Building the ultimate NOS DAC using TDA1541A

Anyway, amorphous and nanocristalline results in different animals in general. At least the core seller said to my boss and to me ;-)

Of course he would.

In doubt, measure and wonder or reject...

Indeed.

So here am I.

I want to put some audio transformers into project. I can readily buy 80% permalloy cored item's at around 100 USD each. Delivery to Thailand, as fast as I'm willing to pay. Next day if I need.

They are well documented and I fully understand the application implications.

A friend suggest AM/NC from Brand X. I look at them. They three to four times as expensive. To call the "datasheet" light on data is to be very British.

I email and ask for more data and am told "we don't publish or provide this data (I need to do my design)".

And I'm told "You don't need data, these transformers are the best ever, just read the reviews on our website."

"Send us money now, we start making them, you get them in 6 weeks."

If you were in my shoes, what would you do? Especially if you did test some samples with some kind of AM/NC cores before and we're rather disappointed.

That's before I read that the NC core costs 5 bux in volume and I know that the nickel in the 100 Bux OTS ones costs a lot more.

Thor

PS, the solution was to go transformerless and shift isolation to the digital side.
 
if you have any recommendation on how and what to measure, please tell me. I have a Tektronix SG505 audio generator 10 Hz to 100 kHz, THD <0.0005% and a matching AA501 level meter/THD analyzer, 24-bit 48 kHz sound card, FFT analyzer...

Look at the datasheets from Jensen Transformers, they show pretty much what and how and actual results.

Thor
 
Absent suitable data only empirical testing can confirm or reject a specific 'amorphous' or 'nanocrystalline' core as suitable in a specific application.
I'm not sure what exactly this means?
All I can say is this: when I switched from using permalloy EK cores to Finemet toroidals a quarter of a century ago, of course I went through the process of measuring and comparative auditioning.
The fact that there are no socalled "audio-related" specifications for Finemet is not strange: Finemet is not an audio-development but is made to serve non- audio applications. However this does not mean that the specifications do not give information on their possible suitability for audio transformers! Are all tubes you can find in tube amplifiers socalled "audio" tubes?? No way! My manufacture of signal transformers like phase splitter input transformers (check http://www.nutshellhifi.com/Karna.gif), MC step up transformers (check https://www.thraxaudio.com/trajan) and inductive volume controls (check https://www.diyaudio.com/community/threads/in-praise-of-tribute-autoformer-volume-control.298053/) might be "some" proof that high quality audio transformers can be made with Finemet.
 
My manufacture of signal transformers

Comes with a full set of technical data, allowing their quality to be evaluated?

That absent, it is hard to judge.

So in this case I have no choice than to go with something that is a known quantify that I can evaluate.

I cannot comment on the quality of your transformers. Because I have neither transformers to measure or a sensible datasheet.

I understand why modern cores lack the data I would like to see. They are not designed or meant for audio. Absent data, again, how can we judge?

I had the guy who made transformers for a brand I designed deliver some new ones with AM cores, he insisted were amazing.

In a direct AB comparison to nicely permalloy they sound fairly opaque, they measured rather high overall HD next to 80% nickel.

He was adamant they were so amazing,he came over from the other end of China, with more samples. He walked out totally crestfallen, after he heard what his transformers sounded in comparison (blind) and saw the measurements.

So I'm skeptical. And the more I see ipse dixitisms absent hard data, the more sceptical I become.

A QA403 is cheap, available and you can instantly provide a set of data using a "DIY Standard" that will be recognised. Your tests will be repeatable by anyone else.

Any doubts or scepticism will be dissolved.

In the time we we both spent here debating either of us could have characterised several different transformers.

Thor
 
I'm not sure what exactly this means?

It means that because audio specific data for these cores is not published, because objective test data for transformers using such cores is not published, the only way to tell is to test an actual core or transformer.

If I want a system with a certain (T)HD (profile) I need data that allows me figure out if a given device will be suitable.

If there is no data, I either need to buy and test, or I buy something where the data needed is provided.

Thor
 
@thorsten: What is the source of this picture?
That's in the Philips paper that shows details of the DEEM Circuit, including HD vs frequency et al., IIRC.

I did have a xerox of the whole thing way back.

Philips DEM info 1.jpg


PhilipsDEMinfo2.jpg


This is what is floating around on the net.

Note that the HD vs frequency comes from the free running oscillator, not synchronous. The free running oscillator is quite jittery with poor longer term stability, which may be the root cause.

If @lcsaszar is bored, he could try this on his test board. The lower the signal frequency the more DEM clock cycles have time to drift, jump into phaselock and out...

Thor
 
  • Thank You
Reactions: Alexiss
It means that because audio specific data for these cores is not published, because objective test data for transformers using such cores is not published, the only way to tell is to test an actual core or transformer.
Thor
Thorsten, which audio specific data are there for the various grades of permalloy?
I tried to find, but actually I don't see more than the usual and appropriate data (permeability, maxB, B-H curves) which are equally known for amorphous and nanocrystalline alloys.
Having said this, I think that the discussion on core materials here at this thread is starting to do harm on the main subject (ultimate dac using TDA1541A), so I will back off.
 
  • Like
Reactions: Logon
I've been thinking about this a bit more, couple of my naive thoughts, worst case I learn something from it 🙂

Two TDA is more a challenge layout wise. Making it with a two layers pcb with ground tracks IS a bad idea. And a 4 layer will be always quieter (for a standaloneic as well).

Completely understand, as a matter of fact I was thinking to have separate PCBs for each channel, this would then allow the user to either run a single TDA1541 normal in stereo, or for a single channel in balanced or parallel mode. Having separate power supplies for the left and right channel should have some sonic benefit? as well as increased channel separation. I also understand there is no free lunch and it might add other cons such as differences between TDA1541 chips.

My interest in balanced is not because of I have a balanced setup, but more in some of what intuitively could bring some sonic benefits. As a matter of fact, in the following post I'm trying to come up with a "unbalancer" : https://www.diyaudio.com/community/threads/sanity-check-differential-amplifier-for-dac.418850/, as adding 3/4 additional capacitors doesn't appeal too much 🙂

As for the intuitive benefits, my idea would be to have a balanced output, and then close to the output convert it into single ended, so that some of the noise / unwanted artifacts get cancelled/subtracted.

Whether the noise cancellation in terms of H2, H3 is beneficial, I'm not qualified to form an educated opinion on that. I'm not sure, but I'm suspecting the 0 crossing problem (mentioned in 2018 around post #6203) and extra current demand because of that would be somewhat compensated as both halves of the signal flip in opposite direction?

Also, I suspect it might provide an easier load on the power supply as you have a more constant current draw (currents flowing in opposite direction in the two halves), which should be easier on the PSU, perhaps a CCS could be a good PSU?

Again thanks for your patience whilst listening to my novice naive ramblings. In the meantime I'm ordering some components to build the CCS for the "unbalancer" / differential amp mentioned above and see if this provides any sonic benefit over my current singled ended setup.
 
If @lcsaszar is bored, he could try this on his test board. The lower the signal frequency the more DEM clock cycles have time to drift, jump into phaselock and out...

Thor
I am currently busy with my custom I/V transformers, which is a bit off track from the subject of this thread.
Anyway, I believe there are not many options for the ideal DEM clock in a NOS setup. According to Philips the clock frequency should be >=4x sample frequency, but it should be below about 200 kHz. And it should have 50% duty cycle, jitter-free, etc. The best option is, in my opinion, a divided BCK. I use the circuit described here and it works perfectly for me.
Or do you want me measure the THD with a synchronous DEM clock >200 kHz? If not free running, the THD won't increase, and we could use smaller filter capacitors (or better filtering with the standard 100nF), is that the idea? I can try 352.8 kHz, one divide-by-2 flip-flop less.
 
Some data: 2x50 turns primary, 2x 800 turns secondary.

What is the target impedance and what is primary inductance? The turns seem low.

THD at 1 kHz is 0.02%.

That is 100 times or so what a high quality OTS transformer delivers.

The TDA1541A with NE5534AN I/V measured 0.00085% THD at full scale, for comparing.

I'd expect comparable HD using a transformer. Make sure to use a 400Hz highpass so you do not measure hum as HD.

Thor
 
Or do you want me measure the THD with a synchronous DEM clock >200 kHz? If not free running, the THD won't increase.

I am hoping THD doesn't increase, you already tested?

Otherwise (say) 0dBFS @ 31/125/500/2000Hz with 176.4/352.8/705.6kHz DEM clock would probably interesting.

I can try 352.8 kHz, one divide-by-2 flip-flop less.

Yup. See if you can get 16X too.

Thor
 
I was thinking to have separate PCBs for each channel, this would then allow the user to either run a single TDA1541 normal in stereo, or for a single channel in balanced or parallel mode.

So double up everything?

Having separate power supplies for the left and right channel should have some sonic benefit?

Why?

as well as increased channel separation.

Try finding an LP setup with > 30dB channel separation at 1kHz. Still, LP has a great stereo image.

My interest in balanced is not because of I have a balanced setup, but more in some of what intuitively could bring some sonic benefits.

Hmmm, I'm not so sure.

As for the intuitive benefits, my idea would be to have a balanced output, and then close to the output convert it into single ended, so that some of the noise / unwanted artifacts get cancelled/subtracted.

Exactly what will be cancelled? And what is the source?

I'm suspecting the 0 crossing problem (mentioned in 2018 around post #6203) and extra current demand because of that would be somewhat compensated as both halves of the signal flip in opposite direction?

Why would cancel the 0 crossing problem instead of making it double?

Also, I suspect it might provide an easier load on the power supply as you have a more constant current draw (currents flowing in opposite direction in the two halves),

TDA1541 is basically constant current.

The only exception is +5V which is kinda of a dumping ground for the unwanted signal current (8mA PP theoretical worst case) and currents flowing in the digital input circuitry.

Some but not all of these will be canceled.

BCK & LE/WCK will not cancel.

The signal current can be cancelled by returning AOL/AOR to +5V.

I have considered this back and forth, the "balanced" benefits are minimal for a lot of extra effort and cost.

While John eventually abandoned the TDA1541, he first cancelled any multi-IC version he tried.

If you really want to do something worthwhile, use 4 TDA1541 in colinear to get 18 Bits and use additional error scrambling to swap the different channels in the block in a DEM like style at 352.6/384kHz, to scramble errors and no, this would not dual mono, but instead 4 X IC's with all "L" on L and all "R" on R.

Thor
 
  • Like
Reactions: Matthew78
Thorsten, which audio specific data are there for the various grades of permalloy?
I tried to find, but actually I don't see more than the usual and appropriate data (permeability, maxB, B-H curves) which are equally known for amorphous and nanocrystalline alloys.
Having said this, I think that the discussion on core materials here at this thread is starting to do harm on the main subject (ultimate dac using TDA1541A), so I will back off.
Agreed, ditto.

Thor
Also agreed.