|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
115.132.40.123
In Reply to: RE: cMP - the open source high-end Memory Player posted by cics on December 30, 2007 at 05:42:01
Proudly presents FIRST TUBE DIGITAL sound card in the world!
Highly appreciated to :
God of digital Mr. Lukasz Fikus (Dr. Lampizator)
Jackwong96
My guru
Edits: 06/15/10Follow Ups:
Hi,
> Proudly presents FIRST TUBE DIGITAL sound card in the world!
If you count external soundcards I think I got you beat to the first:
http://www.diyhifisupply.com/node/749
http://www.diyhifisupply.com/node/794
BTW, for the doubters. As standard the Musiland uses multiple parallel Buffers as driver. Comparing the signal from that with the Tube Stage shows the tubestage has a much faster risetime and a very clear trace.
Certainly the signal is much better. How much that impacts on sound quality - who can say, who wants to say? Subjctively, using a DAC with a cirrus logic receiver the Tube stage plus musiland sounds cleaner and less edgy, compared to the already heavily modified unit which normally gives a quite clean trace already.
Ciao T
Sometimes I'd like to be the water
sometimes shallow, sometimes wild.
Born high in the mountains,
even the seas would be mine.
(Translated from the song "Aus der ferne" by City)
Quote:
Question: What makes the tubed usb to spdif converter better the non tubed version?
Answer:
The tubed output stage operates differently to the solid state logic one. The Tube stage operates as linear Class
Amplifier and does not "saturate". The normal logic IC based output buffer circuitry operates in "saturation" meaning
it is slower to react to signal changes. The Tube stage also has much less input capacitance than CMOS logic IC's, so
the rise-time of the signal is faster.
The technical result is a signal output that is cleaner and closer to the theoretically "ideal" SPDIF Signal.
As a result the DAC's SPDIF receiver can recover the signal and clock with less jitter, in technical terms. This in turn
generally produces a better, more natural and non-fatiguing sound.
Sreaming audio efficiently is a key design goal for cMP².
Hi,
Another factor is the current that the SPDIF output circuitry draws dynamically if it is used.
Driving a minimum capacitance short twisted pair and a cathode follower tube draws much less current than a resistive divider.
All in all SPDIF connections are not easy to get right, but they are what everyone seems to insist on using...
Ciao T
Sometimes I'd like to be the water
sometimes shallow, sometimes wild.
Born high in the mountains,
even the seas would be mine.
(Translated from the song "Aus der ferne" by City)
I'm tube inclined and using DHT in my DAC as well as the entire system. I like the idea of a tube SPDIF buffer, however, the tube buffer circuit in question fall short of a good design. This is simple electronics but no magic. I believe the true potential of a tube SPDIF buffer is yet to be unveiled, with improvement of circuitry, such as White's Cathode Follower.
Hi,
> I like the idea of a tube SPDIF buffer, however,
> the tube buffer circuit in question fall short of a good design.
How do you know? Did you see and analyse my schematic for the module diyhifisupply sells? They only came back from the factory.
> This is simple electronics but no magic.
Yup. We just need around 0.5V PP into 75 Ohm. That is 6.66mA Peak-Peak, so we need to have a bit more than halve that as quiescent current, not very hard to do.
> I believe the true potential of a tube SPDIF buffer is yet to
> be unveiled, with improvement of circuitry, such as White's
> Cathode Follower.
I tested my buffer as under 1pS Risetime including the the drive signal from the FPGA. How would you like to improve this?
Ciao T
Sometimes I'd like to be the water
sometimes shallow, sometimes wild.
Born high in the mountains,
even the seas would be mine.
(Translated from the song "Aus der ferne" by City)
Those are impressive risetimes at 1ps! Here's the digital waveform of a 1kHz tone (at 96000 Fs) done a few years ago:
Risetimes well into 10s of ns together with over- & under-shoot.
Due to commercial obligations you're unable to share your circuit design, however it would be interesting to see equivalent from your DAC's SPDIF output.
Hi,
Apologies for the delay in posting that, I just had to get around to measuring this and to taking a picture...
