Hi slawney,
I don't have any personal experience when using considerably lower rail voltages for these TDA chips, but I am sure that you are right here in that the distortion is reduced under these circumstances. That being the case, one could doubtless get away with the 'enhancements' due to using Caddock TF020 for I/V or V-Ref, and still enjoy the resultant sound.
With 8+v like I use there is already far too much distortion for my liking to risk adding anything which is not sonically-benign in these locations, and the results are surprisingly good still - so long as all other components are very carefully chosen. Especially in shunt applications, these naked Vishays (also if used universally in any one circuit) can give rise to a somewhat 'bland' sounding performance, but where accuracy and lack of added artifacts are concerned, the only resistor which beats the TX2352 in series locations is the TX2575.
I haven't experienced any damage after handling several hundreds of TX2352 and a few TX2575 resistors (yet!), but like you I don't like their cost at all!
Imagine the production costs for a remote-controlled 6 channel 64-step relay-switched attenuator which I developed for a commercial project a few years ago.
This was a constant input impedance design with absolutely accurate 1dB steps from 0dB to - 63dB (0.1% tol. resistors calculated and made to 5 decimal places), and it used all TX2352s for both shunt and series locations. It is the most 'transparent' attenuator I have ever (not!) heard and it has been used for demos in the UK, and I still use the prototype in my domestic system.
I have promised myself that one day I will construct the same circuit with all TX2575s whenever I am rich again (!), or perhaps if another commission comes along for such a good-sounding attenuator.
Regards,
Series Regulator
I wonder if either Peter or Bobken could comment on their preferred components in the series regulator?
thanks
Walter
I wonder if either Peter or Bobken could comment on their preferred components in the series regulator?
thanks
Walter
Hi Walter,
I have been using variations of this relatively simple series reg for a long time, and have tried a great many alternative components along the way.
It is not a very critical circuit except that the AD811 is quite a 'fast' chip and needs some care in layout etc to avoid any stability problems.
For this DAC the best sonic results for me were (unsurprisingly!) using Black Gate caps and bulk-foil Vishay resistors, although I have not constructed this circuit using all naked vishays (TX2352s) throughout. As usual with uncomplicated building blocks like this, the choice of components used does have a marked affect on the sonic results, and for resistors I have tried Holco, Caddock MK132, Takman, & PRP, but ended-up with mainly Vishay VSRJ which have the same resistive element and construction as S102, but with fractionally wider lead-spacings.
I also tried alternative nom. 1v2 voltage regs for VR1 but stayed with the type Peter shows on the schematic, although almost any low-noise well-specified component would probably be similar in results, I guess.
Maybe Peter will comment as he drew the schematic, but I wonder if the cap attached to pin 3 of the chip is an unfortunate error here as it is shown, although maybe he tried 10N and found it to be a good result. This cap will help to reduce noise & ripple, and I would normally suggest at least 10uF here, although during listening-trials I have run the reg in circuits with merely a BG 0.1uF. Again, I would suggest a 10uF BG myself, but I have also had good results when bypassing this cap with a 0.1uF BG (NX Hi-Q) which will spice-up the results a little, if this is desirable.
I didn't care for the results when going a lot larger than 10uF (with BGs) as the sound seemed to become heavier and a litle 'lazy' as if slugged or over-damped, but using a 10uF BG 'N' in parallel with a 0.1uF BG NX Hi-Q does provide a very revealing result if the remainder of the system is up to this.
A single 33uF 16v BG 'N' was somehow less-interesting and less lively-sounding in this location, although I have experienced worse results with other experiments!
Some capacitance is also desirable close to the output, and normally I would use at least a 10uF BG very near to the output, and the original circuit for the TDA chip used 47uF which is also fine, although the sonic results are slightly altered.
Making changes in these caps especially (and to a lesser-extent the resistors), can (and does!) have quite a marked effect overall and is a handy method of fine-tuning the 'voicing' of the DAC to one's own preference.
With relatively few components involved, using more-costly parts is not (quite) such a problem for me, and very rarely do I find that using better quality parts gives a less-good result, so it is usually worthwhile 'investing' in HQ parts for the best overall results, in my opinion.
Regards,
Last edited:
The pictures here: http://www.diyaudio.com/forums/audio-sector/187748-pushing-limits-tda1543-nos-dac-2.html#post2551745 show the circuit with series regulators. So I was indeed using 10uF BG N on pin 3 of the op-amp.
I don't recall now what was my reasoning here, but as usually, it's recommended to try the other options as well.
Hi Peter,
I misunderstood the schematic value here because I took it to read 10n (F), whereas you obviously intended it to be a 10uF 'N' BG cap, but this has at least clarified the matter for anyone in doubt.
I fully agree with others trying variations for themselves, and to be quite clear for members I have never suggested that what I have recommended is the absolute best choice which can ever be found, of course. It will simply be the best overall result that I have discovered (so far!) and I will have tried a great many alternatives before making any suggestions for others to try-out.
As you know, its not unusual for me to find an even better result at a later date, after some improvements in an alternative region of the DAC circuit have increased the overall resolution, and this is all part of the iterative process for me.
Indeed, I am presently working (once again!) on the TDA chip P.S. reg and its choice of associated caps in the hope of achieving some better subjective results in the extreme HF regions, so watch this space!
Regards,
Hi everyone, I have only noticed this thread recently, and I wasn't aware that Peter posted my mods list here. I wrote it before I actually did anything, and I would change a few things now.
