| Passive I/V vs. traditional I/V, FFT pics. - Click HERE for Original Thread |
| A 8 |
The following pics are taken with SpectraLab using a laptop on batteries and a slightly modified USB soundcard also run on batteries.
The passive I/V DAC is a balanced PCM1704K with a resistor (100ohms) between I out on the dacs, nothing xtra to ground.
The measurements for this dac are taken at the poweramp output.
The PCM1704 passive I/V is being compared to the Denon 2900 DVD using the BB PCM/DSD 1792 and standard I/V etc basically (different opamps) as described in TI's datasheets. These pics are directly from line out. The spec on this player says 0.0009% THD @ 1kHz.
First out is the Denon @-20dB.. |
|
|
| A 8 |
| PCM 1704 passive I/V @ -20dB |
|
|
| A 8 |
| The passive PCM1704 @ -60dB |
|
|
| A 8 |
| Denon @ -100dB (16 bit data) |
|
|
| A 8 |
| The passive PCM1704 dac @ -100dB (16bit data) |
|
|
| A 8 |
Considering my expectations based on the general consensus on this forum about passive I/V I think this is pretty good and way beyond my expectations.
The following pics are direct spdif dumps from the same source. |
|
|
| A 8 |
| -100dB which will be the last. |
|
|
| ojg |
Great measurements, certainly gave me something to think about.
It is interesting how the harmonics of the PCM1704 is very similiar to the measurements I did on my own dac.
Some questions:
- You say you used a USB soundcard, which one?
- From what I understand you run the balanced out from the DAC direct to the input of the power-amp. What does the power-amp input circuitry look like?
- Would it be possible for you to post the results of 0dBFS measurements? If it is too loud for your soundcard just use a resistor divider.
Thanks! |
|
|
| Pedja |
Good posts, A8.
Why do all -60dB plots have similarly looking hash? Is it embedded in signal? |
|
|
| Bernhard |
It shows that @ -60 dB a 20bit DAC is not better than the good old TDA1541.
Also harmonics are odd order which I find a bit critical... |
|
|
| jean-paul |
| quote: | | It shows that @ -60 dB a 20bit DAC is not better than the good old TDA1541. |
:D |
|
|
| Bernhard |
Or just bad implementation ?
My DENON DCD3300 is very noisy. Harmonics like a TDA1543 :cannotbe: |
|
|
| ojg |
| quote: | Originally posted by Bernhard
It shows that @ -60 dB a 20bit DAC is not better than the good old TDA1541. |
I think the -60dB test signal is screwed up. Look at the picture in post #10 which is the reference signal, does not look good... |
|
|
| A 8 |
Hi guys,
For those of you who can't resist getting into a ****ing contest I suggest you save your TDA1541/nonos vs everything else for another thread.
I did not expect the passive I/V to measure as good and figured some of you would be interested considering this could open for a "lean & mean" signalpath without sacrificing traditional THD performance.
The two balanced dac chips have a resistor between each I-out, I-out on each chip is then directly coupled to the balanced input on a LC Audio The End Amp with all the resistors in series and to ground removed before the amps input transistors (se pic below).
The -60 dB signal looks bad but the noise is at a low level and as OJG points out it does so on the Denon analog out as well as the direct spdif dump from the Denon which is being used as source for all measurements.
For now I am assuming it is not my DAC, but rather the denons digital processing or the HiFi News&Record review Test CD II I used. I am going to try with a real cdp once I get a chance. It is strange though that it does not show on any other level measurements
All in all I think the denon is far from noisy and measures pretty good.
At 0dB the soundcard distorts heavily on both the denon and the passive I/V dac but as you say I've been thinking about using a divider once I get time for it.
The soundcard is (lord behold) a creative Audigy 2NX with a 24/96 ADC.
The first tests I did showed it was very noisy and the SNR readings I got was not better then around -93dB above 1kHz. It improved significantly to -120,-122dB when I put it on a 5volt NiMh pack and added oscons on the adc and digital processing supplies.
The line input on the soundcard is directly coupled to the adc input ie no additional op-amps other then whatever sits in the adc chip.
Note that the passive I/V measurements are taken at the poweramp speaker terminals with a pair of 3-4 ohms speakers connected so some of the harmonics might be related to the poweramp.
The denons are taken directly at line out which uses NJM2068 (bal) for I/V and AD825 for buffers. |
|
|
| Bernhard |
| Would be interesting to compare left and right channel of the Denon... @ - 60dB |
|
|
| Brian Guralnick |
| For a passive solution, instead of just a resistor, why not use a DAC I/V transformer. It should do an even better job. Get that DAC to see a as close to 0 ohm impedance as possible... |
|
|
| janneman |
Brian,
I thought even for the DAC transformer you need a resistor on the DAC output to develop the input voltage for the transformer?