Here is a picture of a 48KHz signal from a commercial USB-> SPDIF converter which incorporates my design, the tube stage is fed directly from the FPGA output pin.
This picture is at a pretty extreme degree of zoom in, compared to most we see posted here. The whole picture covers around 1/3rd of a single period of a 3MHz signal (the base rate is 48KHz).
This trace is for the full system, that is USB-SPDIF converter, with tube stage, Canare "75R" RCA connectors (both ends) and 2m Cable. It tests the signal at the 75R terminated "far" end of the cable. I find this gives a far better representation of the actual performance than just testing the output.
Note that the loaded output voltage is just a touch over 0.8V Peak-Peak BTW.
Ciao T
Sometimes I'd like to be the water
sometimes shallow, sometimes wild.
Born high in the mountains,
even the seas would be mine.
(Translated from the song "Aus der ferne" by City)
Hi,I need to take a new trace when I get to test another unit. But the traces shown above are not that good at all.
I previously posted on my modifications for the Musiland Monitor 01 USD, which includes the traces from my mods without the tube section.
http://www.audioasylum.com/forums/pcaudio/messages/6/66884.html
The tubed output stage does better. The figure comes from the NEC Digiscope. I did a doubletake too, but the numbers came up like this.
I think it calculates the rise time from 30% to 70% around the zero crossing. Still, the output is very square, a lot better than the above.
Ciao T
Sometimes I'd like to be the water
sometimes shallow, sometimes wild.
Born high in the mountains,
even the seas would be mine.
(Translated from the song "Aus der ferne" by City)
Edits: 06/17/10
I used a cheap Creative soundcard (which outputs TTL and yes its crap) for that measurement (using ELAB-080 DSO). Your Musiland trace looks good and even better after mods.
A tube output stage makes a lot of sense. Given that BMC encoding is used for both S/PDIF & AES/EBU and they can be readily converted to/from, can the tube design be extended to AES/EBU?
Hi,
> can the tube design be extended to AES/EBU?
Sure, it just needs a different wiring up. This will all go into the Manual.
Ciao T
Sometimes I'd like to be the water
sometimes shallow, sometimes wild.
Born high in the mountains,
even the seas would be mine.
(Translated from the song "Aus der ferne" by City)
hey Bro, remember dr. fikusz had tried on using AES/EBU however the result is NEGATIVE..
What he said:
I am sure that a single ended connection is best for S/PDIF signal. I do not like toslink and I don't like AES/EBU either.
The "pro" looking AES/EBU connection is using XLR and it looks OH SO PROFESSIONAL WOWIE ZOWIE. We buy by the eyes. But the AES/EBU is not invented to be better. It is invented to be LONG DISTANCE.
The RS422 industrial data transmission from which AES/EBU derives can travel as long as 1,8 km (over one mile) on the twisted pair balanced cable. The S/PDIF in coax cable can travel up to few single meters. THAT IS THE DIFFERENCE. But the S/PDIF signal on both ends exists ONLY as simple single ended asymetrical signal. We must make it balanced artificially by means of adding a balancing transformer which badly distorts the signal. Then we use in our systems one metre cable (not one mile !) and we must use de-symmertization transformer on other end again. That's two unnecessary transformers, two too many for me.
In analog systems, the ballanced XLR has more meaning - as described HERE, but in digital - the signal itself is never trully ballanced, it is the sending media that is ballanced to make long distance possible.
so why is the AES/EBU sounding better on many transport/dac combos ?
Answer is easy. Because the S/PDIF on all transports that I have seen, and I have analyzed about 40 different ones - is made wrongly and it uses an ugly separation transformer as well. So in that scenario - S/PDIF can not beat AES/EBU because they both have the biggest limiting factor - the transformer."
Like my previous post, removing the pulse trans will help in sound revealing transparency.
Hi,
> hey Bro, remember dr. fikusz had tried on using AES/EBU
> however the result is NEGATIVE..
Yes. But that this mean it is universally applicable?
Generally when done to equal levels of implementation, AES/EBU usually sounds better.
> But the S/PDIF signal on both ends exists ONLY as simple
> single ended asymetrical signal.
Actually, wrong. Almost all receivers out there have balanced inputs.
