I recently decided that the solid state I/V converters in my diy DAC were not really in keeping with the rest of my system which is tube based with a lot of high quality transformers in the signal path. In addition it seemed to me that the TD-124/II with Schick arm, Meister Silver SPU, and Muscovite phono rather readily embarrassed the dac even with high res material. This resulted in my designing the I/V converter "upgrade" presented here.
I ordered a pair of Sowter 9545 I/V transformers with the OFC winding upgrade, they're built on mu metal cores in a mu metal can with lots of electro-static shielding. They're quite good to my relief, almost enough gain and headroom to run without amplification, but not quite I'm running them pretty hard so I think something like 20 - 30dB of gain after them might be suitable after reducing the I/V resistor values. (I'm currently running 20 ohm resistors, I suspect something like 4.99 is the right tradeoff - I've tried 1 ohm to 39.2 ohms, anything higher results in audible distortion.)
I'm using them with parallel mono mode PCM1794A.
I'm trying to find the best tradeoffs between transformer saturation, the voltage compliance of the current output of the PCM1794A, noise, and available tube choices for gain with suitably low output impedanc.. Juggling act. I'd like to keep the amplifier circuit simple and my current choices are the 5842 or C3G, the 5842 would be choke based mu follower, the C3G probably just a choke load.
Shame there is not more voltage gain available, I've also determined that the transformer sounds best with a relatively high load resistance on the secondary and I will run it straight into the grid of the amplifier stage and them try different load resistances if necessary..
Edit: If you intend to build this after reading the thread please note that the last schematic of the I/V converter is the one to build. I've left the intermediate versions but they were part of the learning curve. Note also that this could easily be adapted to other DAC Ics like the AD-1955. See post # 33 for the latest schematic (could be subject to change)
**The Salas SSHV2 is used to power the high voltage sections of the I/V converter.**
I ordered a pair of Sowter 9545 I/V transformers with the OFC winding upgrade, they're built on mu metal cores in a mu metal can with lots of electro-static shielding. They're quite good to my relief, almost enough gain and headroom to run without amplification, but not quite I'm running them pretty hard so I think something like 20 - 30dB of gain after them might be suitable after reducing the I/V resistor values. (I'm currently running 20 ohm resistors, I suspect something like 4.99 is the right tradeoff - I've tried 1 ohm to 39.2 ohms, anything higher results in audible distortion.)
I'm using them with parallel mono mode PCM1794A.
I'm trying to find the best tradeoffs between transformer saturation, the voltage compliance of the current output of the PCM1794A, noise, and available tube choices for gain with suitably low output impedanc.. Juggling act. I'd like to keep the amplifier circuit simple and my current choices are the 5842 or C3G, the 5842 would be choke based mu follower, the C3G probably just a choke load.
Shame there is not more voltage gain available, I've also determined that the transformer sounds best with a relatively high load resistance on the secondary and I will run it straight into the grid of the amplifier stage and them try different load resistances if necessary..
Edit: If you intend to build this after reading the thread please note that the last schematic of the I/V converter is the one to build. I've left the intermediate versions but they were part of the learning curve. Note also that this could easily be adapted to other DAC Ics like the AD-1955. See post # 33 for the latest schematic (could be subject to change)
**The Salas SSHV2 is used to power the high voltage sections of the I/V converter.**
Interesting Kevin. Keep us up to date on what you find. I have a pair of Sowter 1:9 transformers that I might try in a similar mode. Did I get those from you, SY? I also have a pair of multi-tap Sowter that sound dreadful, never figured out why.
Quite interested to see what you come up with as the tube section. I have a Chinese current output DAC called "CityPulse" (don't remember the TI chip) that sounds a bit nasty. It uses a discrete transistor I/V output stage. Maybe it would improve with a Kevin K make-over. 🙂
Quite interested to see what you come up with as the tube section. I have a Chinese current output DAC called "CityPulse" (don't remember the TI chip) that sounds a bit nasty. It uses a discrete transistor I/V output stage. Maybe it would improve with a Kevin K make-over. 🙂
Wrestling with what to do next, it sounds good enough I am almost inclined to leave it as is, but I think lowering the I/V resistors is probably the right move long term. I'm now just thinking of a simple fix biased SRPP as it provides sufficient gain with 1 ohm I/V resistors and a low output impedance. A mu follower would be more linear so that will get evaluated as well.