Jan Didden |
|
|
| carlosfm |
| quote: | Originally posted by Bernhard
It shows that @ -60 dB a 20bit DAC is not better than the good old TDA1541.
|
:D :D :D :devily: |
|
|
| Brian Guralnick |
| quote: | Originally posted by janneman
Brian,
I thought even for the DAC transformer you need a resistor on the DAC output to develop the input voltage for the transformer?
Jan Didden |
Correct, but, the a current out DAC really performs best with a 0 ohm load. Otherwise, use a voltage output DAC.
This is the point of the electronic IV stages trying to ge that magical 0 ohm load with enough gain & minimal noise.
Now, using a transformer & 5 ohm resistor, example setup, the load on the DAC with resistor may be 5 ohms & the output wil be a puny .1v, but on the other side with a gain of 10x, you will have your 1v rms.
The idea here is the transformer shrinks your 100 ohm DAC load, which is crummy for the DACs current switches making an un-optomized circuit, to 5 or less ohms, then it passively gives you the gain.
Here is my favorite IV stage transformer:
http://www.sowter.co.uk./specs/9762.htm
(how's that for a flat wideband transformer) |
|
|
| MBK |
I don't quite understand the engineering issues in a transformer based I/V.
The transformer you quote has a DCR of 6 Ohms. In that case could one not do without the I/V resistor altogether?
And: does the transformer really "shrink" the I/V resistor's value? Would the effective input impedance of a transformer not depend on the load impedance of the transformer at its secondaries? So, if you have this transformer with a 6.5:1 turns ratio and a load of 10k on the secondaries, would the effective input impedance not be 10000/6.5=1.5k? In that case, would it not rather look like you still do need a fairly small I/V resistor OR a small load? Or do I misunderstand this and the load recommended in this site, 100-200 Ohms, is to be placed at the secondaries only?
Example, no I/V resistor at DAC, DCR 6 Ohms, secondaries load of 39 Ohms, would you not get 39/6.5=6, + 6 Ohms DCR Ohms = 12 Ohms effective primaries impedance? That being close to the maximum load a typical current output can drive, if we leave it at that, we get 16 dB voltage gain, or 0.15 V output, which is quite OK indeed. With a 200 Ohms load on the secondaries and no load on the primaries, we'd have 30 Ohms on the primaries, may still be OK. So, assuming no resistor at the DAC side, I can see how this works. I just can't figure out how this works with a 100 Ohms load on the DAC and a highish load on the secondaries.
And - for unipolar current outputs, what to do with the DC through the transformer? Put a beefy electrolytic at the DAC ouput? or will the transformer tolerate the say, 2 mA of a TDA1541 (through 6 Ohms DCR that is just 0.024 mW, but what about saturation?) |
|
|
| Kuei Yang Wang |
Konnichiwa,
| quote: | Originally posted by MBK
I don't quite understand the engineering issues in a transformer based I/V. |
There is nothing to understand nor are there any issues. A transformer does not convert Current to Voltage.
| quote: | Originally posted by MBK
The transformer you quote has a DCR of 6 Ohms. In that case could one not do without the I/V resistor altogether? |
Nope.
| quote: | Originally posted by MBK
And: does the transformer really "shrink" the I/V resistor's value? |
Yes, a transformer transforms the impedance by the square it's winding ratio. And it will transform the current or voltage by the winding ratio (voltage up current down for stepup transformers).
In other terms, if you have a DAC with an Output current of +/-1.2mA for full scale and you apply this current to the primary of a 1:10 Transformer you will give +/- 120uA on the secondary. To get a 5.65V Peak Peak signal from this current you need around 23k5 as I/V resistor. This 23k5 will be transformed into the transformers primary by the square of the 1:10 winding ratio, namely 100. So the load reflected upon the DAC's current output is around 235 Ohm.
Clearly, an output impedance of 23KOhm is impractical, so if we follow the Transformer plus I/V resistor with an active stage we can reduce the I/V resistor.
Let's use a 6922 as Valve for the analogue stage. It has a gain of around 29db with a 22k Anode load and around 3K output impedance. For 2V RMS (5.65V P-P) from the Anode we need around 200mV P-P or 100mV Peak. So we require 100mV from 120uA or 833 Ohm. Let's use 1k for arguments sake, so we require a 1:10 transformer that can be loaded with 1k and this will then reflect an impedance of 10 Ohm plus all direct current resistive losses.