> We must make it balanced artificially by means of adding a
> balancing transformer which badly distorts the signal.
Wrong again. AES/EBU can be made "transformerless", secondly, the traces shown from the Musiland (as well as the Tube Stages I designed) use transformers. Transformers only distort the signal if used wrongly.
The key to understanding is that all these interfaces are radio-frequency designs. We can make wireless routers that work reliable at 2.4GHz. The 25MHz needed for SPDIF and/or AES/EBU are actually quite easy, this merely Shortwave radio frequencies.
> Then we use in our systems one metre cable
Ouch. A 1 meter coax cable is precisely the worst length or an SPDIF cable.
> Like my previous post, removing the pulse trans will help
> in sound revealing transparency.
Actually, you may wish to read the blog posts over at www.diyhifisupply.com. You are right insofar - removing the original tranformers (which are basically inappropriate) improved sound quality.
HOWEVER, correctly implementing some higher quality transformers improved the sound quality even further.
Then getting rid of the two daisy chained inverters that drive the digital outputs on the Musiland and using instead a linear tube stage while retaining the transformer produced further improvements.
The bottom line - as I often say, it is not what you do (e.g. use transformers in SPDIF or AES/EBU interfaces) but how you do it that makes the difference.
Ciao T
Sometimes I'd like to be the water
sometimes shallow, sometimes wild.
Born high in the mountains,
even the seas would be mine.
(Translated from the song "Aus der ferne" by City)
Hi Thorsten
> HOWEVER, correctly implementing some higher quality transformers
improved the sound quality even further.
What output transformer would you recommend for the Juli@ to replace the Hanrun?
Regards
Hi,> What output transformer would you recommend for the Juli@
> to replace the Hanrun?First, I do not know the Juli@ first hand, I have no idea of circuit implementation etc. So it is impossible to make an informed suggestion.
Second, simple transformer swapping gets you nowhere fast.
You need to be able to find a transformer that is sufficiently consistent in production (harder than you think) and characterise it's parasitic (unwanted) behaviours and then design a circuit that gets the best out of this transformer.
For example people ask me "Which transformer do you use for your musiland mod's, I want to try them."
My answer is always:
"Knowing Brand and type will not help you, unless you know my exact application circuit, which is not for public dissemination. And to avoid negativity from people using the transformers wrong and then complaining they are not good and this reflecting back badly on my designs I'm not telling the first part either."
I hope you understand.
Ciao T
Ciao T
Sometimes I'd like to be the water
sometimes shallow, sometimes wild.
Born high in the mountains,
even the seas would be mine.
(Translated from the song "Aus der ferne" by City)
Edits: 06/18/10
hi there... mind to show your tube circuit of buffer stage?
Hi,
> hi there... mind to show your tube circuit of buffer stage?
The circuitry is applied to a commercial product, so I am not at liberty to disclose it, sorry.
Ciao T
Sometimes I'd like to be the water
sometimes shallow, sometimes wild.
Born high in the mountains,
even the seas would be mine.
(Translated from the song "Aus der ferne" by City)
Thorsten, Your tube buffer design is perfect! I believe it can drive a 75 ohm headphone too. In such case, not much I can contribute.
Edits: 06/16/10
Hi,
> I believe it can drive a 75 ohm headphone too.
It can. To 0.5V Peak-Peak. Or 4mW RMS. And I doubt the output transformers are capable to go down into the audio range...
Ciao T
Sometimes I'd like to be the water
sometimes shallow, sometimes wild.
Born high in the mountains,
even the seas would be mine.
(Translated from the song "Aus der ferne" by City)
there are a couple of questions (I presume you've copied the spdif tube buffer circuit from the Lampizator site):
1. a tube cathode follower (CF) cannot drive super low impedance satisfactorily, (75 ohm in your case), although a Whites CF with suitable tubes may drive as low as 100 ohm. A good tube circuit design requires that loading of a tube to be 3 to 5 times the Anode load, the same applies either to Anode follower or Cathode follower circuits. This CF circuit has a loading of 150 ohm, and the tube (looks like a Chinese 6N1 or Russian 6H1) has a Anode resistance of a 2-3 Kohms. The load line of this CF circuit is very steep, it has a low voltage swing and high current characteristic, low but not super low impedance. The tube will distort more than a traditional design due to in appropriate cathode loading.2. I presume the SPFIDF signal is taken from Tx Pin 16. Suppose the CF has a gain of 0.9, the 3.3V SPDIF signal will be reduced to 2.97V and than halfed by the resistor network=1.485V. This exceeds the SPDIF spec. requirement(0.5 to 0.6 V peak to peak voltage, i.e. 1.2V).
