Tube-based or transformer cpl:ed stages cannot resolve 16 bit from 20-20 kHz:
The specific implementation with the specific transformers used in the example you reference cannot do so.
https://www.bramjacobse.nl/wordpress/?p=6233
There are multiple reasons here, including the of a poor quality transformer and an application of this transformer that maximises distortion (as opposed to minimising it).
That doesn't mean nothing with transformers and tubes can offer 16 bit 20-20k. But you cannot buy a random transformer, hook it up according to (wrong) instructions and expect a SOTA result.
I am aware of options that certainly can do so.
I originally had something on the drawing board. 1:30 step-up input and diffential output with cathode and anode feedback, 6SL7+6SN7 per channel, fully differential. Basically studio circuitry derived.
But at > 2k in transformers per DAC, the neccesity for a suitable powersupply tube rectifiers etc. we were at > 3k USD for just the DAC Analogue stages.
I decided to look elsewhere.
Thor
Last edited:
Is the Sowter 1465 really that bad for the 1541A. I was going to buy a pair just to see how they sounded vs opamps and active i/v stages. But, at over $300 for a pair my money might better spent elsewhere
YesTips:
Make sure the I/V stage is capable of resolving at least 16 bits.
Tube-based or transformer cpl:ed stages cannot resolve 16 bit from 20-20 kHz:
This is because transformer is higly complex rative element. And the DAC output is very high impedance current source.
For my opinion, DAC output is not valid "driver" for high reactive element. And current is factor (not voltage)
.
That is why people reporting and even measuring diferent results even with "just" simple Riv at the output of current DAC... Because of R capacitances and inductnces. That is why EC design used NON-inductive, NON-capacitive R component as Riv.
.
I made some tests with isotan non-reactive R and it is much better sounding than any same value R...
.
Even tried Graphite R it was also superior, but it is not easy to match values with graphite. Wotking better in speaker crossover too.
.
I think that trafo is OK, BUT it should be used:
DAC - NON-reactive Riv - low impedance buffer for low signal - Transformer 1:1
That it is where DA aomain ends. At the primary of thansformer.
.
Buffer coul use DAC power supply as it is part of DA segment.
Buffer is dealing with low signals so it has low THD.
.
At the secondary is pure analog and after that.
Transformer has low impedance driver, does not deserve much primary inductance for superior BW.
Wire could be "fat", with lower Rdc. Capacitances are significantly lower because of lower turns.
Transformer distortion is minimal because of very low signals. Sound signature is minimal too...
And DA module is compleatly isolated from Analog domain...
.
If DAC has galvanicaly isolated inputs, own power supply then it is compleatlely isolated from pure Digital domain and pure analog domain.
.
The DAC module is not partialy digiatal and partialy analog, it is DA in the same time...
The power suplies are marked as Va and Vd ground references too, BUT in fact it is all AD
Most people wants traffo because they make shortcuts in their head : passive = purer, less colored, etc ! In reality they take non good enough adapted for the task so they lost something.... Often less lively, etc, because not made sota ! Their ego and the mortgage they took let them believe and convince others they are fools. We have good example there of false prophets with low average knowledge !
Yes for sure !But, at over $300 for a pair my money might better spent elsewhere.
Thorsten had a post some pages ago about the traffos and the requirement needed for that task (load equilibrium and so on iirc) and why the one you can find on the shelf are not the good one for that ic, at least at the level of expectations discussed here !
It transfering the all reactive components trough the square of transformation ratio to primary - DAC output....1:30 step-up
Making the things even more complicated for DAC.
It is a differen with buffwer before the step up if it has to be used at all. I think that is 1:1 a good choice because transfer tatio is just 1.
And with that high ratio 3rd harmonic is dominanat from transformer. I am not saying that it giving sound signature but it could be...
YesMost people wants traffo because they make shortcuts
Take a look at the older CD players designs, with transformer at the end. Almost every good model has them...
BUT they driven with OP based circuit, powered with BJTs - for lower impedance source to drive these very good transformers...
.
DAC is simply totally opposite characteristic driver...
.
And does not like reactive load 🙁
True, 2 major issues with trafos, one concommitant with other: up to task, price... (the main issue of it)
Well, I have some experience with it, anyway:
For my DAC with PCM1704 I made a custom nanocrystalline trafo, making sure about performance at least for my needs (not inexistent but low 20Hz distortion, >50kHz flat HF response). Then, the I/V resistor is at primary side (or the leakage inductance will NOT load the DAC at HF, in case of secondary I/V placement), and some cap to ground, as first filter. Plus the trafo, forms a good filter for the signals way above audio.
