transformer for I/V conversion in DAC?

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

dirkwright

Well-known member
Joined
Mar 24, 2012
Messages
83
Hello, I'm an amateur hobbyist and I'm designing a DAC for my own use. I'm using a current output DAC by TI, the PCM1794A.
http://www.twistedpearaudio.com/digital/cod.aspx
http://www.ti.com/lit/ds/symlink/pcm1794a.pdf

I read, and partially understood, an application note from Analog Devices about using transformers with current output DACs:
http://www.analog.com/static/imported-files/application_notes/AN_912.pdf

After studying this article and what others have done, I have concluded that I really need a high level, low impedance transformer with either a center tapped primary or two primaries that can be put in series. I would prefer a voltage step up also. I looked through the Jensen and Cine Mag catalogs and didn't find anything. I really like Jensen transformers, but in this case it looks like the Lundahl LL1674 is a good choice. I know that K&K Audio use it for this purpose, but I'm not sure if they use it with a current output DAC. (I haven't studied their web site well enough to know yet) I'm aware that Sowter probably makes one that would work as well. There seems to be conflicting requirements for this kind of transformer, from what little I know. On the one hand, line level input transformers are usually high impedance, on the other hand, low impedance input transformers are usually low level, as in microphone input transformers. The expected output voltage on the secondary would be about 2Vrms. So, I want to know if I've exhausted the available transformers or if there are others out there that I don't know of that would work well and maybe cost less than the Lundahl. The Lundahl sells for about $140 each. Any help is appreciated.
 
abbey road d enfer said:
Why you would spend nearly 300 bucks when you can have better performance with $10 of components is beyond me.
This is typical audiophoolery IMO.

.... be ... be ... because TRANSFORMERS ARE BETTER!
 
Well, I didn't mean to start an argument. The AN-912 seemed positive about using transformers for this application.
 
sahib said:
Although you mentioned, have you actually checked what Sowter got?

Yeah, I've looked at some of them. There are a lot to pick from! Is there a dealer in the US for them? Shipping from the UK is not cheap. I've bought transformers from them before and I've been impressed with their quality.
 
This is my first stab at it. The DAC is represented by the current source on the left. I don't have a model for it.
 

Attachments

  • DAC balanced transformer.JPG
    DAC balanced transformer.JPG
    44 KB · Views: 109
This isn't really an argument (yet), but if it becomes one you are on the side arguing against decades of design experience.

I looked at the ap note earlier today, and to be honest I don't see all the benefit claimed for hanging a transformer on a current output DAC.. I couldn't think of anything kind to say so I didn't say anything (earlier).  8)

While some of the claims are true, like the transformer "will" act as a band pass filter, I worry that the source/output impedance used that way will be impractical (high), and that TI current output DAC is not even characterized for use that way. I looked at the app note for part you want to use, and all their distortion data is measured into virtual earth opamp terminations (look at the footnotes on the data sheet).

So a 7+ma p-p output, into a resistor to turn that into a voltage. Then feeding that voltage source into a transformer, means you will need a high enough resistance to make decent voltage swing, while still low enough for decent source impedance. That resistor becomes your source impedance, unless you try to get some voltage step-up from the transformer, then it's that impedance times the turns ratio squared.

Of course don't take our word for it.. It's only money, so build one and let us know how it turns out, but before you drop heavy money on a special transformer maybe look at how just the DAC looks driving open resistor terminations, instead of virtual earth op amp inputs like the DAC app note specifies. 

JR



 
Well, you are correct that the performance of this DAC into a transformer is unknown. Another thing I don't know is how to determine which transformer can handle +/-7mA on their primary without saturating the core.

Also, I assumed that transformers were still commonly used in professional equipment. Is that incorrect?
 
dirkwright said:
Well, you are correct that the performance of this DAC into a transformer is unknown. Another thing I don't know is how to determine which transformer can handle +/-7mA on their primary without saturating the core.
This should be relatively easy; let's say you take a xfmr that's rated for 600 ohms. 7mA peak (14ma p-p) is 5mA rms. Into 600 ohms it computes at 600.5m = 3000m = 3V or +12.7 dBu. Just check your xfmr is rated for at least +15dBu for a 600r type.
In the AN, they target an impedance of 50r, which is a sign that this guy is clueless about audio. Is he designing oscilloscopes or VHF equipment as a day job or what? With a 50r load, the output voltage would be 250mV or -10dBu, which would be correct for most HiFi gear. Now good luck if you want to find an audio xfmr with these specs...
Also, I assumed that transformers were still commonly used in professional equipment. Is that incorrect?
Xfmrs are less and less used in professional equipment because they are very expensive and tend to impair performance in terms of distortion and frequency response. Their only undisputable advantage - galvanic isolation - is less and less necessary, because installations are better and better planned, and the quality of electricity and grounding in modern venues is normalised and enforced.
There are a number of people (particularly in broadcast) who swear by xfmrs, though. I must admit that, in an OB configuration, an isolation xfmr may be a life-saver. But on any planned installation, there is no reason to have a single line-level xfmr.
There are many audio boxes that use xfmrs; it does not necessarily reflect any level of "professionalism". In most cases, (particularly solid-state devices) it is as technically unjustified as having a mechanical VU-meter instead of an LED array.
 