The tube buffer circuit is fun but far from ideal. You should conduct blind listening tests with and without the tube buffer, better ask someone to listen rather than yourself. Without tube buffer means taking signal from the Tx with the recommended resistor network on P.22 of the data sheet, with or without transformer. Let us know the result.
Edits: 06/16/10
1. i am not sure bout it. its abit technical however, one of my friend was using sony transport he find out the sony transport has this transistor buffered circuit. After that, he also follows digilampizator in his another philips CD Player just that with positive result!
2. Yes the output is higher, but it doesnt effect the sound. my journey of this was:-
i.) Signal out from the output trans. "Sound is normal"
ii.) Signal without the output trans. (impedence measure was 100ohm!) More revealing)
iii.) Signal with multiplexer bypass with 75ohm resistor to ground. "Very transparent!"
iv. )Signal from TUBE buffered circuit. "This is where magical inspiration take place, similar transparency as without multiplexer but with more liquid. perfectly defined highs, deep and powerful precice bass, huge stage, you can really pinpoint the instruments in a orchestra, different is night and day. Its not upgrade of the sound but like changes entire sound.
Great! Congratulations! Please construction's details, schematics and what about SQ ? You use an external clock for Juli ?
due to the resource limitation, i just do anything more than capturing using my cheap HandPhone. U don't listen to the recorded video quality.
Hmmm, however i am truly blive u can simply compares the flow of music without looking to in-depth details.
This is how a music flow should sound.
I don listen michael jackson in CD format b4 i tube the digital, this is bcos Michael Jackson sounds not it shoud be comparing to VINYL that i have listen.
BUT.! listen here, this tube stage brings back the life of music. Mentioned that it realigns the ingredients of MUSIC.
So... Yes i do enjoy listen to more music!
Good music flow! I'm with you, in Hong Kong.
Clock: Modded LCLOCK XO3 with dedicated power supply.
SQ
i.) Signal out from the output trans. "Sound is normal"
ii.) Signal without the output trans. (impedence measure was 100ohm!) More revealing)
iii.) Signal with multiplexer bypass with 75ohm resistor to ground. "Very transparent!"
iv. )Signal from TUBE buffered circuit. "This is where magical inspiration take place, similar transparency as without multiplexer but with more liquid. perfectly defined highs, deep and powerful precice bass, huge stage, you can really pinpoint the instruments in a orchestra, different is night and day. Its not upgrade of the sound but like changes entire sound. This is the way for a computer transport to inject musicality like CD transport!
in fact audio note does this.
You're sure you have connected LCLOCK XO3 correctly?
wow.. thanks for ur concern, hmmm i think it should be correct but without the 0.1uf coupling cap. Does that a matter?
OK. I nevertheless would mount the capacitor 0.1 µF.
Congratulations. Yet another step forward!
I'll put this on our site in Advanced section under Juli@ mods.
I hope you will like it and I am happy of contribute and bring us to a better enjoyment!Please let me know if there is any correction.
GOOD LUCK!
Edits: 06/19/10
I done some calculation of the circuit according the tube data and textbook reference:6n14P data: http://frank.pocnet.net/sheets/113/6/6N14P.pdf
mu=25 S=6.8 Ra = mu/S = 3.67K
Output impedance of a cathode followerRout = Ra / mu+1
= 3.67K / 26 = 141 ohm
Voltage gain of a cathode follower
Where Rk= (75+75)=150 ohm
V= (mu * Rk) / (Ra + (mu+1)*Rk) = 3750/7570 = 0.495
With a 3.3V input, this output will be about 0.4V PP.
