Hi,
The Signal/Noise ratio and/or dynamic range of the PCM1704 are around 120dB A-Weighted, typical. That is the same as the PCM63 and well below a true "20-Bit" resolution level. So I would not worry about loosing the last 4 bit. Given the SNR of any available recording microphone you are not going to find anything but noise there anyway.
It cannot be the same. The output currents differ widely! Also, the PCM1702/1704 does not really passive I/V conversion, the PCM63 does (another reason to use it).
Well, it is a matter of comparison. If you gave me a choice between CS4398 or PCM1704 I'd also take the latter. But if you give me Ad1862, PCM63 or a TDA1541 with a few added bits I'd not look at the PCM1704.
Ciao T
The Signal/Noise ratio and/or dynamic range of the PCM1704 are around 120dB A-Weighted, typical. That is the same as the PCM63 and well below a true "20-Bit" resolution level. So I would not worry about loosing the last 4 bit. Given the SNR of any available recording microphone you are not going to find anything but noise there anyway.
It cannot be the same. The output currents differ widely! Also, the PCM1702/1704 does not really passive I/V conversion, the PCM63 does (another reason to use it).
Well, it is a matter of comparison. If you gave me a choice between CS4398 or PCM1704 I'd also take the latter. But if you give me Ad1862, PCM63 or a TDA1541 with a few added bits I'd not look at the PCM1704.
Ciao T
About 1 of 100 pcm63 might be 20bit . About 0 pcm1704. Why philips never manufactured a 20bit multibit dac? For the same reason the 1541 is unusable for correctly dithered 16/44, the DNL spec is just lame.
What we were talking about is the fact that the PCM1704 doesn't like any impedance for a load. It has been shown to start to sound good below 25 ohms with 10 ohms ideal (Swenson). The idea is that if you paralell DAC chips you can use a lower I/V resistor but what we don't know is if this will help the sound because as you say the voltage seen on the output of the DAC chip is the same.
I personally don't think it will work or someone would have reported it as the cure to passive I/V + PCM1704. I think the best solution is a high quality 10 ohms I/V resistor and a low noise high high gain phono tupe output stage. Something I've been meaning to try but haven't had the time.
The PCM63K sounds fine with a 47 ohm I/v resistor and it gives twice the voltage so end result is you only need 1/4 to ~1/8 the gain on your tube stage as the PCM1704. Thats the big reason people prefer the sound of PCM63 vs PCM1704. IME the PCM1704 matches the PCM63K in SQ with a <10 ohms load which usually means an active I/V.
Paralleling PCM1704 in this case will not help because the voltage variation across the 10R/n passive resistor at the current output pin is the same, which is what causes distortion. Paralleling DAC does reduce noise by square root of n.
At least that is the theory.
Patrick
At least that is the theory.
Patrick
I want to use my Lundahl LL1931 (1:16) trafos as passive I/V.
Currently I'm using it with AD1865 balanced DAC (1xAD1865 per channel, feeded with true and negavtive data) but destination will be PCM1704 in balanced mode.
I read somewhere that those 10ohms is a result of max acceptable output voltage swing of 25mV. It is diffirence because if I use more than 1 PCM1704 and I use 10ohm load, every chip will see 10ohm but voltage swing will be n times higher.
So I need to realize what is important to DAC to get max sound performance- just load or maybe allowed voltage swing?
Currently I'm using it with AD1865 balanced DAC (1xAD1865 per channel, feeded with true and negavtive data) but destination will be PCM1704 in balanced mode.
I read somewhere that those 10ohms is a result of max acceptable output voltage swing of 25mV. It is diffirence because if I use more than 1 PCM1704 and I use 10ohm load, every chip will see 10ohm but voltage swing will be n times higher.
So I need to realize what is important to DAC to get max sound performance- just load or maybe allowed voltage swing?
I think Evel just answered the question.
So we need to design a quiet DAC stage with ~100x gain. I vote 6EW7 SET OTL.
Hi,
Yes, PCM1704 is mostly active I/V only. The problem is that even the best open loop active I/U converters seem to give up something to the simple passive approach. So the PCM63 sounds better because it allows a better sounding analog stage.
As for those who wish to make a high gain, low noise gainstage - I think a D3a is the only game in town and even then you will struggle hugely.
Here is one case where using a low value resistor with the right kind of stepup transformer brings dividends.
If we are boring and use a E180F as gainstage (or C3g, WE 417A etc) we have a gain of around 40, at fairly high Zout. An output transformer with a really good 4:1 stepdown would be a good choice, probably a parallel-feed, superpermalloy cored variant (funny, I just recently ordered some such up for a linestage project). We have remaining a gain of around 10.
If use the Cinemag CMQ-3440 MC Stepup we have a gain of 36, so our resistor can be around 6.8 Ohm. The effective impedance at the tube grid is now 8.8KOhm, high enough that tubes own noise should be removed from the equation.
So what remains is the noise of our DAC and I/U resistors. At 6.8 Ohm we get a Johnson noise of -146dBV.