The PCM helps due to output being zero at zero signal, so no detectable magnetization even with toroidal signal trafo.
Is my main DAC at the lounge room. Difficult to surpass it in naturality.
For my TDA1541A DAC I decided not to use trafo, using what I described. I can say for fun, and not to repeat the PCM choice = intelectually boring, but... the cost of trafo materials at least 😉 . Moreover, magnetizing current makes a bad trafo, and I not wanted to use balanced approach, needing one more TDA (this need a balanced primary trafo - not big deal, bifilar is welcome with such low impedances involved).
Well, I have some experience with it, anyway:
For my DAC with PCM1704 I made a custom nanocrystalline trafo, making sure about performance at least for my needs (not inexistent but low 20Hz distortion, >50kHz flat HF response). Then, the I/V resistor is at primary side (or the leakage inductance will NOT load the DAC at HF, in case of secondary I/V placement), and some cap to ground, as first filter. Plus the trafo, forms a good filter for the signals way above audio.
The PCM helps due to output being zero at zero signal, so no detectable magnetization even with toroidal signal trafo.
Is my main DAC at the lounge room. Difficult to surpass it in naturality.
For my TDA1541A DAC I decided not to use trafo, using what I described. I can say for fun, and not to repeat the PCM choice = intelectually boring, but... the cost of trafo materials at least 😉 . Moreover, magnetizing current makes a bad trafo, and I not wanted to use balanced approach, needing one more TDA (this need a balanced primary trafo - not big deal, bifilar is welcome with such low impedances involved).
sounds like my wife lol... Somewhere buried in here is the output stage with "diamond" in the name. I think that one was said to be the best by Pedja Rogic. I have the schematic somewhere. If I remember it was for single ended, not sure if balanced was or is an option with that one. That's the one that interests me the most. Will the end result with this version top that one?
Core matherial is not from the vital importance. Just has to be slight more of relative permeability. And with thin laminations. Old types are the best. They ususaly M laminations with small or moderate gap. Siemens, Telefunken and other german top brands almost exclousively useed VAC or Wasner laminations. permaloy, permenorm etc...
.
There are no magnetization current so the gap can be very small. There is just AC voltage if it is used after Riv.
.
BUT primary has very hard complex reactive components, as secondary too and with reversal transfer ratio from secondary to primary with self reactive components superimposing to primary reactance. All that i parallel with Riv
AND DIRECTLY into the output of DAC current Iout...
.
Even few 3-4pF of input capacitance of next device, transfered in reverse to primary, with say 1:30 ratio, become
90-120pF in addition to transfer of self capacitances from secondary added to primary and to DAC output...
.
There are no magnetization current so the gap can be very small. There is just AC voltage if it is used after Riv.
.
BUT primary has very hard complex reactive components, as secondary too and with reversal transfer ratio from secondary to primary with self reactive components superimposing to primary reactance. All that i parallel with Riv
AND DIRECTLY into the output of DAC current Iout...
.
Even few 3-4pF of input capacitance of next device, transfered in reverse to primary, with say 1:30 ratio, become
90-120pF in addition to transfer of self capacitances from secondary added to primary and to DAC output...
. . . . trafos works great ( better than active I/V , sound wise ) when they are correctly implemented , you have here some examples 😉
This is also what said John Brown from ECdesign ( with a lot of experiences , not just words ), especialy in the real world , the one I Iike and prefer 😎
I tend agree with u on the sonic qualities of the transformers - when correctly applied.
However, my point was to underline the necessity to measure the final trafo+tube based I/V stage to make sure it resolves 16 bits.
Cheers
My own experience proved the task to be far more difficult than anticipated.The specific implementation with the specific transformers used in the example you reference cannot do so.
I am aware of options that certainly can do so.
Thor
Many yrs ago there was an article (MJ Techonology) describing a tranformer-based design using a multifilar-wound amorphous core trafo with remarkable k-factor @ 20hz + discrete transistor circuitry that had 16 bits resolution @ 20Hz.
Is this the option you are referring to?
If not, I would appreciate any information you can provide on high res transformer-based I/V's
Cheers
Nice so what trafo are you using, is it sowter or & what follows after tube or ss stage ?
I am using Sowters 1495 with secondariers in series giving 1.0 V rms output , enought to drive my Pass F6 amplifier , the great result is also there , the two match perfectly in every way , impedance and so on
as I said before if there is something to learn here I will do , as I did from John , until now nothing new , waiting from words to become FACTS .......