I built the circuit to test its performance. Since I don't have any 6n14P, I used an equivalant tube, ECC84, instead.
Below pix showed the signal at the input (upper trace) and the tube output
(lower trace). The singal was drawn directly from the FPGA and hex inverter(74hcxx)was removed. No loading was applied to the tube output.
Below pix showed the signal at the input (upper trace) and the tube output
(lower trace) wuth a 75 ohm resistive load applied.
Great! This tube buffer can handle a 75 ohm load, albeit at a reduced output(about 0.32V PP). Not bad for such a simple buffer circuit.By the way, LGA 775, there are some mistakes on the power supply circuit.
The bridge rectifier was not correctly connected, and AC is applied directly to the caps.
I do not know the application of the voltage stablizer in your circuit as it was placed in series with the B+ rail. It is supposed to be connected in parellel with the PS so as to stabilize voltage. It acts like a shunt regulator and a resisitor needs to be put in series with it whereas the cap in parellel should suitably reduced. More, if it is intended to be used as a shunt reg. the rating of 75V does not seem sufficient for this circuit. 125V would be better. please clarify.
Finally, the filament needs a voltage reference. In this case, reference to the ground.
Edits: 06/21/10 06/21/10 06/21/10 06/21/10 06/21/10 06/21/10 06/21/10 06/21/10
Did tried with Resistor as CRCLC, very bad ripple filtering capability.
Sound is not good. when u use metal film will sound brighter and sharp.
Using carbon will sound not transparent.
although B+ has got 75V, but still better than previous circuit.
Higher capability of filtering ripple and more clean sound.
Edits: 06/23/10
.
hi there.. i was quite shocked when i look at ur measured scope that there is some fluctuating. hmmm just wanna report that yes the documentation has a mistake at the AC-DC rectification. After the 0C2 tube there is another shunt 0.1uf bypass capacitor.
Also i measured my B+ i got 115V nominal with +/-1% voltage ripple. Absolute max voltage swing was +/-2V. Inital startup B+ was about 131-132V.
The tests continue. The below pix shows a stock Musiland 01 USD with its BNC output connected to a 75 ohm (3-feet) cable, terminated at the other end with a two 150 ohm resistors in parellel. Not bad for this cheapee! Although noise is a bit high.
Next, a waveform test for symmetrical output. A saw wave is used. Upper trace 3.3V PP to the input of the tube buffer, the lower trace about 0.4V PP take at the output. See the voltage difference between the positive and negative swings of the lower trace.
The below pix shows the output swing at about 1V PP, the diffference is more profound. The negative swing is clipping. The buffer distorted. If you look carefully at the last pix on my previous post, you will notice the same.
What next? As a loading of 150 ohm is extremely low for the tube which has a RP of 3670 ohm, the cathode resistor value should be increased....
Edits: 06/22/10 06/22/10 06/22/10 06/22/10 06/22/10 06/22/10 06/22/10
> The negative swing is clipping.The buffer distorted. [What is the significance of this distortion? Will it manifest in incoherence in the music?]I havn't conducted a listen test on the distorted signal, but I guess I would not like it.
> What next? As a loading of 150 ohm is extremely low for the tube which has a RP of 3670 ohm, the cathode resistor value should be increased....
[Any suggestion to address this shortcoming?]
The tests continue. The ECC84 did not give a good voltage swing and its output impedance was high. Therefore, I switched to ECC88, a more popular tube with a lower Rp. Used as a cathode follower, it has an output impedance of 78 ohm (almost one half of 6n14P) under the given configuration and a better output swing. The ECC 88 is suitable for low voltage applications (in the old days I've built an ECC88 pre-preamp using 24V B+ for my mc cartridges), the value of cathode resistor is relatively low, therefore, 150R of cathode resistor is very suitable for the application and resulted in a better headroom.
Below two pix shows a ECC88 buffer @90V with a cathode resistor of 75R + 75R. The output was taken from the lower 75R where it was loaded with a 1.5M video cable terminated with a 75R resistive load on the other end. I've tried increasing the cathode resistor value using a negative supply rail which resulted in better voltage swing but not lower output impedance or improved waveform. I've also tried using a digital trans. but the result was not satisfatory.First one, SPDIF signal locked
Second one, SPDIF working with 1Khz sine wave signal. This one looks very much like the "AFTER" pix in the PDF posted by lga775, albeit more symetrical.