Our signal is 1.2mA * 6.8 Ohm = 8mV peak or 5.7mV RMS or -44dBV.
So we have a S/N ratio of -102dB remaining after we have used an essentially noiseless Tubestage and a low value I/U conversion resistor.
Hence we have made a 17 Bit DAC at the output of the analog stage, despite all this effort.
Might as well stick to the AD1865 if you want tubes and > 16 Bit resolution. At least the AD1865 tolerates passive I/V very well and we can at least get the full dynamic range it offers (with 220R I/V we can get 119dB SND at the I/U conversion with 150mV out opening up a wide range of tubes to use...
The PCM63 is even better due to +/-2mA Iout, as is the (16 Bit only, sadly) TDA1541.
Ciao T
Yes, PCM1704 is mostly active I/V only. The problem is that even the best open loop active I/U converters seem to give up something to the simple passive approach. So the PCM63 sounds better because it allows a better sounding analog stage.
As for those who wish to make a high gain, low noise gainstage - I think a D3a is the only game in town and even then you will struggle hugely.
Here is one case where using a low value resistor with the right kind of stepup transformer brings dividends.
If we are boring and use a E180F as gainstage (or C3g, WE 417A etc) we have a gain of around 40, at fairly high Zout. An output transformer with a really good 4:1 stepdown would be a good choice, probably a parallel-feed, superpermalloy cored variant (funny, I just recently ordered some such up for a linestage project). We have remaining a gain of around 10.
If use the Cinemag CMQ-3440 MC Stepup we have a gain of 36, so our resistor can be around 6.8 Ohm. The effective impedance at the tube grid is now 8.8KOhm, high enough that tubes own noise should be removed from the equation.
So what remains is the noise of our DAC and I/U resistors. At 6.8 Ohm we get a Johnson noise of -146dBV.
Our signal is 1.2mA * 6.8 Ohm = 8mV peak or 5.7mV RMS or -44dBV.
So we have a S/N ratio of -102dB remaining after we have used an essentially noiseless Tubestage and a low value I/U conversion resistor.
Hence we have made a 17 Bit DAC at the output of the analog stage, despite all this effort.
Might as well stick to the AD1865 if you want tubes and > 16 Bit resolution. At least the AD1865 tolerates passive I/V very well and we can at least get the full dynamic range it offers (with 220R I/V we can get 119dB SND at the I/U conversion with 150mV out opening up a wide range of tubes to use...
The PCM63 is even better due to +/-2mA Iout, as is the (16 Bit only, sadly) TDA1541.
Ciao T
Look at the 6EW7 datasheet, it can be setup as two anode follower stages (single tube) gain I believe 100x, The nice this is the output section has a low Rp (800 ohm), It would be a simple solution with CCS feed to each section OTL, just don't know how noisy it would be.
I tried the cinemag mic transformers and they basically sucked, sounded hollow/echoey just plan terrible.
Hi,
on't need to. Know the tube. So, what is the input noise of the triode section? Lower than a 6.8 Ohm resistor, which on it's reduces our S/N ratio to 102dB?
Not sure why. I have come across them several times in phonostages and they are not really bad.
Ciao T
on't need to. Know the tube. So, what is the input noise of the triode section? Lower than a 6.8 Ohm resistor, which on it's reduces our S/N ratio to 102dB?
Not sure why. I have come across them several times in phonostages and they are not really bad.
Ciao T
Here is a PCM1704 internal. This schematic come from ANJ1092
We can see that Iout must be kept as near as possible to zero volt. When you have chance to deal with +/-1.2ma use it! It's a very strange idea to use PCM1704 with output tension!
Attachments
Last edited:
Agreed but I don't think it is possible to build a 24 bit (144 S/N) analog stage and even less so with tubes. We will be lucky to have 90dB (15 bit) with tubes. I can't find an equivalent noise resistance for the 6EW7 but since the first stage is lower gain than the D3a (70 ohm eq noise res) I assumed that it may give an overall less noise? Plus there are other issues with D3a.
Hi,
True. Even 20Bit equivalence with tubes is going to be VERY non-trivial.
I recently designed a Phonostage with Ein -146dBV. I had to go hybrid to go anywhere near that.
Well, I can do a fair bit better than that... 😉
Well, noise resistance is generally the reverse of the transconductance, so more transconductance = less noise. The 6EW7 first stage has a transconductance of 2mA/V, the D3a has 35mA/V transconductance and around 150 Ohm equivalent noise resistance in reality (70 Ohm per milkmaid calculation) or -133dBV, which is around 5dB noisier than a 2SK170 for reference and 13dB noisier than my Phonostage... 😛
I would estimate the noise resistance of the 6EW7 first stage (low current/highergain) as 1.25KOhm (or -124dBV Ein) in theory, in reality probably around 3dB worse. So with a signal level at 10mV (-40dBV) you end up with a SN ratio of maybe 80dB, so 13Bit equivalent.
Of course, this is all conjecture and best guesses, rather than certainty, so by all means go ahead.
The D3a at least gets us into 16 Bit Territory with this.
If we use TDA1541 or PCM 63 with optimum I/U conversion the D3a can get us -113dB SN ratio, so we are at least around 18.5 to 19 Bits.
I'd personally probably try a E810F, with it's 50mA/V transconductance and 110 Ohm Ein. On second thought, maybe just a 6N30 cascoded with a bunch of 2SK170/LSK170?
There are? I used it in Phonostages, Mike Pre's and SE Amplifiers as driver, I did not note any.
Ciao T
True. Even 20Bit equivalence with tubes is going to be VERY non-trivial.
I recently designed a Phonostage with Ein -146dBV. I had to go hybrid to go anywhere near that.
Well, I can do a fair bit better than that... 😉
Well, noise resistance is generally the reverse of the transconductance, so more transconductance = less noise. The 6EW7 first stage has a transconductance of 2mA/V, the D3a has 35mA/V transconductance and around 150 Ohm equivalent noise resistance in reality (70 Ohm per milkmaid calculation) or -133dBV, which is around 5dB noisier than a 2SK170 for reference and 13dB noisier than my Phonostage... 😛
I would estimate the noise resistance of the 6EW7 first stage (low current/highergain) as 1.25KOhm (or -124dBV Ein) in theory, in reality probably around 3dB worse. So with a signal level at 10mV (-40dBV) you end up with a SN ratio of maybe 80dB, so 13Bit equivalent.
Of course, this is all conjecture and best guesses, rather than certainty, so by all means go ahead.
The D3a at least gets us into 16 Bit Territory with this.
If we use TDA1541 or PCM 63 with optimum I/U conversion the D3a can get us -113dB SN ratio, so we are at least around 18.5 to 19 Bits.
I'd personally probably try a E810F, with it's 50mA/V transconductance and 110 Ohm Ein. On second thought, maybe just a 6N30 cascoded with a bunch of 2SK170/LSK170?
There are? I used it in Phonostages, Mike Pre's and SE Amplifiers as driver, I did not note any.
Ciao T
ThorstenL thanks for the info, this is an area I've always wanted to explore but never had the time. Your phonostage sounds like quite the performer.
The more I think about it without your 149 dB S/N analog stage this project will just be an exercise in poor tube dithering. Probably better to resample/dither the 24/96 files with Izotope Ozone down to RBCD and use a TDA1541 DAC.
As far as the cinimags go, I'll revisite it. I add good luck with them on phonostages but with an R2R DAC (PSM63K) they just plain didn't work. Sounded like an AM radia. Better to get the Sowter transformers designed for the BB DAC's I'm afraid.
As far as the cinimags go, I'll revisite it. I add good luck with them on phonostages but with an R2R DAC (PSM63K) they just plain didn't work. Sounded like an AM radia. Better to get the Sowter transformers designed for the BB DAC's I'm afraid.
Hi,
Well, the stage is a MC Preamp and the noise is -146dBV. So with passive I/V on the PCM1704 and 10mV full scale signal even this stage would only give 106dB SNR.
Bottom line - we need a DAC with better tolerance to passive I/V.
Or maybe resample on the fly in the player using libsamplrate build into the player (as I do it).
This may have been down to the the actual application. Do you remember the exact schematic you used.
Ciao T
Well, the stage is a MC Preamp and the noise is -146dBV. So with passive I/V on the PCM1704 and 10mV full scale signal even this stage would only give 106dB SNR.
Bottom line - we need a DAC with better tolerance to passive I/V.
Or maybe resample on the fly in the player using libsamplrate build into the player (as I do it).
This may have been down to the the actual application. Do you remember the exact schematic you used.
Ciao T
There aren't any unless you throw SQ out the window and go S-D.
How about a simple SS I/V without gain then your phono gain stages?
That may prove interesting.
Hi,
I don't know. The AD1865 is 18 Bit with a decent SNR. We could use 4pcs and use these as effectively two extra bits (turn them on and off as the MSB's). Then we have a 20 Bit DAC that does all we want.
Why, if you use a SS I/U conversion (current conveyor) you may as well go straight to line level. I'd use one of the BB Diamond Transistors for the I/U conversion. As we obviously have to have a tube somewhere, why not as output follower?
Ciao T
I don't know. The AD1865 is 18 Bit with a decent SNR. We could use 4pcs and use these as effectively two extra bits (turn them on and off as the MSB's). Then we have a 20 Bit DAC that does all we want.
Why, if you use a SS I/U conversion (current conveyor) you may as well go straight to line level. I'd use one of the BB Diamond Transistors for the I/U conversion. As we obviously have to have a tube somewhere, why not as output follower?
Ciao T
To me there is no point in adding a cathode follower after a SS I/V.
I've heard of a China DAC that just used a single transistor collector resistor cicuit to convert the low i-out to a low v-out then had a tube gain stage. IME if we want tubes we want an anode follower gain stage.
- Status
- Not open for further replies.