I started out with some clever ideas which have been discarded because I need much less gain than I initially believed based on working with the actual transformers which have a pretty wide window in terms of signal handling.
The original goal was to end up with no solid state in the signal path and a distortion characteristic typical of an all analog signal path. I fully expect to throw away a couple of bits on the linearity front, but we'll see how that goes. Since the upgrades to the analog end of my system I've found my digital hardware pretty unpalatable sounding by comparison. Interestingly I'd say unbuffered these Sowter 9545 sound significantly better than the carefully built THS4130 based I/V I was using previously, although I've no doubt it measures nowhere as well. (Perhaps around 19 - 20 bits for distortion and a bit better for noise performance, very hard to measure reliably here as the sound card analog inputs are not better.)
FWIW, I have by far the best digital of anyone in the pack I run with locally, and probably if not the best analog very close to it. (Unsubstantiable, and probably unseemly bragging on my part.. 😀 )
I think if nothing else I wouldn't hesitate to recommend these transformers.. 😀
1:1:5:5, I've got the primaries independently wired since the dacs have balanced current outputs, and secondaries wired in series.. This gives me 26dB of gain, almost enough to avoid buffering, not quite enough headroom though - they are intended to used with an amplifier stage.. <sigh>
If I use 4.99 ohm I/V resistors the 6SN7 will give me sufficient gain I think, but if I use the 1 ohm resistors I would need at least 12dB more gain which puts me in the territory of the 5842 or C3G again. Probably need to think about this some more.. 😀
Probably closer 3.5 - 4.0Vrms my system is deliberately designed for higher signal levels in an attempt to make a good tradeoff between dynamic range and noise. Most commercial high bit dacs I suspect the LSBs are below the analog noise floor in the amplifier stages that traditionally follow I/V conversion. You can see evidence for this in the way current output dacs are specified and the recommended design where FS may be 6.0Vrms to in order to achieve something close to the theoretical dynamic range the dac is capable of. (i.e to get the last truly usable LSB above the analog noise floor..)
I try to place as much of the gain as possible nearest to the respective transducer or dac, so I don't follow old nominal standards slavishly.
Maximum output of my phono stage for example is in the range of 3.50 - 4.0Vrms, the dac is intended to be pretty close.
I try to place as much of the gain as possible nearest to the respective transducer or dac, so I don't follow old nominal standards slavishly.
Maximum output of my phono stage for example is in the range of 3.50 - 4.0Vrms, the dac is intended to be pretty close.
Hi Mike,
Here is what I have come up with.. Going with a single 5842 running at 12mA into a 270H lundahl choke, with 1.5V battery bias. I/V resistors are now 2 ohms which puts the transformers in a happy place wrt to core saturation, and results in an output of about 100mVrms which with a single 5842 will result in about 4.5Vrms which is adequately close to the desired output levels.
I've attached the circuit and I plan to eventually power it with Nick's nifty high voltage shunt regulator. 😀 Note that the input transformer wiring as shown is configured for ease of simulation, and that the windings are actually separately driven by the dacs. I will wire them so that the output is overall non-inverting. I have to say I am quite pleased with these transformers.
I will select from amongst my stock of Amperex and Raytheon 5842 to find ones that are in the appropriate transconductance range. Since they last practically forever I should have plenty for the long haul. (I've got more than a dozen nib ones on hand)
I'm currently running with 2 ohm I/V resistors straight into my line stage, sounds very good, but I am at the limit of my volume controls - system gain is maxed out. 32dB of additional gain would be ideal, and I am going to probably be a little over that.
Here is what I have come up with.. Going with a single 5842 running at 12mA into a 270H lundahl choke, with 1.5V battery bias. I/V resistors are now 2 ohms which puts the transformers in a happy place wrt to core saturation, and results in an output of about 100mVrms which with a single 5842 will result in about 4.5Vrms which is adequately close to the desired output levels.
I've attached the circuit and I plan to eventually power it with Nick's nifty high voltage shunt regulator. 😀 Note that the input transformer wiring as shown is configured for ease of simulation, and that the windings are actually separately driven by the dacs. I will wire them so that the output is overall non-inverting. I have to say I am quite pleased with these transformers.
