I tried the Digitecs without any load and still had a very tinny sound. At any rate I'll install them again with 510R resistors on Sunday and report back.
As an aside, been playing music through the Behringer all day now and it has finally started to 'loosen up' at long last; those SoniQ caps must take a while to sound their best. Still not sounding as good as my Gigaworks CS4398 DAC running into UTC-A20s output trannies, but I can finally hear its potential at last now.
Last edited:
I also have the smaller CS4397 dac with separate PSU:
CD DAC Lampucera lampizator
Using this with UTC A-43 transformers. Resisters and filters as per the Jensen scheme for these dacs.
Behringer SRC2496 with Sowter 3603 didn't compete with the CS4397 with UTC. CS4397 dac had bigger scale, better dynamics, tighter more powerful bass and more extended HF. This is the reason I swapped in the Lundahls into the Behringer, to see if it was the Sowters that I didn't like. Lundahls are very different, but I'm not convinced that the Behringer is any closer to the CS4397 dac.
Another possible issue is capacitors. I swapped out all the caps with tantalums on the CS4397 dac, and all caps with nichicon solid polymer in the Behringer (except the big caps - Rubycom ZLH used there).
Annoying thing is that the CS4397 dac sounds better with the SRC2496 upstream, used to upsample to 24/96 or 24/88.2 and set to use internal clock (think this helps reduce jitter from source).
Too many variables to draw conclusions but the CS4397 with UTCs is the better of the two dacs to my ears. Not given up on Behringer yet though.
CD DAC Lampucera lampizator
Using this with UTC A-43 transformers. Resisters and filters as per the Jensen scheme for these dacs.
Behringer SRC2496 with Sowter 3603 didn't compete with the CS4397 with UTC. CS4397 dac had bigger scale, better dynamics, tighter more powerful bass and more extended HF. This is the reason I swapped in the Lundahls into the Behringer, to see if it was the Sowters that I didn't like. Lundahls are very different, but I'm not convinced that the Behringer is any closer to the CS4397 dac.
Another possible issue is capacitors. I swapped out all the caps with tantalums on the CS4397 dac, and all caps with nichicon solid polymer in the Behringer (except the big caps - Rubycom ZLH used there).
Annoying thing is that the CS4397 dac sounds better with the SRC2496 upstream, used to upsample to 24/96 or 24/88.2 and set to use internal clock (think this helps reduce jitter from source).
Too many variables to draw conclusions but the CS4397 with UTCs is the better of the two dacs to my ears. Not given up on Behringer yet though.
Last edited:
Sharpi: Interesting to hear about your results. Like you I think I'm just extremely happy with the sound of the CS4398 chip.
Well tried the digitecs with 510R resistors on each primary leg and still no-go. Very faint sound and appearing to lack bass too. To be honest I give up with it and will stick to the Gigaworks with UTC-A20s. I could buy new trannies for the Behringer but then might have the same problem again and I can't afford to do that at present. At least it works via coupling caps I guess. And I could always put back that 11k resistor group and it should then work as new via the stock outputs.
Bit disappointing as this is the first project in over 10 years of DIY that hasn't worked out. C'est la vie I guess and chalk it up to experience!
- John
Well tried the digitecs with 510R resistors on each primary leg and still no-go. Very faint sound and appearing to lack bass too. To be honest I give up with it and will stick to the Gigaworks with UTC-A20s. I could buy new trannies for the Behringer but then might have the same problem again and I can't afford to do that at present. At least it works via coupling caps I guess. And I could always put back that 11k resistor group and it should then work as new via the stock outputs.
Bit disappointing as this is the first project in over 10 years of DIY that hasn't worked out. C'est la vie I guess and chalk it up to experience!
- John
Have you completely isolated the chip's output traces from the board? There is no reason in the world it should not work. Either there is wiring left on the board interfering with the operation or the chip is toast.
I honestly don't know Bill. The wiring looks fine to the naked eye, and passes continuity tests with a multimeter too.
The DAC outputs are only connected to the wires, and then to the resistors/transformers. All proceeding wiring on the circuit board is completely out of circuit.
I would have thought if the chip was toast it wouldn't have worked through the coupling caps, unless it's possible for the '-' outputs (which aren't used when using caps) to be damaged only?
At any rate sick of the thing for the moment - will leave it for a week or two then get back to it with a fresh pair of eyes.
The DAC outputs are only connected to the wires, and then to the resistors/transformers. All proceeding wiring on the circuit board is completely out of circuit.
I would have thought if the chip was toast it wouldn't have worked through the coupling caps, unless it's possible for the '-' outputs (which aren't used when using caps) to be damaged only?
At any rate sick of the thing for the moment - will leave it for a week or two then get back to it with a fresh pair of eyes.
I understand your frustration. Another thing to consider, especially if using caps from the plus, the Vcom pin is the return (neutral) and as on the datasheet it should be connected to ground through caps also. Is this handled correctly in the Berringer.
I'd imagine the Vcom is correct as nobody else seems to have had these problems using Behringers with output transformers. I'll have a look at the Behringer circuit though.
Good call - I'll try the '-' through caps too. Or I will once I get a new soldering iron - my trusty 60W Weller decided to pack it in this week and start glowing orange! I think it's just one of those weeks hahaha!
Thanks Bill! 🙂
Good call - I'll try the '-' through caps too. Or I will once I get a new soldering iron - my trusty 60W Weller decided to pack it in this week and start glowing orange! I think it's just one of those weeks hahaha!
Thanks Bill! 🙂
The transformer will provide very little series resistance to the legs of the DAC.
The DCR on a transformer will be in the ballpark of 25-100 Ohm. On the AKM 2493 (not sure about the your version), the required resistance is 600 Ohm or greater for the AC. Don't worry about the DC since you have wired the transformer to get rid of the DC, IIRC.
Why are assuming that the transformer is providing suffiecient series resistance? Are there series resistors on the other side (secondaries) that the transformer is reflecting back to the DAC?
See page two of the Jensen document:
http://www.jensen-transformers.com/an/an002.pdf
With 1:1 transformers, as long as the input impedance of the load on the secondaries is sufficiently high for the dac chip (>600 ohms for the AK4393) it should be happy.
The bit I'm not clear on is what happens at very low frequencies (<20Hz).
http://www.jensen-transformers.com/an/an002.pdf
With 1:1 transformers, as long as the input impedance of the load on the secondaries is sufficiently high for the dac chip (>600 ohms for the AK4393) it should be happy.
The bit I'm not clear on is what happens at very low frequencies (<20Hz).
At lower frequencies the trafo's inductance enters the picture and affects the rolloff frequency. The higher the inductance the lower the rolloff frequency, generally speaking. I don't know the formulas but a little digging will find them.
Thanks Bill.
So, in terms of the load the dac sees, the impedance will be that reflected by the load on the secondaries until below the LF limit of the transformer. Below the LF limit the impedance will drop from the reflected secondary load to the primary DCR resitance at 0Hz. Is this correct?
If this is right, there shouldn't be a problem with low series R on the primaries as long as there isn't much happening below the LF limit of the transformer (<20Hz with a decent audio transformer).
Pure speculation on my part - could easily be very wrong (please correct me if so). Just trying to learn 🙂
So, in terms of the load the dac sees, the impedance will be that reflected by the load on the secondaries until below the LF limit of the transformer. Below the LF limit the impedance will drop from the reflected secondary load to the primary DCR resitance at 0Hz. Is this correct?
If this is right, there shouldn't be a problem with low series R on the primaries as long as there isn't much happening below the LF limit of the transformer (<20Hz with a decent audio transformer).
Pure speculation on my part - could easily be very wrong (please correct me if so). Just trying to learn 🙂
I would say that your assumptions are correct but the trafo physics are much more complicated than that. You are right in assuming that lowering the source impedance that the trafo sees will also lower the rolloff frequency of the trafo. You can find graphs in the Jensen datasheets that clearly show this. I'm surely no trafo expert but this is what I know.
I just don't know exactly what is going on in these dac chips. If the DC on the outputs is truly only a bias voltage then a trafo should work fine directly wired to the output pins, provided it meets the impedance requirements. I just don't know if that is true.
The secondary load from a connected preamp input circuit is so high that it is not in play.
I just don't know exactly what is going on in these dac chips. If the DC on the outputs is truly only a bias voltage then a trafo should work fine directly wired to the output pins, provided it meets the impedance requirements. I just don't know if that is true.
The secondary load from a connected preamp input circuit is so high that it is not in play.
Correct, and not only does the inductance determine this cut off freqeuncy, but also the output Z of the source (DAC) and the input Z of the load (pre-amp). As I recall from the Jensen docs, the higher output Z from the source (DAC) will move the cutoff freq into the audible band.
Now, I will confess my confusion since these terms are being used inter-changably.
The AKM specs refer to a minimum resistance required (and this resistance is in series) for the + leg and also for the - leg. However, when we talk about load we usually mean the resistance (impedance) of the hot to the ground (essentially a resistance in parallel not in series).
Yes, I am aware that a series resistor on each leg leading to the transformer will add and create a "virtual ground" when the transformer cancels the DC (for a balanced-to-unbalanced conversion)
So is the input Z of the pre-amp effectively providing the required resistance that the DAC needs. The DCR of the transformer is only providing a trivial amount of resistance in series with each leg. I am also trying to understand this stuff also for a project that I am in the middle of.
Well, one thing at a time.
First off, the trafo does not create a virtual ground, the DC is still there. Both legs are at the same DC potential so no current flows, so the trafo ignores it. It's the same as biasing a filament circuit with a B+ voltage in a tube amp. Measure either leg to ground with your meter and you will see.
A voltage out dac chip ideally wants to see an infinite resistance for a load. It is a voltage source, not a current source. The tradeoff is you must have SOME current flow through the trafo to generate the magnetic field, but you also don't want to overload the dac chip, a balancing act.
A high impedance secondary load is reflected through the trafo to the primary circuit but it generally is many times the source impedance of the dac chip, which is generally around 100-120ohms. It is an ideal ratio for a voltage source. You must realize that we are dealing with transferring voltages, not power. That would get into impedance matching.
As far as the dac specs, the minimum R to ground for each leg can be substituted with 2x R from leg to leg, it would pull the exact same current because the potential would be double of one leg to ground. Putting a trafo between the 2 equal Rs forces the current through the trafo, and the dac chip is even happier because you have added the trafo impedance to the circuit, lowering the current further.
Inductance can be thought of as a complex frequency dependent impedance and I'm not sure how to explain what I know about it. The higher the inductance and/or the lower the source impedance the lower the F3 cutoff will be. As the frequency goes down the impedance of the transformer goes down also, so now we are entering into power transfer where impedance matching becomes involved. When the trafo impedance becomes the same as the source impedance there is a 6db loss in the voltage transfer, and as it gets lower the losses become greater. That is it in a nutshell.
The real experts can probably shred my explanation, but it works for me, and has since my school days long ago.
I'm wondering if the problem is not so much lack of bass as a peaking treble due to incorrect secondary termination.
Unfortunately 90% of this discussion is beyond me, but the lack of bass I experienced might simply be a result of the very low signal levels I guess? I had to turn the volume all the way up.
I tried with Jensen's recommended filter on the secondaries (which worked when I used the Digitecs with CS4398 DAC **), and without. Also tried no load on the primaries, 10R on each leg, and 510R on each leg. Same results every time.
The only time I've gotten acceptable sound is leaving the '-' wires unterminated from the DAC, and running the '+' connections direct into 4.7uF caps, and then directly to the output sockets (pins disconnected from circuit board).
Only thing left for me yet to try is Bill's recommendation of trying the cap on the '-' chip outputs instead to see if they're OK.
** Is the CS4398's requirements from a transformer really that much different to the AK4396? I can't help thinking the chip is damaged.
Cheers,
- John
I tried with Jensen's recommended filter on the secondaries (which worked when I used the Digitecs with CS4398 DAC **), and without. Also tried no load on the primaries, 10R on each leg, and 510R on each leg. Same results every time.
The only time I've gotten acceptable sound is leaving the '-' wires unterminated from the DAC, and running the '+' connections direct into 4.7uF caps, and then directly to the output sockets (pins disconnected from circuit board).
Only thing left for me yet to try is Bill's recommendation of trying the cap on the '-' chip outputs instead to see if they're OK.
** Is the CS4398's requirements from a transformer really that much different to the AK4396? I can't help thinking the chip is damaged.
Cheers,
- John
Last edited:
Bill, Thanks for taking the time.
Yes, I agree about impedance mis-match as a way to transfer voltage (not power, hence the technique of an impedance bridge).
We are at odds over your statement about the transformer not eliminating the DC. This, of course, is common use of a transformer for balanced to unbalanced conversion and getting rid of the DC that the + leg and the - leg are riding on. With two legs going to the two primaries on the Jensen, the voltage at the secondries will still be inverted, but the DC will have been cancelled. I am sure you know this, so maybe something was lost when you glanced at my words.
The term "virtual ground" may have been at bit informal, but it is the way that I had learned things.
Agreed about the inductance, although the cutoff freqeuncy not only depends on the inductance but also the source load (output Z). I did not know if you were disagreeing with that. Whitlock goes over some examples in his white paper if others want to have a look. Perhaps this is why the OP is getting a low freq roll off. I am only guessing at this point.
I am still confused about the resistance that the DAC "needs to see". The spec sheets clearly show that this is a series resistance; however, folks keep referring to a load resistance (which to me implies a resistance to ground).
Yes, I agree about impedance mis-match as a way to transfer voltage (not power, hence the technique of an impedance bridge).
We are at odds over your statement about the transformer not eliminating the DC. This, of course, is common use of a transformer for balanced to unbalanced conversion and getting rid of the DC that the + leg and the - leg are riding on. With two legs going to the two primaries on the Jensen, the voltage at the secondries will still be inverted, but the DC will have been cancelled. I am sure you know this, so maybe something was lost when you glanced at my words.
The term "virtual ground" may have been at bit informal, but it is the way that I had learned things.
Agreed about the inductance, although the cutoff freqeuncy not only depends on the inductance but also the source load (output Z). I did not know if you were disagreeing with that. Whitlock goes over some examples in his white paper if others want to have a look. Perhaps this is why the OP is getting a low freq roll off. I am only guessing at this point.
I am still confused about the resistance that the DAC "needs to see". The spec sheets clearly show that this is a series resistance; however, folks keep referring to a load resistance (which to me implies a resistance to ground).
- Status
- Not open for further replies.