dirkwright said:
Well, you are correct that the performance of this DAC into a transformer is unknown.
The kit PCB you show talks about a passive 382R to convert I/V with no performance data listed.
Another thing I don't know is how to determine which transformer can handle +/-7mA on their primary without saturating the core.
You are not putting +/-7mA into a transformer. The spec says differential current of 7.8mA p-p. So differential means one output swings up while the other down summing to 7.8mA peak to peak. So more like + 3.9mA and - 3.9mA. So driving this current into the passive 382 termination will be around 1.5Vp or 3Vp-p.  This seems practical* from a max 6.5V power supply. (while I still have no idea how linear the current source will be used that way.)

You need to be looking for a transformer that accepts roughly 3vp-p with 750 ohm source impedance.
Also, I assumed that transformers were still commonly used in professional equipment. Is that incorrect?
I have avoided even input transformers since the late '70s in professional recording products, while many professional products used for live sound reinforcement benefit from transformers to deal with rouge facility problems. Transformers are still fashionable for some high-end recording products.

There is a segment of the recording market attracted to transformers for their euphonic coloration. I make a distinction between designing a linear path (straight wire with gain), and an "effect" that intentionally sounds different and presumably "better" than the dry linear path.

If designing to sound different than clean linear, your transformer has some merit. To "accurately" convert digital 1s and 0s to linear audio signals, not so much.

Enjoy.

JR
 
While I very much agree that there is little objective support for the use of a transformer for this application I'd at least use a "zero field" transformer (e.g. LL7101) such that the DAC output sees a nice virtual earth.

Samuel
 
ok this is a subject that polarises people.
some people might say "too much colour" - "colour" is good and has its purpose in the recording arts.

check this out
http://www.sowter.co.uk/dacs.php

Brian also recommends 3195 or 4935 for more budget conscious.
you need a centre tap on the primary for the PCM1794a because it has a bias voltage on the output.

Now we all know in audio something might not measure up well (like all the old classic analogue gear) but sounds like bliss.
I definitely want to try this out but the audiofool kits are completely insulting.

dirkwright are you going to post your PCB layouts here??
It might be worth doing in a modular manner. One pcb is the basic DAC and then you can add output stages to that like discrete opamp IV convertor or just simple transformer out.
Fred Forsell does the DOA outputs on his high level DACs - as do Burl audio.

the beauty of the current output DACs is that you can add your own output stage as opposed to a voltage output DAC which has miniaturised opamps as the IV convertors. I am sure we can do better than that. Ideally you would chose your favorite output stage.

DACs and ADCs are topics barely been touched upon in this forum. Why should we settle for what is available on the commercial market. We certainly dont limit the quality of our signal chains this way. we need more roll your own. The more experimentation the better.






 
tardishead said:
ok this is a subject that polarises people.
some people might say "too much colour" - "colour" is good and has its purpose in the recording arts.

check this out
http://www.sowter.co.uk/dacs.php

Brian also recommends 3195 or 4935 for more budget conscious.
you need a centre tap on the primary for the PCM1794a because it has a bias voltage on the output.

Now we all know in audio something might not measure up well (like all the old classic analogue gear) but sounds like bliss.
I definitely want to try this out but the audiofool kits are completely insulting.

dirkwright are you going to post your PCB layouts here??
It might be worth doing in a modular manner. One pcb is the basic DAC and then you can add output stages to that like discrete opamp IV convertor or just simple transformer out.
Fred Forsell does the DOA outputs on his high level DACs - as do Burl audio.

the beauty of the current output DACs is that you can add your own output stage as opposed to a voltage output DAC which has miniaturised opamps as the IV convertors. I am sure we can do better than that. Ideally you would chose your favorite output stage.

DACs and ADCs are topics barely been touched upon in this forum. Why should we settle for what is available on the commercial market. We certainly dont limit the quality of our signal chains this way. we need more roll your own. The more experimentation the better.

Thanks for the support and encouragement everyone! There is much to think about for this project.

On the digital side, I'm using the Twisted Pear COD board, so only the I/V converter and output stage are my responsibility.

Sure, I can post board layouts but it depends on which transformer I pick, since the Lundahl's are mounted on PCB's and the Sowter's are not, usually. I've modeled two versions so far. One uses a step up transformer and a unity gain follower stage and the other uses a 1:1 transformer with some gain in the following stage. The one with the step up and no gain measures better but in the real world is unknown which is better. Both are completely balanced end to end.
 
JohnRoberts said:
I looked at the ap note earlier today, and to be honest I don't see all the benefit claimed for hanging a transformer on a current output DAC.. I couldn't think of anything kind to say so I didn't say anything (earlier).  8)

Look at the source of the document.

AD has very few audio DACs. They do have lots and lots of 100+Megasample/sec DACs (both as separate parts and as outputs for their DDS products). At >>100MHz, it is very hard to do a wideband I/V stage in silicon that doesn't impact converter performance more than an RF transformer would. At audio frequencies, not so much.

JD 'context' B.
 
jdbakker said:
JohnRoberts said:
I looked at the ap note earlier today, and to be honest I don't see all the benefit claimed for hanging a transformer on a current output DAC.. I couldn't think of anything kind to say so I didn't say anything (earlier).  8)

Look at the source of the document.

AD has very few audio DACs. They do have lots and lots of 100+Megasample/sec DACs (both as separate parts and as outputs for their DDS products). At >>100MHz, it is very hard to do a wideband I/V stage in silicon that doesn't impact converter performance more than an RF transformer would. At audio frequencies, not so much.

JD 'context' B.

Thanks that explains why they wrote it, still not why to use it (for audio).. but again whatever floats your boat.

@tardishead it seems to me if there is any transformer magic to be found, why not make a dedicated stand alone transformer magic black box, that you can add to a linear path, or not as taste dictates. When you build it into the primary path it is always there.  I like effects to have bypass switches.

JR
 
yeh its all down to taste I suppose. Personally I always wanted a digital rig that sounded close to my Ampex MM1100. I have spent lots of money on nice digital rigs over the years and compared to the Ampex they all leave a lot to be desired.
My whole signal chain is discrete analogue equipment. When it hits digital it always sounds disappointing. It sounds "like" what it should do but not quite close enough. I have tried the Forsell digital stuff and some from Burl audio and that to me sounds miles better than the ususal "high end" ADC and DACs. I was very impressed - since then I have been on a mission to find out more and considered the DIY route.
 
So you want to emulate the nonlinearity of magnetic tape?  Again that sounds like an effect that would benefit from a bypass switch.

The engineers designing tape recorders would have given their left nad to realize modern digital performance, without magnetic tape's sundry limitations.

Note: I do not dispute the merit of selectively applying tape compression to some kinds of sources. But that is little reason to saddle the entire path with that transfer function.

JR
 
Yes thats what I state a few posts before - would be good to be able to add DOA output or transformer output to taste.

Its down to what you want. I have enough "clean" convertors - looking for a different sound. Its like different mic preamps for different sounds.

No I don't necessarily mean tape compression and non linearities............I can see 35khz measure up on my Ampex. Could not say the same for digital - can you???

I totally see your point but this website is about customising to your tastes.
Speaking as a studio owner for nearly 20 years -  what people consider "clean digital" is just as much a colour as anything else.
 
tardishead said:
Yes thats what I state a few posts before - would be good to be able to add DOA output or transformer output to taste.
That would satisfy my bypass switch request.
Its down to what you want. I have enough "clean" convertors - looking for a different sound. Its like different mic preamps for different sounds.
I expect more useful differences between microphones, than between mic preamps, where the differences are more likely to be inaccuracies reproducing the inputs they receive.
No I don't necessarily mean tape compression and non linearities............I can see 35khz measure up on my Ampex. Could not say the same for digital - can you???

If it was useful, it would be available, digital is capable of higher bandpasses when desired, and paid for. IMO  higher sample rates in current digital studio gear is more a fashion statement or to satisfy uncertainty about whether it may be better. 

I totally see your point but this website is about customising to your tastes.
I guess I missed that  :-[ . It appears this website is about different things to different people.
Speaking as a studio owner for nearly 20 years -  what people consider "clean digital" is just as much a colour as anything else.

I don't argue with people about what they hear, so I won't argue about what they think other people think or hear.

Clean linear reproduction is an objective metric that can be used in design. While for recording, this is just the clean stew pot a good chef starts with, not an end target.

I designed some studio efx gear (Loft delay line-flanger) 35 years ago, and efx are all about how they sound, not how they measure. Different strokes for different applications. 

JR
 

Latest posts

Back
Top