Output transformers for DACs

[ArthurDK]

Off course a strong preamp would do the job too.

[ggetzoff]

Exactly....

[ggetzoff]

Originally Posted by tolikas

Hello, just to share my experience with 0.5+0.5 : 22 transformers on a SM5865 DAC Chips. They are really superb , sound is more transparent and 3D than ever before. Tried many kinds of Opamps on my DAC output - none of them can compare to this transparency. Of course i have weak output ( 0,47V) and a driver / buffer is needed to match my amp impedance, but once everything is set right - the potential of this transformer i/v is superb. Bravo!
SM5865 — ImgBB

Maybe Ivan can just swap them for a ratio that works better with your gain structure?

[tolikas]

I have not decided yet the final source - pre - amp setup, for now jfet buffer is fine , i have less gain but enough to drive the speakers easily. Once the final amplifier is known i will redesign the gain /vol structure . It was an experiment to check the transformer i/v solution for DAC output and it was a great success

[bisesik]

Originally Posted by ArthurDK

It seems you should have bought a 1:4 transformer. A buffer is not a solution, because it will give you no gain, and it is not easy to make a high quality stage with low gain and low output impedance.

The beauty of transformers is that you should not need anything else.

I suppose you are trying to compare it with your 4497 DAC, but there are too huge differences compared to SM5865 chip. I do not know is 0.47V is a 0.47Vp-p, 0.47Vp or 0.47Vrms, but peak-to-peak I suppose (please tolikas correct me if I am wrong). Lets compare SM5865 with AK4497 on the output capabilities:

0.47Vp-p for SM5865 on 10k load
7.5Vp-p for AK4497 on 10k load

There is about 16 times difference in signal power between DACs.

There is always a fork with any type of DAC caused by wanted i/v load and wanted signal voltage swing after i/v. Repeating again, as it is a current-output-type DAC, then the best possible way in terms of SQ is to work on to as low i/v impedance as possible. So as you can see, there is 0.47V only at the output with the 10k output impedance. That means SM5865 sees 20 Ohm in case of 1:22 turns ratio reflected from 10kOhm.
So the 1:4 version will simply not decide the situation with SM5865 as it is a weak-current-output DAC from the past. 10kOhm at the output will increase the output voltage swing for ~5.5 times (changing 1:22 to 1:4), but the impedance seen by the DAC itself will be rised to 625 Ohms from 20 Ohms, THD will be increased disasterably.

Audio as a part of our lifes is a compromiss - as always.

So, for SM5865 will be better to work on to high turns ratioed transformers (higher output i/v impedance/resistor), but there is another limitation caused bu LF response. Very interested what is the LF losses on 20Hz compared to 1kHz with such output (10k/0.47V)?

All correct in my vision - the best way in this case is to work onto the jfet buffer (current follower) or to use tubes as a power amplifier if both the voltage and the current of the signal is a target to be amplified (I not heard transistor power amplifiers that sounds better than tubes). Followers - another story. I suppose if to use transistors, then as a current follower only, not as a power amplifier (buffer, voltage gain should be less than 1). I know that 170V/74V is a pretty good solution for this role. All you need is to get the enough voltage of the signal on the jfet gates before transferring the signal to Amp using capacitive load (RCAs-cable--RCAs/input capacitance of the Amp side). The target is the slew rate IMHO.

[nautibuoy]

Hi Ivan.

I'm interested in a pair of your transformers for an AD1862N project, what arrangement do you suggest please. around 1 - 1.25V RMS output level would be great.

Cheers

Ray

[bisesik]

Hi Ray,

Just couple months ago I have implemented my transformers in to Sunaudio SV-DA800 DAC which is based on 8xAD1862.

IMG_20201204_193806.jpg

I have got the following output results using 1:22 transformers on 4xAD1862.

Output impedance = 14kOhm (20k across the secondary -> 47k after cable):
1.7Vrms@1kHz=0dB
1.5Vrms@20Hz=-1.1dB
1.63Vrms@20kHz=-0.4dB

So if to stay at the same loss on LF, then for single AD1862 the output will be: 1.7/4=0.4Vrms on the same transformer.

1Vrms can be achieved with ~40kOhm of the i/v resistor = output impedance.

Before continue, you should really understand that such a big output impedance is a matter of hard thinking regarding communication with the next stage. I would prefer to go with the buffer circuit (current follower) on a matched pair of 2sk170V/2SJ74V jfets, right after the transformers - it works fine, btw, I recommend.

[nautibuoy]

Thank you Ivan. Much as I would like to experience the fruits of your labours I think for the AD1862 project I will stick with the Borbely hybrid op-amp arrangement that I've been collecting the parts for.

I'm sure I'll get to try your transformers another day in another project.

[bisesik]

The most funny is that the high output impedance after transformers based i/v is simply borned to work onto high input impedance of tubes (fets). Not instead but together.

[bisesik]

If the mentioned is this circuit ( Borbely low-voltage hybrid Tube-MOSFET opamp as a line amp... ), then you can set the gain to x10 (the lower gain, the better SQ result usually for any OpAmps).

1:22 transformer will give you 0.15Vrms for i/v resistor=5kOhm, which you will feed to your Borbely circuit to get 1.5Vrms. Increasing the i/v resistor across the secondary and lowering the gain of circuit will gives even more sonical advantage I suppose.

Anyway - it is up to you what to do.