I will select from amongst my stock of Amperex and Raytheon 5842 to find ones that are in the appropriate transconductance range. Since they last practically forever I should have plenty for the long haul. (I've got more than a dozen nib ones on hand)
I'm currently running with 2 ohm I/V resistors straight into my line stage, sounds very good, but I am at the limit of my volume controls - system gain is maxed out. 32dB of additional gain would be ideal, and I am going to probably be a little over that.
Attachments
I should be able to report on my success or failure later Monday evening. I've finished about 90% of the wiring, fitting all of the pieces into that small dac box has been a bit of a problem. Hopefully I haven't created an oscillator.. 😱 5842 can be temperamental at times..
The HV supply is going to have to be in another box as there is just no room for another transformer and supply, but I can control it all from a single power switch. I hope I can still get the "post" sescom chassis to match the dac..
The HV supply is going to have to be in another box as there is just no room for another transformer and supply, but I can control it all from a single power switch. I hope I can still get the "post" sescom chassis to match the dac..
As promised and just squeaking in under the wire..
Initial results are more than satisfactory, and I am enjoying some music off of the server through this dac as I write this.. It easily outperforms the THS4130 into UTC A-20s that I was using previously.
I've included some pix of the mess, err, project, except for the 301 ohm resistors in series with the outputs - this to prevent VHF oscillations with the 5842 due to interactions with the cable capacitance. 😀
The dac boards are TPA boards that I've either modified or bought bare and built my own implementation on their boards. (It was way cheaper this way, although the board design leaves a LOT to be desired - with some tweaks they work quite well.)
The circuit is exactly as described in a previous post.
Initial results are more than satisfactory, and I am enjoying some music off of the server through this dac as I write this.. It easily outperforms the THS4130 into UTC A-20s that I was using previously.
I've included some pix of the mess, err, project, except for the 301 ohm resistors in series with the outputs - this to prevent VHF oscillations with the 5842 due to interactions with the cable capacitance. 😀
The dac boards are TPA boards that I've either modified or bought bare and built my own implementation on their boards. (It was way cheaper this way, although the board design leaves a LOT to be desired - with some tweaks they work quite well.)
The circuit is exactly as described in a previous post.
Attachments
It took quite a lot of effort to figure out how to fit all of that new stuff into the existing box. 😛 There is not a lot of room, but the circuit is so simple that it would either work or not. I'm pretty comfortable with high transconductance triodes and the operating point is actually within < a couple of % my prediction based on curves and a spice simulation.
My biggest concern was I would take what I considered to be a pretty good dac and turn it into rubbish.. Fortunately the new design seems to accomplish both of my goals; a real improvement in subjectively perceived performance, and the removal of the last audio IC in the audio signal path. It certainly seems to dig a bit deeper into the background, I'm hearing details that were previously buried. I've not yet listened to any high res material, but the difference on some very familiar material ripped to flac is rather startling.
It also unfortunately tells me that the analog path is still a problem for digital playback.
My biggest concern was I would take what I considered to be a pretty good dac and turn it into rubbish.. Fortunately the new design seems to accomplish both of my goals; a real improvement in subjectively perceived performance, and the removal of the last audio IC in the audio signal path. It certainly seems to dig a bit deeper into the background, I'm hearing details that were previously buried. I've not yet listened to any high res material, but the difference on some very familiar material ripped to flac is rather startling.
It also unfortunately tells me that the analog path is still a problem for digital playback.
I aim for none, but I needed a chassis ground connection that I thought I oughtn't.. 😀
nice build Kevin......
Thanks Tony! 😀
This was about a six hour build from start to finish. I was a bit concerned about cannibalizing my dac but it worked out OK.
At some point I will do some measurements, but it is extremely quiet and FR peters out around the predicted bandwidth of the 9545 so I am pretty pleased.
This was about a six hour build from start to finish. I was a bit concerned about cannibalizing my dac but it worked out OK.
At some point I will do some measurements, but it is extremely quiet and FR peters out around the predicted bandwidth of the 9545 so I am pretty pleased.
six hours!!! It would take me 6 days (more likely months) to do something half as complicated as that (and that is assuming I knew what I was doing before I started!)
Tony.
Tony.
- Status
- Not open for further replies.