DIY discrete dac chip.

Status
Not open for further replies.
Jump to Latest
#61
Absolute linearity

Hi again.

I've just been looking in the PCM1702 datasheet.

http://focus.ti.com/lit/ds/symlink/pcm1702.pdf

I'm not impressed by the documentation of the performance. They state that monoticity is archived but the test only shows monoticity for the 10 LSB.
Monoticity for the 10LSB should be very easy to archive, since the monotocity is best at low signal levels. I wonder how the monoticity is at +-½ fullscale were the 18LSB is shifted to 1 MSB ?

Or is it less importance ? I can imagine that a linearity error is gives a very digital sound. Because distortion of this kind isn't present in analog systems.

I think absolute linearity is of very high importance. And that Monoticity tells little about the perfomance, you can archive monoticity even though there is linearity error greater than several bits.

Any comments ?

To get a very good absolute linearity, temperature controlled DAC is the only way to go 😀
 
#63
Re: Re: Absolute linearity

Bernhard said:


I have tested PCM56 with and without adjustment.

Without it, it is very temperature stable, but with adjustment...
Click to expand...

With adjustment the DAC isn't temperature stable ? The distortion rises, more on an adjustet version than a DAC which isn't adjustet ?

It must be because the adjustment circuit dosn't have the same temperature coefficient as the internal resistors. Or is it something else ?

A DIY dac needs to be adjusted because buing ultra precise resistors, would be to high a cost. This means that temperature drift would nit be the same for the different bit. This indicates that a temp controlled DAC would be very nice.
 
#66
The potentiometer has to be of good quality. Unfortunately most brands used cheap carbon potentiometers of average quality. I am afraid you were measuring the quality of the potentiometer. Anyway, you may want to try it with a good type.

In the case of PCM63 BB stated themselves that omitting the potentiometer(s) was the best solution to achive long term stability.
 
#70
Cost of the DIY DAC

rfbrw said:
What is the projected unit cost of this marvel?
Click to expand...

The weighted current source DAC:

1 voltage reference http://www.maxim-ic.com/quick_view2.cfm/qv_pk/1657 :2,25$

8 low Vos Quad opamp http://www.analog.com/UploadedFiles/Data_Sheets/141253328AD8605_6_8_d.pdf :1,63$/pcs

32 jfet : 0,5$ /pcs

32 precision resistors http://www.elfa.se/pdf/60/06031066.pdf :0,1% 0,5$ pcs

16 dual SPDT fast switch http://pdfserv.maxim-ic.com/en/ds/MAX4635-MAX4636.pdf : 0,9$/pcs

Total about 54$ for a 16 bit stereo DAC.

Beside this comes powersupply, PCB, decoupling cap's, logic IC's, (low jitter clock). And probably more... 🙂
 
#77
unanswered questions

I had e few questions which I didn't get a answer to, so I'll try to post them again :

I really have no idea of how much intrisic jitter is added by a conventional dac itself. But the settling time is pretty high around 200ns for a PCM1704. I think this indicates that some jitter is added by the decive itself.
Does anyone have an idea of the jitter added by the DAC ?


How much current noise is there in a cheap 1% metalfilm resistor ?

PCM1704 documentation:
I'm not impressed by the documentation of the performance. They state that monoticity is archived but the test only shows monoticity for the 10 LSB.
Monoticity for the 10LSB should be very easy to archive, since the monotocity is best at low signal levels.
I wonder how the monoticity is at +-½ fullscale were the 18LSB is shifted to 1 MSB ?

Or is it less importance ? I can imagine that a linearity error is gives a very digital sound. Because distortion of this kind isn't present in analog systems.


A DAC can have monoticity even though the linearity isn't good.
e.g. a DAC can have a linearity error of 5LSB, with no error in monoticity.

Anonther dac can have monoticity errorss, but a max linearity error of e.g. 3LSB, which one of the DAC is the best ?

I would say the one with the least linearity error.
 
#78
Re: unanswered questions

Lgrau said:
I had e few questions which I didn't get a answer to, so I'll try to post them again :

I really have no idea of how much intrisic jitter is added by a conventional dac itself. But the settling time is pretty high around 200ns for a PCM1704. I think this indicates that some jitter is added by the decive itself.
Does anyone have an idea of the jitter added by the DAC ?
Click to expand...

No. However from memory, one of the lowest jitter DACS
Stereophile ever measured with the Julian Dunn "J test" was the
Sonic Frontiers Processor 3. This was with the transport and DAC
in linkback mode (ie; no PLL, directly reclocked) . All other PLL
based DACs have measured slightly worse.

What does this tell us? The SF P3 uses BB PCM1704 (similar to
1702) so my guess is that your jitter bottleneck is going to be
elsewhere. Even the Lavry DA2002 with it's discrete ladder and
XO driven sample + hold reclocking measured very similar at
around 125pS.

You might also compare the low level linearity plots of the SF P3 to
the discrete Lavry DA2002. The PCM1704 actually measures
slightly better.

How much current noise is there in a cheap 1% metalfilm resistor ?
Click to expand...

Current noise is inversly proportional to the size of the R.
So to make low noise current sources you need big R's NOT
little ones.

PCM1704 documentation:
I'm not impressed by the documentation of the performance. They state that monoticity is archived but the test only shows monoticity for the 10 LSB.
Monoticity for the 10LSB should be very easy to archive, since the monotocity is best at low signal levels.
I wonder how the monoticity is at +-½ fullscale were the 18LSB is shifted to 1 MSB ?

Or is it less importance ? I can imagine that a linearity error is gives a very digital sound. Because distortion of this kind isn't present in analog systems.

A DAC can have monoticity even though the linearity isn't good.
e.g. a DAC can have a linearity error of 5LSB, with no error in monoticity.

Anonther dac can have monoticity errorss, but a max linearity error of e.g. 3LSB, which one of the DAC is the best ?

I would say the one with the least linearity error.
Click to expand...

Are you going to use S + H at OP to get around glitch problem?

Cheers,

Terry
 
#79
Re: Re: unanswered questions

Terry Demol said:


No. However from memory, one of the lowest jitter DACS
Stereophile ever measured with the Julian Dunn "J test" was the
Sonic Frontiers Processor 3. This was with the transport and DAC
in linkback mode (ie; no PLL, directly reclocked) . All other PLL
based DACs have measured slightly worse.

What does this tell us? The SF P3 uses BB PCM1704 (similar to
1702) so my guess is that your jitter bottleneck is going to be
elsewhere. Even the Lavry DA2002 with it's discrete ladder and
XO driven sample + hold reclocking measured very similar at
around 125pS.

You might also compare the low level linearity plots of the SF P3 to
the discrete Lavry DA2002. The PCM1704 actually measures
slightly better.



Ok, the jitter is below 125pS in a PCM1704 then. I was afraid that the jitter was higher.
But still the jitter is high compared to jitter from e.g. Guido Tents clock.

I tought the Lavry DAC would be better.

I would expect the PCM1704 to measure better at the low level linearity, because of their Signed Magnitude design.
But is the low level linearity most important ? I think the overall linearity of the Lavry DAC is better.
IMHO linearity at -100db isn't important to the sonic performance.
But I could be mistaken here.


Current noise is inversly proportional to the size of the R.
So to make low noise current sources you need big R's NOT
little ones.


A higher reference voltage will then lead to less resistor current noise, if the output current is equal.



Are you going to use S + H at OP to get around glitch problem?
Cheers,

Terry


It should be possible to reach 16 bit precision without using S + H
I've not been Investigating how to make a "current s/h" circuit
Click to expand...
 
Status
Not open for further replies.
Top Bottom