I would start modifications with the destructive mods, as they seem to have the biggest impact on the sound. Swapping one capacitor type for another (one resistor for another etc.) only alters the "flavour" of the sound, and doesn't change the fundamental qualities.
So my list of mods would now look more like this:
Now, from here, you have the choice of:
And I think this pretty much exhausts the current state of my knowledge
I'm sure Peter / Bobken will correct if anything is wrong, or is no longer recommended.
I would start modifications with the destructive mods, as they seem to have the biggest impact on the sound. Swapping one capacitor type for another (one resistor for another etc.) only alters the "flavour" of the sound, and doesn't change the fundamental qualities.
So my list of mods would now look more like this:
- Get rid of the SN75179 and all the parts that feed it (U5, U1, C5a, C5b, L3). Remove the two film capacitors at Digital In (C6 and C7) - this is not generally recommended by "engineers", and you may want to leave these in place if you experience problems with ground loops etc. - it has always worked fine for me though. Now, to make DAC usable again, reroute the signal from the 75R Caddock to CS8421. IN+ goes to pin 9 and IN- to pin 10.
- Remove the 3.3nF Wima capacitor (C10) from pin 20 of CS8421. Remove the two dual inductors in the PSU (L1 and L2), restore the connections with jumpers. Replace the last remaining inductor L4 with a jumper.
Now, from here, you have the choice of:
- Replace both Black Gate STD 1000uF caps with Black Gate FK 2200uF (good luck finding any)
- Add one Black Gate Nx 0.1uF bypass across the input and one across the output of each regulator (U2, U3, U4).
- Build Bobken regulators (or your super regs of choice), and replace the AN800x ICs with them. Doing this obviates the above mod.
- Replace the Caddock at Digital In (R5), as well as the Caddock at pin 20 of CS8421 (R6) with same value nude Vishays (check below for R6).
- Replace the 0.47uF Black Gate Nx cap (C11) with two BG Nx 0.1uF, but measure them so that you get 0.22uF in total. Alternatively, replace the R6 resistor with a different, but fairly close value, and aim for this formula:
Either: C11 = 49 / R6 -or- R6 = 49 / C11
For example, if you have two caps that in parallel give you exactly 0.245uF, get a 200R resistor (and make it a nude VIshay, while you're at it).
- Replace the two 20R resistors in PSU (R1 and R3) with same value Caddocks MP930.
- Get an extra transformer - so that the PSU for the digital section and the PSU for the analogue section are completely independent, each powered from its own transformer.
- Replace the I/V and Vref resistors (R7, R8 and R9) with Caddocks TF0200 (2K7 and 1K5).
- Replace the output electrolyte caps (C13 and C14) with superior film caps. If you have Black Gates N here, you will need to aim rather high to beat them.
Cut the PCB board to separate the PSU section, and house it in a separate enclosure (leave C2 and C4 with the DAC).
And I think this pretty much exhausts the current state of my knowledge
I'm sure Peter / Bobken will correct if anything is wrong, or is no longer recommended.
It would be interesting to see if Lundahl or Jensen had any transformers that would convert balanced to unbalanced without coloring the sound.
Regards,
Dan
Have you tried the Lundahl 1545a? I have two unused of those that I have not tried yet but will shortly if time allows me!
Very interesting thread!Peter, I have one question. Looking at the new circuit, it still uses only a single resistor as I/V conversion. Looking at different threats here I know that you have earlier tried different active I/V circuits
...are you conclusions that a simple resistor is better?
...if so, is that better on all parameters?
...any related experinces?
At the moment I am very satisfied with my passive version, but having considered building an active I/V stage I got interested in you evaluation!
Yes, it uses single resistor at the output and I'm not tempted to change things here.
We were experimenting with adding a simple Boberly buffer and it was effecting transparency. In a past I also tried Rudolf Broertjes I/V stage, and again, went back to single resistor. As they say, if it's not broken, why fixing it?
We were experimenting with adding a simple Boberly buffer and it was effecting transparency. In a past I also tried Rudolf Broertjes I/V stage, and again, went back to single resistor. As they say, if it's not broken, why fixing it?
Those cables are not made by me. I wrote more about them here: http://www.diyaudio.com/forums/audio-sector/151938-phono-stage-17.html#post2655116
Quite by accident I came across pictures of this system today, while browsing local audiophile club meetings: New Page 2
This is the older setup, from last October. You can still see there the TVC line stage I've built for him. The most recent picture in attachment. It's sort of a spy pic, that's why the quality isn't to my usual standards
This is the system when we (me, Marlen and George from Audio Oasis) do all the critical listening comparisons. While all technical aspects of the design come from me, those two other guys always contribute to the final voicing of the device.
BTW, That system, with the DAC, was recently used at TAVES and this is what they just wrote about it: Blueberry Hill Audio Rhapsody 3D | Stereophile.com
Yes, it uses single resistor at the output and I'm not tempted to change things here.
We were experimenting with adding a simple Boberly buffer and it was effecting transparency. In a past I also tried Rudolf Broertjes I/V stage, and again, went back to single resistor. As they say, if it's not broken, why fixing it?
How about the "sin(x)/x" filter for correction the HF-response, have you tried it?
How about the "sin(x)/x" filter for correction the HF-response, have you tried it?
Subjectively, HF response seems to be perfectly fine, so we didn't try any additional filtering.