was planning :)

Cheers. Fred

freddiecas
17-03-2009, 23:10
if I use a 13V zener ref to get 14.25V out (giving a little extra safety across the regs) is 240R also ok for R3 (3 x 240 in all)?

and how do I put cap discharge protection diodes now, one across R3, one across R1 and one from output of 2nd reg to input of first?

thanks :)

NRG
18-03-2009, 09:15
Hi Fred,

RC filtering is a good idea but probably unnecessary with the pretracking configuration for ripple. Also can you afford to lose .4v across each 2R resistor?

The PSRR of the pretracking cct is very good, the benefit of an RC cct is probably the low pass filter it would form… 2R and 2200uf creates a LPF at 36Hz…maybe worth forming a 2200uf = 2R = 2200uf PI filter after the bridge….if you have enough volts….

I experimented with resistors in the secondary when messing about with gainclone amps and found they slowed down the bass and transients…although the current demands where much greater than your cct, I dunno I would tend to leave them out and use soft switching Schottky diodes.

I loathe ceramic caps, they suffer from Piezo electric effects, the capacitance varies with applied voltage and they can inject noise into the cct. I know some are better than others in these respects but I would avoid them if it where me.

The input to the 317 would only need a small decoup cap if they are positioned some way away from the supply, otherwise no need. I would decouple the supply at each opamp with a 0.1uF Wima cap and also place one directly across the – and + supply right under each opamp keeping the leads short as possible.

240R for R3 is fine. I’ve not used protection diodes but it may be prudent to place one from output to input on each reg if you are planning to use a large cap on the output, see Figure 4. I would not bother with the Adj pin diode.

If going for the Zener you can raise the output voltage by .6v with a diode to ground. Ff you use two Zeners in parallel you will halve their impedance. There’s a good info about his on Acoustic.org on the first page about 3 terminal regulators.

--Neal

freddiecas
18-03-2009, 09:57
thanks again for your stirling advice :) I apologise for not mentioning that the torroid was planned to be remote in a plastic box and 1 meter of 3 core cable would take the secondaries to the circuit box containing bridge and rectifiers to give you the full picture. The primary will end up connected all the time with the secondaries switched through a 3 pole for power.

Ok I'll ditch the ceramics for polyester or polyprop, and ditch the secondary 2 ohms, keep the 2 ohms between the two 2200uF and up the torroid to 18-0-18 to be safe. That would be around 9-10V across the regs for a 13V zener, say 5V max on each so around 1W so would one of those small U section heatsinks with fins down each side for upright TO220s be ok?

NRG
18-03-2009, 22:34
Yes, Fred should be OK...you'll soon find out though, if they are painfully hot when you touch them then you'll need larger ones!

NRG
19-03-2009, 08:05
I've implemented the pretracking supply for the -5 and +5v TDA feeds. Only thing left to-do is a couple of pretracking regs for the CS supply and I'm done....I think! Can't see what else I can do with the current layout / board, so time to case up and enjoy!

http://i94.photobucket.com/albums/l82/_NRG_/other/Alpha_DAC/DSCF4066.jpg

http://i94.photobucket.com/albums/l82/_NRG_/other/Alpha_DAC/DSCF4067.jpg

http://i94.photobucket.com/albums/l82/_NRG_/other/Alpha_DAC/DSCF4073.jpg

freddiecas
19-03-2009, 10:09
e really informative thread Neal :) The pre-tracker must have a much better performance than 2 regulators in series and the 317/337 is so cheap there seems to be no reason not to use it whenever possible.

I count 4 elecs, so one of them is decoupling R1 or R2? which one and which way round please? ;)

NRG
19-03-2009, 13:53
Hi Fred, yes, I have a 47uf bypassing the set resistor R2 and also the LED's/Zener...and 100uF on the o/p of each regulator.

Ah! you noticed I rotated the R2 bypass cap on each of the first regs. There's a small offset voltage across these (the 1.25 reference voltage) and as the 'ground' is now derived from the o/p of the second reg the orientation needs to be correct. The Adj pin is therefore 1.25 above the 'ground' so for the negative regulator the negative leg of the bypass cap goes to the Adj pin and for the positive reg the positive leg of the cap goes to the Adj pin...

Hope that makes sense...