In my last post the output waveform of the ECC 84, with no load applied, was respectable in terms of square wave performance, but not its symetrical swing. This was improved by the use of ECC88. However, when loaded with a 1.5M cable and terminated with a 75R resistor on the other end, the output of ECC88 ran out of gas and the square wave was rounder than it should be. My attenpt to put a digital trans. at the O/P was not satisfactory too. The reason for this, I believe, was because the output impedance of the tube was not low enough to drive the load. Although the ECC88 cathode follower has a respectablely low output impedance, it was using its maximum strength to drive a 75 ohm cable terminated with 75R resistor (a capacitive load (of the cable) + resistive one of the R), not to mention adding an inductive load (digi-trans.). Consider this situation, would someone be satisfied with the performance of a set up with an amplifier of 8 ohm output impedance driving a 8 ohm speaker? No. I would not. A much lower output impedance would be required to drive a 8 ohm speaker to satisfaction.What next? I need a tube buffer with a super low output impedance. The White Cathode Follower (WCF) immediately comes to mind. See this link:
http://www.tubecad.com/2006/10/09/cathode%20followers.png
So I built a WCF circuit using an ECC88. It has a gain of 0.97 and an output impedance of around 10 ohm. You can Google for info on WCF. The inherent uses of WCF were video buffer in the old days and headphone amp. Since it has a much higher voltage swing, a resistive network was used to bring the signal down to SPDIF level. The resistor network was connected to a digital trans. before output to the 1.5m video cable(terminated with 75R resistor).Below pix : WCF o/p SPDIF signal loack @96K
Below pix : WCF o/p working with 1Khz of sine wave.
Not bad in terms of square wave perforance, although rise time is marginally slower than the original signal. There was slight undershoots which may be improved upon refinement of the circuit.After some burn in, I will talk about the sound quality of the three configurations, i.e. Two tube configs (CF & WCF) and the original 74HCxxx output.
Edits: 06/26/10 06/26/10 06/26/10 06/26/10 06/26/10 06/26/10 06/26/10 06/26/10 06/26/10 06/26/10 06/26/10 06/26/10 06/26/10 06/26/10 06/26/10 06/26/10 06/26/10 06/26/10 06/26/10 06/26/10 06/26/10
Hi,
Looking at your 'scope traces and noting your comment on the SPDIF transformer I cannot but conclude that you are doing a lot of things rather wrong.
Above I posted a trace from my implementation, with ECC88, follower and SPDIF transformer. That is what it looks like if it is implemented well:
Ciao T
Sometimes I'd like to be the water
sometimes shallow, sometimes wild.
Born high in the mountains,
even the seas would be mine.
(Translated from the song "Aus der ferne" by City)
hi bro, good to hear that u have done the comparison of using different tubes with different circuit, may b u can also try 6n14p it is relatively cheap to try out.
Hmmm. did you hear any improvement of sound when u switch from CF to WCF?
Hi Jack
Great follow-up on the design by lga775. Hope your input will result in a better design for the tube buffer.
> The negative swing is clipping.The buffer distorted.
What is the significance of this distortion? Will it manifest in incoherence in the music?
> What next? As a loading of 150 ohm is extremely low for the tube which has a RP of 3670 ohm, the cathode resistor value should be increased....
Any suggestion to address this shortcoming?
Best Regards
Hi Jack! Could you show the oscillogram of standard spdif output without the tube buffer (stock version of juli, musiland etc.) to see advantages of the tube buffer.
Download finnaly successful. Thank you for unselfishly sharing your hardwork.
The file is successfully downloaded. Thanks!
Me cannot download. Can anyone email the file instead at smicyta@singnet.com.sg?
tia
i am currently busy with my academy exams... i've been struggling with this project for a month plus.. until jackwong96 and Dr.Lamp told me that has to bypass the multiplexer.
Post a Followup:
FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
This post is made possible by the generous support of people like you and our sponsors: