CD Player Output Stage - a few questions on implementation!

[Mikey]

Hi gang,

I recently picked up a Denon DVD-2900 universal player, and I've got the urge to tinker with it a bit.
I'd like to bypass the opamp outputs and come up with a better sounding solution.

The CD player uses a DSD1790 DAC chip, very similar to the DSD1792.
It is a differential, current output chip...see the spec sheet below.

http://focus.ti.com/lit/ds/symlink/dsd1792.pdf

It seems like there are three ways I can accomplish this task:

1. Vacuum tube output stage (active)
2. Transformer output stage (passive)
3. Transformer, followed by vacuum tube output stage (combination)

My line stage preamp is an Extended Foreplay 3.

Can anybody suggest a way to implement any of the aforementioned schemes?
Maybe there are schematics out there that you've tried?

Thanks in advance,

Mike

 

[Paul Joppa]

To answer the question, you need to know how much voltage fluctuation the outputs can support without loss of sound quality. I couldn't find anything in the spec sheet about that (OK, I didn't read every word of the 59 pages!) so you'll probably need to get hold of someone experienced with this chip. Swenson, are you reading this thread?  :^)

 

[Mikey]

Hi Paul,

Thanks for the reply!

I sent a PM to John Swenson yesterday, but he's buried with work for the next two weeks.
I figured I'd post up here to see what kind of interest I could generate in the meantime.

I brought the spec sheet over to our resident digital expert (here at the office), but he was not
sure what you meant by 'voltage fluctuation the outputs can support without loss of sound quality'.
Is there a specific spec (or specs) that you need to get an idea of what will (or will not) work?

Also, would the schematic of the Denon's analog output stage help at all?
I'l try to attach a portion of the schematic below.

Mike

 

[Paul Joppa]

All of the schematics show an opamp current-to-voltage converter at the analog outputs. The output pin is loaded with a virtual ground at the opamp circuit input. If a resistor is used as the current to voltage converter in order to get rid of the opamp, the voltage on that pin will rise due to the current through the resistor.  How high that voltage can rise without compromising performance sets a limit on the available output voltage, which in turn determines what kind of gain and impedance is needed to get to a normal output voltage.

Actually, though it does specify 7.8mA peak to peak at each output pin for a full-scale signal, it does not specify the output current at zero signal (quiescent) either. Really, all they say is "use this circuit and this opamp". At last that's all I could find.

Some chips can tolerate nearly a volt on that pin, so no further gain is really needed. Others can't tolerate, or don't sound as good, if the voltage exceeds 10mV - which means they behave more like phono cartridges and need lots of gain.

 

[Mikey]

Hi Paul,

So....what's the next step?  A few options come to mind:

- call (or e-mail) T.I. and ask them for some application assistance
- prototype the passive transformer scenario, and see if there is enough gain
- simply assume that it won't sound good exceeding 10mV, and design a solution accordingly
- poke around inside the machine with a meter and try to learn something emperically?  (I'd need help with that one)

Have you got any idea what JT does inside his modded Denon players?
I know these machines were popular with him a few years back....

Mike

 

[Paul Joppa]

I'd just hit the digital forums and search for the particular chip - I'll be somebody has tried everything you can think of!

 

[coca]

I know of many folks who have claimed they have greatly improved the performance of Their players such of yours, by going from your players outputs, then into a YaqinCD1 tube buffer. Not an expensive device, and can be purchased from The Pacific Valve and electric company near Chicago.

Bernie.

 

[Paul Joppa]

This is getting a bit confusing with another thread on the Magnequest forum. In that thread, John Swenson mentioned the K&K transformer passive approach. I'll put in my two cents here though, since I'd rather not discuss other people's transformers on Mike's forum...  :^)

The K&K board uses Lundahl transformers and is described as suitable for the PCM1794 chip. Does the DSD1790 have the same output characteristics as the PCM1794 chip? I don't know.

It appears that the Lundahl is used in 1:4 stepup mode running 187.5:3K ohms with 0.5vRMS appearing between the chips differential outputs. That's rather a lot of voltage, which is why I asked about the chip's voltage output capability. John suggested a reduced load impedance for better performance, but by my interpretation of the small amount of data available on the Lundahl, it should not be loaded with much less than 2500 ohms with the secondaries paralleled lest the high frequency resonance become underdamped.

Direct tube outputs are also practical, it's just necessary to be careful about grounding and RFI when interfacing digital to analog circuitry. That's the main reason for considering transformers, IMHO.

 

[John Swenson]

I'm still snowed under but just woke up and can't get back to sleep so I thought I'd peruse here! The 1792 has exactly the same output stage as the 1794, so if the 1790 is the same as the 1792 then a 1794 design should work.

I've done several 1792 output stages and have found that most people put too high an impedance on them. Because the output has such high current the temptation is to go full passive, this works but doesn't run it at its best. If possible you should try and run it at less than 20 ohms into each output. Its possible to run it single ended (ignore one output) but its best to make use of the balanced output.

The outputs are designed to drive current into ground, so current flows out of the pin in all cases, with 6.2ma at no signal. The chart at the bottom of the spec sheet lists the currents. (why they say current flowing out is negative I don't know!)

To me the transformer in this case serves several purposes, it does balanced to single ended conversion, it provides very effective low pass filtering and you can get some voltage gain (depending on the ratio).  The exact design that K&K uses may not be exactly the most optimal solution. As I stated by going full passive I think they are putting too high an impedance on the outputs. The recommendation of the 1K was purely based on getting the load impedance down to a reasonable range for the DAC chip, I haven't done it so I don't know if its really a good fit for the transformer in question.

The designs I've done with the 1792 used MOSFET based I/V conversion for each output feeding a LTP differential amp, it works very well but is fairly complex. I think its possible to get the same (or better) level of performance with a transformer and gain stage.

Paul, one big difference with this chip over what we are doing is that the current range is much larger, you have to carefully choose the transformer so it doesn't saturate with those higher output levels. Some of the low power transformers I've tried before will overload on this chip. I was thinking that since the K&K design is already known to work with this chip it might be a good place to start.

John S.

 

[Mikey]

Hi John,

Good to see you over here!

So...in your opinion, a transformer followed by a gain stage would be ideal?
My only concern would be too much gain, as I'm already using a Foreplay III,
and when listening to CD's I never even approach 12:00 on my volume control.
To me, it would seem counterintuitive to add a step-down transformer after
the gain stage.

I can start to research suitable transformers if you tell me what specs to look for.
I've had decent luck with transformers from Magnequest, Cinemag, and Jensen
(among others) so I can begin with those manufacturers...

Mike

 

[Paul Joppa]

John is talking about 20 ohms maximum each output, which would call for a 40 ohm CT primary. It will be carrying 6.2mADC in each side, so any imbalance will show up as a magnetic bias, risking saturation especially if the core is high-nickel. That means many microphone transformers are risky - they may be extremely sensitive to unbalanced DC. Also, the high output (7.8mA pk-pk IIRC) could easily be more than the transformer can handle. That's about 110mV rms, ( about -5dBm) and you're looking at a MC step-up type transformer. You only need a total gain of about 20 times to get full 2vRMS output, so with a single tube gain stage a 1:1 transformer would do the job.

Based on that (and assuming my quick arithmetic is accurate) I'd look for something rated 40CT:40 ohms at 0dBm, and rated to carry some unbalanced DC. You won't find many. You may find something with 20-50 ohms primary and 600 secondary, giving something like a 4x voltage increase and calling for a gain of 5 in the tube - perhaps a 12B4?