.
Is the Sowter 1465 really that bad for the 1541A.
My experience with Sowter Transformers makes me disinclined to even give them a look.
At any extent, only someone who has no clue about transformers would suggest to feed them from a near infinite impedance with current and then use a resistor on the secondary to convert the now distorted current to voltage.
The lower the source impedance for a transformer and the higher the load impedance, the wider the bandwidth and the lower distortion.
The lowest distortion and widest bandwidth actually happen if the transformer is fed with a negative impedance exactly equal to the primary winding DCR.
I was going to buy a pair just to see how they sounded vs opamps and active i/v stages. But, at over $300 for a pair my money might better spent elsewhere
For starters, better transformers.
There are excellent and affordable 1:10 high nickel core microphone transformers out there. With +/-25mV we should keep the load impedance to 12.5 Ohm. In this case 12...15 Ohm at the TDA1541 will produce around 20mV into the transformer and thus only 200mV out of our transformer.
Distortion is going to be low.
We now need a tube circuit with a gain of around 10 and really low distortion.
The E182CC in SRPP has around this gain and cancels the tubes high H2. I'd not stake my reputation on it doing -90dB-ish THD but it stands a chance to come close. It was my solution for the first TDA1541 based DAC's I build ages ago.
We could also use a circuit with a high gain tube (6072A as cascode?) followed by a cathode follower (5687W as white follower?) and use shunt feedback to place the transformer into a near "zero field" condition, further reducing distortion especially at low frequencies.
The key is to understand how circuits work and to select parts and/or design circuits with the appropriate technology.
Thor
It transfering the all reactive components trough the square of transformation ratio to primary - DAC output....
Nope.
Resistor I/U conversion at the DAC, it was considered balanced. Around 40mV RMS (-26dBu) at 0dBFs from a 27 Ohm source).
After the transformer 1.2V balanced or 600mV per phase.
This signal straight into the grids of a dual triode (6SL7).
Output transformer 25k CT: 600 with feedback winding used in the output stage (6SN7) cathodes for local feedback.
Additional feedback from the output stage anodes and from the output winding to the input cathodes.
Pretty bog standard 1960's Pro Audio Stuff.
Transformers modern replicas of period transformers on insanely large high nickel permalloy cores.
The output transformer was rated open loop for 0.1% THD at 50Hz and +20dBu (8V) on the secondary, with 600r load and 25k source (absolutely worst case).
Making the things even more complicated for DAC.
Nope.
It is a differen with buffwer before the step up if it has to be used at all. I think that is 1:1 a good choice because transfer tatio is just 1.
Than why add a transformer? That's totally useless then, innit?
And with that high ratio 3rd harmonic is dominanat from transformer. I am not saying that it giving sound signature but it could be...
Ratio don't come into this. All transformers without DC Bias will be H3 and odd order distortion dominant.
What I see most people try is to go from a DAC directly to 2V line level with a transformer that usually 1:10 or thereabouts, as really high ratios are super b!tchy to wind and nobody wants work that hard.
For a TDA1541 balanced we can get 35mV @ 30 Ohm before pushing it. Some suggest more, but that questionable.
We can only now get 530mV.
So we increase our I/U conversion resistor from 15R+15R to 82R+82R. No DC current cancellation.
Now we have 2V out because we feed the primary with 200mV RMS. The TDA1541 distorts now quite significantly, the transformer source impedance is up and so transformer distortion increases.
We could of course make it worse. Take a 1:10 transformer with a balanced TDA1541, connect the primary without I/U conversion resistors, we get +/-0.2mA current in the secondary so we load with 15k for I/U conversion.
Can you see where this leads?
Ok, say single TDA1541, 12R I/U conversion resistor, so we get 17mV. To get 2000mV we need a step-up of 120. I worked on transformers designed for 10R feed and 1:32 stepup and 320V RMS on the secondary (no prizes for guessing the application) which was not easy.
I am seriously scared of attempting 1:120, but say we make it work...
Maybe make a vertically sectioned "scheiben wicklung" with 6 slices that each have 3 secondary sections between 4 primary sections and use altering clockwise and counterclockwise windings for 42 sections (see, 42 is always the answer) which should it, just who will wind it and what will they charge?
So we have a working 1:120 stepup from 17mV / 12 Ohm to 2V RMS at 173kOhm... So we need a buffer anyway.
TANSTAAFL and physics is indeed a harsh mistress.
So, use the right transformer and tubes correctly, results can be excellent and easily "good enough for 16Bit", but that never was nor ever will be cheap.