--Neal

NRG
19-03-2009, 15:01
I have a bit of temperature problem with the leading 5v regulators...more heatsinking required!

leo
19-03-2009, 18:07
Thats coming on really nice Neal!

freddiecas
19-03-2009, 18:31
Hi Neal,
so you have an elec on the pre Vref resistor R1 ("other way round"), one on output, one across Vref zener or leds, and the last is C2 decoupling between the 2 regulators? if I've got that right :)
and nothing on the input if short and straight from the large decoupling caps.

The ground sense from zener//cap is particularly good advice, I shall connect seperate wires to my "star" ground at the parallel circuits output.

I found your link to acoustica - very good reading, look forward to the 4th part - thanks for your patience.

freddiecas
19-03-2009, 18:39
I have a bit of temperature problem with the leading 5v regulators...more heatsinking required!

do they have more volts across them in this application than the second regs which don't require sinks? I see they had one bigish heatsink on the original board.
cheers, Fred

NRG
19-03-2009, 21:22
Hi Fred, hope the attached diagram makes it a bit clearer, this is for a positive regulator. R4 in my diagram can be replaced with the diodes or Zener. I have a small 0.1uF Wima right in front of each pre reg to bypass the supply as theres a few inches of wire between them and the supply.

The + / - 5v regs are having to drop 20v or so with the Alpha board, previously they where fed from the 15v regs so had an easier time! The second reg only drops 2.5v so does not need one.

--Neal

NRG
19-03-2009, 21:32
I've also added larger heatsinks and had a good listen tonight, the presentation is very easy to listen to, there seems to be no nastiness or edginess to the sound and yet there's loads of detail and depth... very good for long term listening...I'm liking it a lot!

freddiecas
19-03-2009, 23:19
Hi Neal,
that's brilliant, thank you very much for your trouble. I had got the output decoupler wrong as it needs to be on the sense gnd with R4/zener/leds etc. :doh:

Hmmm, 20V across that reg, a lot more power than the other one!

This has got to be one of the most cost effective "super" regs around. I'm really glad all your efforts have paid off in the listening :)

cheers,
Fred

NRG
21-03-2009, 00:28
The TDA1541 has to be one if not the most intriguing DAC chips ever! Just been reading some very old threads on DIYAudio and I have implemented some suggestions for supply decoupling...no real time to listen tonight but through my headphone amp there's some interesting things going on... Just when I thought I was 'done' something else come up! ;)

leo
21-03-2009, 01:37
While your at it try a 10uf Oscon from the TDA1541A's -5v to -15V pins
Theres a reason for this mod;) measure from the -5v to -15v, think its positive leg of the cap goes to -5v and neg leg of cap goes to -15v
I fit the cap under the pcb directly onto the chip pins solder pads

Another good one is to tweak the voltage regs so theres exactly 10v between those pins:eyebrows:

NRG
21-03-2009, 08:07
I'm ahead of you there Leo ;)

Although I've not used Oscons or a single cap, good idea about the -5 and -15...read that as well and may have to derive my -5 from the -15...that'll mean more twiddling of my PSU! Also ground return of the +5 is critical...now I know more about this I see the Arcam board layout is not optimal...

NRG
21-03-2009, 22:52
I decided to leave the pretracking regs as they where... but I removed the zener from the -15V supply and replaced it with a multi turn pot to get 10v between the negative supplies...this and the decoupling between the negative supplies has made a significant change...very similar in magnitude to the opamp change and the introduction of an o/p TX in the squeezebox.

I didn't think there was much more to be gained from this cct but I'm mistaken...there's more I can do as well...where will it stop! :eyebrows:

Tony Moore
22-03-2009, 08:47
Hi Neal,

How close to the 10 volt differential do you have to get before you notice the improvement? What sort of tolerance are we talking here? I've not measured mine recently but I'd imagine they're within +/-0.5v of 10v. I wonder if this is something I should try? :scratch:

Cheers,
Tony

NRG
22-03-2009, 13:12
Hi Tony, I went from a .6v differential to approx +/- .02v :lolsign:

I read that .1v is to aim for. I did this at the same time as the -5/-15 common decoupling so can't really say at what point you would note the difference.

I've also ordered a few TL431's to play with, idea is to set the -15v using one instead of the Zener which is not adjustable.

NRG
24-03-2009, 07:41
The fettling continues...more internet reading and investigation of the MSB decoupling caps for the TDA...looks like Arcam already knew about these and used .47uF Tants vs .22 for the other decouple caps....anyhow due to their age I've replced them with Wimas...

