Anyone using ground isolation trans with bal. out dac?

[homemade]

I've been very interested in newer dac chips with balanced voltage output. In particular the AK4393 used in Berhinger's SCR2496 , given all the positive reviews of this unit.
As most know , good transformers = $$$. In hope of finding a less expensive , yet good alternative , I came across the A.R.T. DTI ground loop isolator. Basically a transformer coupled signal device. The specs look pretty good (at least to me).;

Frequency Response: 10Hz ~ 50kHz, -0.5dB/+4dB
THD: 0.01% @ 1KHz, <0.05% @100Hz

I bought one of these($45) and removed the transformers. I disconnected the power supply to the SCR's output stage and removed the resistor bridge just after the AK4393. I connected my 4 leads for R+,R- L+,L- to these pads and those wires to the 2 transformers. The output side of these is connected to the RCA output connectors I installed (insulated from chassis)
At first I was getting strong hum from these outputs , until I connected the rca ground lug to the circuit board ground(?). It's silent now.

Surprisingly , these transformers appear very transparent and both treble and bass are well extended. And the SCR used with this output as a dac/upsampler is extremely detailed with a very 3D soundstage.

Wondering which other good transformers might make a difference (but not super expensive)

[zigzagflux]

I have six Jensen JT-11-DMPC coming at the end of this week for the DCX2496 (same DAC). I have made preliminary tests with Lundahl's LL1690 with good measured results (no listening tests yet). The Jensens I expect are better than the ground isolators; the specs are much better, at least. They run $73 a piece. Depending on how much bass signal you drive and at what distortion you desire, you could drop the price of the Jensens to a different model with less core mass. Probably as little as $30 each.

Very excited to hear these bifilar Jensens perform.

[homemade]

I must say that anyone looking for another dac project , really needs to check out the many balanced dac boards available for use with transformer output stage. While certain that I'm telling many something they already know , for me , this has been the easiest and cheapest way to get outstanding digital decoding.

[Bill Fuss]

The trafo solution with V out dacs is pretty hard to beat, simple and elegant. The Jensen JT11-DMs are definitely on the top performance tier, and are a relative bargain compared to the Lundahl LL1690. They are dead flat within my test equipment range of 12hz-250khz, but the low frequency rolloff is dependent on the source impedance, as always. Units with smaller cores will have a higher low freq f3 with the same source impedance because of lower inductance. Many are very happy with the 11-EMs. Some have actually switched from the Lundahls to the Jensen DMs.

Actually, the specs you listed are pretty dreary compared to high quality trafos, but the proof is in the listening. My 11-DMs sound glorious, but YMMV as always.

I have not experienced any hum problems with any of the trafos I have used, either in balanced or single ended use. I can only speculate about what might be going on with people's hum problems. It would seem to be a hidden ground loop of some kind, associated with their other equipment, or is acting as an EM antenna. Grounding one end of the secondary seems to eliminate it but that is a bandaid and not an ideal solution.

Best, Bill

[homemade]

I'm quite certain that my transformers are not as well made as those Jensens , but I picked them because the were reasonable , had seemingly good specs(?) and were being used in a commerical design that was tested for use in (pro)audio. Those Jensen's will be on my list to try.

This my be OT , but is there a rule of thumb for 'lresistive loading' of the dac chip before the trans? I'm currently using 1k resistors on each leg. The trans. is rated 600:600 ; measures 250ohms per in/out winding. I picked my values based on the AK4393 data sheets spec these chips to drive 1.5k ohm min. The SCR2496 uses a 11K resistor bridge between the dac outputs and the active output stage , so I did not want to overload the chip on my try , via direct coupling.
Any suggestions on min loading?

[zigzagflux]

Fundamentally, you are best off with the least amount of series resistance between the DAC and xfmr. At least with most output transformers, optimal (flat) performance is obtained with the lowest possible source impedance. There may be a tendency to think the resistor helps limit DC current flow for undesired offset, but if there is DC present in mV ranges you may find a capacitor is needed if the transformer is not gapped (i.e. the series resistor will not benefit). I would not expect a functional differential DAC would have appreciable DC offset to affect you.

There is however benefit to providing some series impedance to ultrasonics. This helps reduce the chances of oscillations from the solid state output and its internal NFB. This is traditionally done with either RF chokes or ferrite beads. See the following link:

http://www.jensen-transformers.com/datashts/oli3.pdf

The final decision is how to load the secondary of the xfmr, which could be straight resistance possibly combined with a zobel network. This is very transformer-dependent, and maybe slightly DAC-dependent. Jensen will normally recommend the highest load resistance practical. In other words, don't add resistive load unless necessary or desired.

I suppose you could play around with your ears and test equipment to determine if loading affects the distortion. For the simplistic tests I have done to date with the Lundahl units connected to the AK4393, I see no difference in FFT spectrum from 5k load to infinity.

[Bill Fuss]

Just FYI, a distortion number @1khz is not going to tell you much. The dist. @ 20hz at rated ouput level will tell you how good they are.

The load ratings on the chips are kind of curious. If you consider that using trafos differentially as you are doing represents no DC load because both pins have the same DC potential, so there is no path for DC current to flow.

As far as max AC load, a 600/600 trafo is rated at 600 ohms, but without a substantial load on the secondary it's actual impedance is in the 10s of thousands. That is not a problem, the chip doesn't want to supply a lot of current to a load, and the trafos don't want a big load either.

If using 1k series resistors does not give you a low bass rolloff you are good to go, but I doubt that you have the bass performance the trafos are capable of. You can safely drop down to 250 or even less.

[homemade]

Originally Posted by Bill Fuss

Just FYI, a distortion number @1khz is not going to tell you much. The dist. @ 20hz at rated ouput level will tell you how good they are.

The load ratings on the chips are kind of curious. If you consider that using trafos differentially as you are doing represents no DC load because both pins have the same DC potential, so there is no path for DC current to flow.

As far as max AC load, a 600/600 trafo is rated at 600 ohms, but without a substantial load on the secondary it's actual impedance is in the 10s of thousands. That is not a problem, the chip doesn't want to supply a lot of current to a load, and the trafos don't want a big load either.

If using 1k series resistors does not give you a low bass rolloff you are good to go, but I doubt that you have the bass performance the trafos are capable of. You can safely drop down to 250 or even less.

Took your advise and changed the 'loading resistors' to 180ohms (what I had on hand). It does indeed sound fuller . Still pretty amazing what you can get for little effort ; once you understand the concept. The SCR2496 does have much potential to it. This is just the easiest approach to get it.

[zigzagflux]

Just for kicks, try it without any series resistors. I just did some quick measurements with the Jensen connected directly to the DAC, and got some really promising results. Just like Bill Whitlock mentions, there is some 3rd harmonic added, along with just a tad of 5th. But they are really low for all practical purposes.

Attached are FFT's of the reference (taken directly at D/A) and at the output of the transformer at 10 Hz. I still need to experiment with balanced vs. unbalanced measurements to keep the noise floor down. Actually cleaned up the 2nd a little.

[Bill Fuss]

Keep in mind that the average distortion of a high quality speaker at those frequencies is in the neighborhood of 5% or higher so you have to decide how good is good enough.

See if there are any tradeoffs up around 300hz, that would be in the vocal range which would be more important.