A handful of 20 cent each BJT's can get similar results, objectively at least, are also non-feedback and essentially "passive I/U conversion". So the extra cost seems like put a 24 Karat Solid gold Gucci branded case on an iPhone.
Too rich for my blood.
Thor
Core matherial is not from the vital importance.
It is. It has a direct relationship to level handling and distortion.
Nowadays we are sold "Amorphous" and "Nanocrystaline" Cores, guess you what, they are essentially > 90% iron cores and that suited for Audio, EXCEPT with a big aigap and for use with a lot of DC offset (SET Amp).
Now if we had 80% nickel in Amorphous or Nanocrystaline Cores, we might have something to discuss for DAC's.
Correct material makes a huge difference, incorrect material selection for the core, distortion performance becomes poor.
Thor
Hello folks,
I asked to CDREAM5 if they have this little board for the ones wanting simultaneous mode for whatever reason (double TDA1541A board; Symetric output; zillions oversampling, digital EQ, etc). It is pretty little and both Thorsten and Zoran said the guy behind makes a very serious work.
http://www.cdream5.com/htm/I2S_MK.htm
It is to avoid the mess for people not confortable with the idea to inject a little prg in a CPLD, because we have eventually two guys on the forum (one in the thread) who copied something already made or made a sim cpld board prg from scratch with a little handy cpld board.
And in order the Thorsten Insurection 's standalone core pcb works with the active sync we need someting:
directly simultaneous but w/o clock stop at the output or a board with I2S input before the cpld if using stop clock output (preferably not).
We could also ask them if theAlixpress store where CDream5 has maybe only the bulky pcb with SAA7020 we don't need, to supress the stop clock if simple to do in their prg?
For now, Thorsten confirmed few posts before most of us with reccordings up to 16 bits /192 K hz or 24/96 will work on a standalone TDA1541 chip in I2S mode with no better sound on his design than if readed in Simultaneous mode.
The 4 layers pcb I made for my needs is versatile though w/o sacrificing the layout quality. For the one wanting ultimate design there is the Ascendant version with two TDA1541. I will print a first 5 pcb batch for myself of the standalone TDA1541A core project here for me and the heads that contribuate to make it feasible. If their measurements will be ok, I eventually make it avialable if any interrest and noone else pop ups a pcb. For the moment, few corrections are needed perhaps on theFlip flop layout, cause I have a doubt on my understanding about the shematic about pin 1,4,10 & 13 of the flip flops.
In the meantime I am to beginn a pcb or two for the I-V , one from Thorsten, and one different, tube or discrete too, (same same but different.)
I asked to CDREAM5 if they have this little board for the ones wanting simultaneous mode for whatever reason (double TDA1541A board; Symetric output; zillions oversampling, digital EQ, etc). It is pretty little and both Thorsten and Zoran said the guy behind makes a very serious work.
http://www.cdream5.com/htm/I2S_MK.htm
It is to avoid the mess for people not confortable with the idea to inject a little prg in a CPLD, because we have eventually two guys on the forum (one in the thread) who copied something already made or made a sim cpld board prg from scratch with a little handy cpld board.
And in order the Thorsten Insurection 's standalone core pcb works with the active sync we need someting:
directly simultaneous but w/o clock stop at the output or a board with I2S input before the cpld if using stop clock output (preferably not).
We could also ask them if theAlixpress store where CDream5 has maybe only the bulky pcb with SAA7020 we don't need, to supress the stop clock if simple to do in their prg?
For now, Thorsten confirmed few posts before most of us with reccordings up to 16 bits /192 K hz or 24/96 will work on a standalone TDA1541 chip in I2S mode with no better sound on his design than if readed in Simultaneous mode.
The 4 layers pcb I made for my needs is versatile though w/o sacrificing the layout quality. For the one wanting ultimate design there is the Ascendant version with two TDA1541. I will print a first 5 pcb batch for myself of the standalone TDA1541A core project here for me and the heads that contribuate to make it feasible. If their measurements will be ok, I eventually make it avialable if any interrest and noone else pop ups a pcb. For the moment, few corrections are needed perhaps on theFlip flop layout, cause I have a doubt on my understanding about the shematic about pin 1,4,10 & 13 of the flip flops.
In the meantime I am to beginn a pcb or two for the I-V , one from Thorsten, and one different, tube or discrete too, (same same but different.)
Last edited:
- Home
- Source & Line
- Digital Line Level
- Building the ultimate NOS DAC using TDA1541A