Post-Xmas update
Having been crouched over multiple alpha-builds with soldering iron at the ready for about a week I figure I'm about done with component tweaking 😀 I've decided to keep the EP17s for the first released version but I was able to obtain some thinner wire for the secondary (we're now down to 0.04mm diameter) which raises the inductance and hence gives better flatness in the bass. The I/V resistors are down from 5k6 to 2k meaning there's almost 3X less signal level at the output of the EP17s - but 1VRMS is probably enough from a DAC for many. 2VRMS can be reached by making the OPTs a 1:2 ratio - going higher than 2X gives too high an output impedance with the cheap RM10 cores I'm using. Some of the signal level is made up by reducing the step-up ratio of the EP17s - its about 1:27 now.
Having gotten the distortion into the acceptable range I then found my simulation results for the FR were astray so the filter's been tweaked too, along with the RC networks on the DAC board. The end result isn't ruler-flat but it is an improvement on the unfiltered NOS DACs out there in the top couple of octaves - Audacity plot attached.
Having been crouched over multiple alpha-builds with soldering iron at the ready for about a week I figure I'm about done with component tweaking 😀 I've decided to keep the EP17s for the first released version but I was able to obtain some thinner wire for the secondary (we're now down to 0.04mm diameter) which raises the inductance and hence gives better flatness in the bass. The I/V resistors are down from 5k6 to 2k meaning there's almost 3X less signal level at the output of the EP17s - but 1VRMS is probably enough from a DAC for many. 2VRMS can be reached by making the OPTs a 1:2 ratio - going higher than 2X gives too high an output impedance with the cheap RM10 cores I'm using. Some of the signal level is made up by reducing the step-up ratio of the EP17s - its about 1:27 now.
Having gotten the distortion into the acceptable range I then found my simulation results for the FR were astray so the filter's been tweaked too, along with the RC networks on the DAC board. The end result isn't ruler-flat but it is an improvement on the unfiltered NOS DACs out there in the top couple of octaves - Audacity plot attached.
Attachments
I agree. There is no reason to follow Red Book 2V RMS requirements. This plot is quite good in low frequency. Once distortion issue is sorted out, it looks like the project will be finished in two or three months. I can't wait any longer. 🙂
All the best in New Year to you and your young wife!
BTW, For the last two weeks I was enjoying music coming out of the KEF Model One speakers and IcePower 125ASX2. KEF - old one, but in a pristine condition. A little Nobsound 8xTDA1387 $42 DAC delivers all required details together with emotions that modern DACs do not do. For instance, I can hear second row of violin in Mozart Symphony No. 40.
On these speakers a distance between Nobsound and Topping D30 (that disappointed me completely) increases even further. What does mean "clean sound", low noise/THD, completely flat frequency plot, top position in ASR review, if I can't hear second row of violin on the Topping? Integrated sound chip in the Sony VIAO laptop (Realtek chip!!!) comes out better than D30 in my opinion. There is no reverbation on a decay of R2R DAC, though.
All the best in New Year to you and your young wife!
BTW, For the last two weeks I was enjoying music coming out of the KEF Model One speakers and IcePower 125ASX2. KEF - old one, but in a pristine condition. A little Nobsound 8xTDA1387 $42 DAC delivers all required details together with emotions that modern DACs do not do. For instance, I can hear second row of violin in Mozart Symphony No. 40.
On these speakers a distance between Nobsound and Topping D30 (that disappointed me completely) increases even further. What does mean "clean sound", low noise/THD, completely flat frequency plot, top position in ASR review, if I can't hear second row of violin on the Topping? Integrated sound chip in the Sony VIAO laptop (Realtek chip!!!) comes out better than D30 in my opinion. There is no reverbation on a decay of R2R DAC, though.
What do you mean 'project finished in two or three months' ? Projects never get finished, they just get put aside for other more interesting avenues to explore 🙂 I think I will get to release this one into the wild sooner than that though, I just need to build a stock of parts and create the BOM, schematic. Oh and re-spin the PCB. So hopefully before the end of Jan - which here is Chinese New Year.
Interesting that you think the low frequency is quite good, that's the part I am wondering where some measurement artifact has affected the results. I need to investigate further but the plot is showing -1dB @50Hz which seems too soon for the roll-off, at least compared to my sims which were showing (from memory) no more than -0.25dB there.
Interesting that you think the low frequency is quite good, that's the part I am wondering where some measurement artifact has affected the results. I need to investigate further but the plot is showing -1dB @50Hz which seems too soon for the roll-off, at least compared to my sims which were showing (from memory) no more than -0.25dB there.
Last edited:
Yup, another dac is just what I don't need, but damn I want one🙂
Abraxalito, what happens when a dac chip dies, if they are in parallel it would be impossible to find the dead one wouldn't it?
Abbrax,
Without going into too much detail on transformers, the short of it is you are heading in to somewhere between MC step up and Ribbon Microphone step
up territory when looking at your transformer turns ratio and DAC OP drive impedance.
There are quite a few factors that control distortion of these very specialized high ratio transformers. My advice is to contact Cinemag and ask what they
recommend WRT core types, impedances, snubbers, transformer loading etc etc. These guys have a huge amount of experience making low distortion,
high ratio transformers and will know how to optimize for your application.
I have used them for various Pro Audio Apps in the past, their transformers
are very high quality, they are super helpful and their prices are very
reasonable.
Don't re invent the wheel. 🙂
TCD
Whilst I've no doubt those guys are experts indeed, why should they want to share their proprietary hard-won information with me when I'm not at all likely to be a customer? Putting a pair of Cinemags on my BOM will completely swamp all the other parts costs.
@avtech23, @Luke - thanks muchly for your enthusiasm for this project 🙂 To answer your question Luke, post #23 has my experience in fault-finding a particular faulty DAC chip. So long as the dodgy chip is still responding to its pin7 reference its a fairly straightforward operation to track it down.
@avtech23, @Luke - thanks muchly for your enthusiasm for this project 🙂 To answer your question Luke, post #23 has my experience in fault-finding a particular faulty DAC chip. So long as the dodgy chip is still responding to its pin7 reference its a fairly straightforward operation to track it down.
Last edited:
It depends on what you consider expensive. A large part of the performance
will be a result of using VHQ core material and there is no real way of getting
around this if your goal is to have a very transparent transformer. They will
be sourcing HQ cores much cheaper than you can. They also offer great qty
discounts. I also believe there is a certain degree of proprietary knowledge
with heat treatment of core materials for lowest distortion which Jensen and Cinemag keep guarded.
As far as using cheaper iron-mix core materials is concerned, these are also
great transformers but they are colored. I use them in certain jobs where
the customer wants the added 'vibe' or color as an option.
One other option is Nanocrystaline cores (BYO winding) which appear to have
very good magnetic properties and aren't hugely expensive from what I've seen. However, I have not found any apples to apples measurements
compared with Hi Nickel cores, just lots of audiophile talk (yawn).
If you want to cut to the chase with transformer performance, look at 20Hz distortion. That will tell a lot.
TCD
Expensive is relative to all the other parts on the BOM, there would be something like a 40dB difference in price between a Cinemag trafo and any other single item. If I were aiming at Cinemag levels of performance and price I'd not start out using dirt-cheap TDA1387s but rather something more like AK4497 or even AK4499.
There's little doubt that at the price levels I'm interested in, there's going to be colouration, the question then is really how subjectively acceptable that colouration is.
On 20Hz distortion, yes that's where I've been focussing. The lowest frequencies are where the highest flux levels are found.
There's little doubt that at the price levels I'm interested in, there's going to be colouration, the question then is really how subjectively acceptable that colouration is.
On 20Hz distortion, yes that's where I've been focussing. The lowest frequencies are where the highest flux levels are found.
Yes, 1/2 freq, double the flux and so on.
So how much are you willing to pay for a decent (1/2 decent) transformer?
TCD
For this particular design, very little. In fact my aim is that an interested builder could wind one themself from the details I provide. The parts cost target is <$1 which is fairly easily achieved with an EP17 core. Labour will of course increase that figure substantially but for the DIYer, labour isn't an accounting cost.
50,000 uF
I wuz just leafin' thru these pages ... and saw that y'all once had massive couplin' on that pin.
Oh ... uh ... hi, again, toots.
I wuz just leafin' thru these pages ... and saw that y'all once had massive couplin' on that pin.
Question is how much we (DIY guys) are willing to pay. How much Cinemag transformers do cost $10-$100? In this post abraxalito wrote he was experimenting with EE35. I would accept -1dB@50Hz, but if colouration can be reduced, maybe EE35 is a way to go without significantly increasing cost.
Today I made some measurements of distortion with EE35 - definitely its lower at low frequencies by 10-20dB. I'll pencil in EE35 for the next revision of the project, it won't increase the BOM cost very much but PCB real-estate goes up quite a bit compared to EP17.
-- LOOKING GOOD !!! -- Not much i can contribute technically - but I can lead the cheering section - Thank you! For using your expertise to do what many of us can't ~ !!!!!!!
-
- 
- - - - - 😀
-
- - - - - - 😀
Last edited:
Progress update - new PCBs are in
We received the re-spin of the PCB today, first build is taking shape. Here is the updated schematic - I've been playing a lot with transformers and the output filter's been updated somewhat as well as test points added for ease of build/debug.
Speaking of transformers - its not looking like we can get an external manufacturer lined up before Chinese New Year, the thin wire is discouraging manufacturers from quoting on the EP17s. We can wind some ourselves initially but I seem to have given my right arm something like RSI from doing way too many turns on my machines....😱
We received the re-spin of the PCB today, first build is taking shape. Here is the updated schematic - I've been playing a lot with transformers and the output filter's been updated somewhat as well as test points added for ease of build/debug.
Speaking of transformers - its not looking like we can get an external manufacturer lined up before Chinese New Year, the thin wire is discouraging manufacturers from quoting on the EP17s. We can wind some ourselves initially but I seem to have given my right arm something like RSI from doing way too many turns on my machines....😱