Pin 13 of the CS8412 was left 'floating' and could possibly pick up noise, so I grounded it and re-introduced the cap multiplier ahead of the CS PSU regs. TL431 are on order ;)

Back to common mode decoupling between -5 and -15, I'd been using the recommended cct as per the data sheet (two 100nf series caps between -5 and +15) and had taken the center of the two caps to Dgnd. This changed the presentation to a faster, open and brighter one with loads of detail, pace etc....somehow its not very 'analogue' like...could be classed as 'digital' (runs for cover) and proved to be a little wearing over a long listen.

However, simply changing the ground placement and dropping it straight onto pin5 (Agnd) smoothes things out considerably, this seems natural to my ears very enjoyable but I'll try a 10uF Oscon between the supplys next to see have that sounds. Strange how one little change can have such a dramatic effect.

Also, I'm going to offset the voltages to see at what point the 10v becomes critical...

--Neal

PS there's a very interesting Charge Pump power supply design over on DIYAudio that I think I'll have a go at ;)

NRG
24-03-2009, 10:46
Hi Neal,

How close to the 10 volt differential do you have to get before you notice the improvement? What sort of tolerance are we talking here? I've not measured mine recently but I'd imagine they're within +/-0.5v of 10v. I wonder if this is something I should try? :scratch:

Cheers,
Tony

Tony, I've listened with a 9v difference and the sound is sat upon. Depth and width collapses and HF dynamics are reduced. Also detail becomes smeared.

At 9.5v the effect is not as noticeable, in fact unless you had heard it at 10v you would probably be less certain to detect it. However at 10v the sound snaps into focus. HF head room increases as does the sound stage, detail comes into focus and you can hear the texture to instruments. 9v is not unlike loosing bit resolution... I think aiming for within .1v of 10v is a good audiophile target. ;)

I've also just gone back for a quick AB test against my modded 7510 and....in comparison....its horrid. It's amazing how much further on the TDA1541 has moved. The 7510 sounds hollow and flat in comparison with a sort of grey blanket over the presentation...

I should really leave alone now but I've still got things to try! :eyebrows:

Tony Moore
24-03-2009, 15:25
Hi Neal,

I'll measure the voltage differential that I have when I get a chance. (Multimeter has finally given up the ghost after many years of service! So now awaiting new one) Thanks for doing the research!

Cheers!
Tony

NRG
24-03-2009, 18:05
No problem Tony.

I tried the 10uF cap between -5 and -15 supply and I didn't like it, I tried various caps a ZA, BG and Oscon. All, to a degree made the sound spit on sibilance and emphasize that particular frequency....going back to the series Wimas with center grounded to pin 5 sounded much more natural without the spit and fatiguing presentation. It could be something to do with the way I have the supply setup with pretracking regs....I dunno...but it was not for me :confused:

The TL431's arrived today and after frying a few :doh: I decided to use one as the set reference voltage for the -15 pretracking supply. I made it adjustable and it gives a much more stable voltage compared to just a variable set resistor. So now the 10v differential is more tightly controlled than before.

leo
25-03-2009, 17:00
Thats interesting Neal, I got quite the opposite to sibilance .
How long did you leave the cap in circuit? Oscon and ZA's tend to be a little electrically leaky when new (ZA more so) ZA are a bit etched anyway IMO, May have been a 10uf Oscon SP series in my AYA, I actually remember trying a film cap there too instead of the lytic, think a 0.1uf RTX gave rather good results going by memory, (could be wrong, tried lots of things since then)
Normally bigger films are a no no but the spacing between the pins was pretty good for that cap so there wasn't lots of leg length

Your spitty could be something else, the regulation etc, these dacs are so damn temperamental especially with the decoupling.
Anyway your getting some nice results and I'm pleased to hear your trying all this stuff out, patience with this old classic pays off, honest:lol:

NRG
25-03-2009, 22:53
Hi Leo, I may not have given them enough time, I may go back and retry with them in cct for a few days and see if it changes.

However, the DAC is sounding damm fine, quite possibly the best I've heard from digital in my system...

I've left the simple gyrator in cct for the CS supply and plan to try some ahead of the TDA supplies. Reading DIYAudio there are some recommendations for attenuating the I2S lines have you tried that?

And then there is the possibility of re-clocking! That may be a mod to far...we'll see...but I'll give the pulse charge PSU's a shot and see how they pan out...

leo
25-03-2009, 23:16
Obvious advice from me mate is leave it if its sounding that good, its so easy to upset things and can be harder getting things back to how you liked it, I've learn't the hard way ;)

Regarding attenuating the I2S signals, I've tried it and mileage can vary .
I started off with 22R's , also tried adding some filter caps etc but settled on just 22R's for all lines as the track was pretty short from the receiver to the dac (it all does effect the sound too)
I had it so links,resistors or even an asynchronous reclocker could be tapped in too
I'm not sure if you'd like Asynchronous reclocking, it sounds great on some material but a little artificial on other stuff, it adds that sounds better than is effect if you know what I mean

leo
25-03-2009, 23:24
The Gyrators are great at filtering crap btw, you can use them like the Teddy regs after the variable 3pin regs, only downside is the higher output impedance which isn't so bad on the receiver/dac IC's , not so good for op-amps though drawing tiny current so the shunts can help.

NRG
28-03-2009, 22:54
Quick update;

With more help from Leo (cheers mate!) I've been looking at the o/p filtering of the DAC. As this board was converted from 4x oversampling to NOS the o/p filter was in effect not working. After a *lot* of reading there's really no way to implement a fully functional anti-aliasing filter on a nos dac but I've managed to tweak the on board 3-pole Bessel filter to, at least, provide some filtering from 20Hz to 10 Khz. Previously artifacts could be seen at 1KHz on a sine wave o/p but now these don't become apparent until 10KHz...

Sound seems to be unchanged! But at least I know the mid band measures better!

--Neal

NRG
15-04-2015, 14:15
Rather than polluting the Analogue Dinosaur RPi thread I thought I'd bring this old thread up-to-date. I moved on from this DAC sometime back and went the Beresford route and have been happy for the last few years. With the release of the RPi2 and some very interesting all-in-one DACs now available for it and the ability to o/p I2S it seemed a good idea to see how the TDA1541A sounded fed from the RPi2 before I purchase an IQaudio DAC+.

An initial lash up proved it worked and also proved it sounded good, very very good! So I've now incorporated an RPi2 into the DAC case in a semi-lashed up way LOL!

Out went the CS SPDIF board and in came the RPi. A bit of filling and drilling got the RPi mounted with access to the USB and Ethernet ports at the rear.

I ran it with battery power at first but have now put together an external un-regulated 12v linear supply and regulated the supply inside the DAC with a 5v ALWSR. The ALWSR pre-reg get a bit hot and I may need to improve the heat-sinking and add case ventilation.

The sound is terrific, probably the best digital I've heard to date in my system and it wasn't too shabby before! Some pics:

External 12v Supply:
http://i94.photobucket.com/albums/l82/_NRG_/other/Alpha_DAC/AE08EAF1-1842-498E-AD73-650FF39E6BA5.jpg

ALWSR 5v
http://i94.photobucket.com/albums/l82/_NRG_/other/Alpha_DAC/727B29D7-F347-4709-A333-52B3D8932A21.jpg

RPi2
http://i94.photobucket.com/albums/l82/_NRG_/other/Alpha_DAC/8AC49B6F-4421-4950-BEB4-8054A117B333.jpg

NRG
01-05-2015, 14:10
I've been running my PSU with the extra stage of regulation for about a week and I have to say the sound has improved again, it seems to be getting better and better or I'm just more accustomed to it now ;) Anyhow, block schematic of what I'm using:

http://i94.photobucket.com/albums/l82/_NRG_/other/Alpha_DAC/RPi_PSU.jpg

Gazjam
01-05-2015, 14:15
Top work Neal.
Be interesting to hear your thoughts on the dac compared to the IQAudio Dac+ when the postie delivers it. :)

NRG
01-05-2015, 21:45
Cheers Gary, the IQ has a lot to live up to given the recent 'enthusiastic attention'...it will have a tough job against the TDA1541.....we'll see. ;)

Audio Advent
02-05-2015, 17:47
To clarify my understanding of what you've done, are you taking I2S straight off the Pi2 and is that a standard thing one can do (no extra hardware or mods)? And then you're feeding that straight to the TDA1541s as is a NOS design (perhaps a little tweeking with shift registers to get the right kind of input)?

I was following a thread on diyaudio.com a while ago, a group buy thread for manufacture of a set of boards to run chips like TDA1541as at high rez like 192KHz (as that's the speed they run at when 4 x oversampling is introduced, so is in their capability). Seems like the Rpi2 can perform this function out of the box with a software front end and all for £35..

NRG
04-05-2015, 07:34
To clarify my understanding of what you've done, are you taking I2S straight off the Pi2 and is that a standard thing one can do (no extra hardware or mods)? And then you're feeding that straight to the TDA1541s as is a NOS design (perhaps a little tweeking with shift registers to get the right kind of input)?

Yes, this is exactly what I've done but there's no tweaking of any registers...


I was following a thread on diyaudio.com a while ago, a group buy thread for manufacture of a set of boards to run chips like TDA1541as at high rez like 192KHz (as that's the speed they run at when 4 x oversampling is introduced, so is in their capability). Seems like the Rpi2 can perform this function out of the box with a software front end and all for £35..

I've not worked out how to get 4x oversampling out of the RPi yet but I'd be interested to know how, if it's possible at all....

Audio Advent
04-05-2015, 22:19
Yes, this is exactly what I've done but there's no tweaking of any registers...

I've not worked out how to get 4x oversampling out of the RPi yet but I'd be interested to know how, if it's possible at all....


That's great to know! It sounds like a much better way of getting audio into a DAC than either a s/pdif receiver or an asychcronous USB receiver.

What I meant in my post is that, with the TDA1541A connected directly to the RPi's I2S output, your dac should be able to play at 16/192KHz. How it would handle the truncation of a 24bit signal to 16bit I don't know (will still work, but might not sound perfect), but I'm sure you could do something in software on the RPi to dither down to 16 bit, if it could handle the processing?

The TDA1541A has a max speed of about 192KHz (could be 200KHz) and it is normally used running at 172.4KHz when 4 x Oversampling is applied inside a CD player of the 80s/90s. In a NOS dac, playing a CD will only push the chip to run at exactly 44.1KHz. Sorry if that's pointing out the obvious - am trying to be clear in an awkward way - I know what I mean.

lurcher
04-05-2015, 22:36
How it would handle the truncation of a 24bit signal to 16bit I don't know (will still work, but might not sound perfect),

No different that if it got a 24bit stream from a s/pdiff receiver. It just ignores anything after the 16th bit position from each L and R set.

Audio Advent
04-05-2015, 23:48
OK, think I'm confusing the thought that with a good noise shaped dither down to 16 bits you keep more info than a straight truncation.

NRG
13-05-2015, 08:23
Right, I've got oversampling to work. Quite straight forward in the end.

I used the MPD based player MoOde as I had it pre-installed on an SD card.

Simple case of configuring MPD via the web interface for MoOde and selecting resampling (I used 4x oversampling 16bit/176.4KHz) and SRC: Medium Quality or SOX Very High Quality.

Output stutters even with over clocking on SRC: Best Quality

Then SSH into the RPi (root/moode) and while a track is playing paste this line into the terminal window: cat /proc/asound/card0/pcm0p/sub0/hw_params

This returned:

access: RW_INTERLEAVED
format: S16_LE
subformat: STD
channels: 2
rate: 176400 (176400/1)
period_size: 22050
buffer_size: 88200

This confirms its playing 4x oversampled.

The only downside is I'll have to tweak the filter stage of my TDA1541a back to the original values as I's implemented a Bessel filter to run it in NOS mode! :doh:

NRG
13-05-2015, 20:35
Implementing the original interpolation filter has made a positive difference. It's subtle but the sound has opened up a touch and I'm noticing more detail coming through.

Interestingly I can run at 24/88.2 but not 24/176.4 I just get digital hash/noise from the 1541

Audio Advent
14-05-2015, 00:36
Oversampled and upsampled being interchangable here?

The thing with oversampling is that is done regardless of the resulting sampling frequency, it's a slave to a fixed mulitple. Whereas with upsampling, the mulitple varies and it is the output's sampling frequency which is fixed. When the limits of the dac chip are low compared to modern sample rates, you really need to do the latter to a fixed output rate.

Just wondering why 176.4 is digital hash..

Reading your post again, maybe it's being multiplied by 4 (oversampled) and then also experiencing a sample rate conversion to something higher still.. maybe it's being pushed outside the limits of the 1541.

NRG
14-05-2015, 07:24
Yes, the results are similar. 16/176.4 works fine, just not 24bit. The Arcam Board I'm using came from an early Alpha CD player and it was fed from a Philips SAA7220 Interpolation filter and 4x Oversampling chip. I have no hi-res material only 16/44.1 ....

NRG
26-05-2015, 09:28
I now have up / re-sampling working with the SoX player inbuilt into squeezelite/max2play

In the GUI front end of max2play go to the audio player menu and click on 'Edit Advanced Options' for squeezelite.

In the ALSA field I've set; 80:4:16: The 16 is the the sample format, I've tried 24 here but as with Moode the TDA1541a doesn't like it and the audio output is horribly distorted.

In the command line field I've set; -r 176400 -u hMX

Click Save.

To check if up sampling is actually working, ssh into the RPi and while playing some music paste this into the command line and hit enter:

cat /proc/asound/card0/pcm0p/sub0/hw_params

You should get a printout similar to this:

access: RW_INTERLEAVED
format: S16_LE
subformat: STD
channels: 2
rate: 176400 (176400/1) <<<<Indicates playback rate is 176.4Khz
period_size: 3528
buffer_size: 14112

This is my first shot at getting it going and there is a *lot* of scope for experimentation here. Full details of what you can tweak can be found here: http://manpages.ubuntu.com/manpages/trusty/man1/squeezelite.1.html

Initial listening shows promise with nothing obviously wrong. http://theartofsound.net/forum/images/icons/icon14.png

NRG
02-06-2015, 19:32
Quick update: I've settled on mIX for the filter settings and kept the sampling at 176400.

NRG
20-06-2015, 12:57
LOL. Gone back to hIX. Sounds more natural with greater depth.

NRG
01-03-2019, 15:52
Flippin heck time flys! Just bringing my old thread up to date. I’ve been using my TDA/RPi2 happily since my last post in 2015. Recently I was made aware of the Allo Kali reclocker for the RPi so went ahead and bought one.

The Kali reclocks the I2S to provide more precise timing and provide correct timing for 44.1khz based material. The RPi master clock runs at 19.2Mhz and is therefore not directly divisible by 44.1khz. The Kali internally buffers the I2S from the RPi and outputs a very low jitter I2S signal at the correct frequency.

Installation is straight forward as it simply plugs on top of the RPi, the output pins are in the same location as the RPi itself. Power is fed to the Kali directly which in turn passes power to the RPi. There is an option to power the RPi separately if you think it will make a difference. I made up a short 9cm cable from a USB lead for the connection to the DAC board. No software mods are required to support the Kali.

To say I’m pleased would be an understatement, the reclocking has made a noticeable improvement. Detail and texture (voices, instruments etc) have improved but it’s the improvement in musical engagement and satisfaction that is the biggest gain for me. Highly recommend.

Installed
https://i.postimg.cc/RV9L2bps/82-F9-BEDA-ECB5-4-F95-8079-D6-D1-D0-CC12-E4.jpg (https://postimages.org/)

In use, signal locked at 176.4Khz
https://i.postimg.cc/rmnfqGd0/1-AE1-CE00-F8-E8-4-FDE-AA0-A-F4-CD3-BB4-FFD4.jpg (https://postimages.org/)

Gazjam
16-03-2019, 09:25
Good work Sir...

Jimbo
16-03-2019, 16:52
Allo are doing some great stuff so not surprised you got a good result.

NRG
18-03-2019, 19:51
I’ve revisited the squeezelite upsampling settings and switched to picoreplayer as I was getting fed up with Max2play not responding and having to have a subscription so as to use a plugin to protect the file system. Picoreplayer runs totally in memory so SD card corruption should not happen.

Reading Archimago’s blog I found a post about digital filters and impulse filter settings for squeezelite http://archimago.blogspot.com/2017/12/howto-musings-playing-with-digital_23.html

And thought I’d give it a go, I messed about with the suggested settings in the post and ended up with the following as it sounded best to my ears and in fact was much better than the hIX setting I was using previously, in fact I’m quite taken aback there is such a difference.

The setting I have settled on is vX::4:20

Translating: very high quality, upsample to the max rate of 176400 for my 44100 material, -4db reduction in volume to avoid sample clipping during the upsampling process, use 20 bit accuracy. This can also be set to 28 bit but I’m not 100% certain it makes a difference...more time is needed.

There are three other user defined settings for passband-end in %, stopband-start and phase response, omitting them sets them to their defaults IE: 91.3% of 44.1khz/2 or 20.13khz, 100% of 44.1khz/2 or 22.05khz and 50 or linear response.

I thought I couldn’t improve on what I had before but these new settings are a bit of a revaluation to me and I’m going to try some more combinations to if it can get better still.

NRG
22-03-2019, 21:18
To wrap this chapter up, I’ve been playing with Archimago’s ‘Goldilock’ settings for squeezelite ending up with the following: vX::4:20:93:100:45

It’s a slight tweak to his Goldilock settings but just seems to hit the spot in a better way for me and my DAC. Blog post here: http://archimago.blogspot.com/2018/01/musings-more-fun-with-digital-filters.html