While looking good on paper this is not a very practical transformer to find of the shelf, so we are back to first converting our current to Voltage with a suitable resistor (say 10 Ohm) and then stepping it up with a traditional 1:10 Stepup Transformer (for microphones and or MC Pickups) loaded with it's usual secondary load (say 10k giving a 100R Load in parallel with the 10R I/V conversion resistor) to give use 100mV Peak from the DAC's output into our Valve stage.
| quote: | Originally posted by MBK
And - for unipolar current outputs, what to do with the DC through the transformer? |
This does not work easily for unipolar DAC's. You will need a suitable offset voltage in the "middle" of the linear range which has a zero AC impedance (or at least a fairly low and consistent AC impedance).
| quote: | Originally posted by MBK
or will the transformer tolerate the say, 2 mA of a TDA1541 (through 6 Ohms DCR that is just 0.024 mW |
If it is designed suitably it will be fine. I took a different take in my Adagio. The transformer used had a DCR of the primary of 180 Ohm and I used a 20 Ohm I/V resistor. This way only a very small offset went into the transformer, one I tested to have no material negative effects. If you have a low DCR primary transformer you can simply use a series resistor.
BTW, with the TDA1541 the best solution I found is no transformer and a rather tricky near "zero bias" implementation of the 7308/E88CC which uses a rather higher than usual value I/V conversion resistor (100R). Just one resistor, one RLC combo for compensating the sinc treble rolloff, one valve, one anode resistor and one output coupling capacitor.
Sayonara |
|
|
| MBK |
Thanks, makes sense. Some misunderstandings along the way but nevermind (I didn't assume that transformers convert current to voltage just by themselves, the "I/V resistor" needs to exist at some point - however the question was whether it needs to be on the primary).
The square relation of the windings is one of the points I had missed... must be of course because power in must correspond to power out (minus losses).
Looks like it could work with a fairly simple setup. The TDA1541 should lend itself to a Pass D1 style I/V as well, but that is more complex to get zero bias reliably.
I now have a JFET and op amp based solution, I wanted to give a transformer a try, also because of inherent RF rejection. |
|
|
| Brian Guralnick |
| quote: | Originally posted by MBK
\
I now have a JFET and op amp based solution, I wanted to give a transformer a try, also because of inherent RF rejection. |
DONT FORGET the voltage isolation!!!!!!!!
At times, it's unbelievable how useful this can be, especially if you want to convert unbalanced dacs to a true unbalanced output, or, feed a simple balanced line stage SOZ type of buffer running on the same psp without any DC reg caps at all in the audio path. |
|
|
| A 8 |
I am disturbed about the -60dB dumps.
I started a new series of tests playing with bit dephts and another data source ie the DAC Tester that was kindly provided by ULAS on the forum and a Sony CDP.
All pics are spdif direct.
First out is a SONY CDP 16/44.... |
|
|
| A 8 |
| This is the DAC Tester at -60dB 16/44 |
|
|
| A 8 |
| Now it gets interesting, this is the Sony CDP playing a -70dB Dithered 16/44. |
|
|
| A 8 |
| Compared to the DAC Tester with no dithering at -70dB 16/44 |
|
|
| A 8 |
| Last out is the dactester @-60dB 24 bits and 44kHz |
|
|
| Pedja |
A 8, if that may help, you can download both dithered and undithered, 0dB and -60dB sine signals (.wav files) from here and burn by yourself to CD. You have there all the files analyses as such, so you can know what you are working with.
Pedja |
|
|
| Bernhard |
| Could it be that compared to the -70dB signal, with -60dB there are more bitd switching in the DAC chip and those extra switching bits are bad by accidant in the Sony ? |
|
|
| A 8 |
Thx Pedja,| quote: | | A 8, if that may help, you can download both dithered and undithered, 0dB and -60dB sine signals (.wav files) from here and burn by yourself to CD. You have there all the files analyses as such, so you can know what you are working with. |
I tried to burn these files but the program states they are the wrong format!?
Any idea as to why I can not burn them?
Also, looking at your pics I see similar hash on the non dithered signals.
Could it be that the "hash" only shows when analyzing 16 bit non-dithered data on a 24 bit processing system? |
|
|
| Pedja |
| quote: | Originally posted by A 8
I tried to burn these files but the program states they are the wrong format!?
Any idea as to why I can not burn them? |
I did not encounter any problem. What program do you use to burn them?
| quote: | Also, looking at your pics I see similar hash on the non dithered signals.
Could it be that the "hash" only shows when analyzing 16 bit non-dithered data on a 24 bit processing system? |
That hash is the regular part of the conventionally generated signal. That’s why dither. The difference can be seen normally on 16 bit system.
Pedja |
|
|
| A 8 |
| quote: | | I did not encounter any problem. What program do you use to burn them? |
Its a HP burner and I use whatever came with it; HP CD writer powered by veritas.
I've checked the files on my Xp laptop and they are indeed 16/44 but only one channel...perhaps thats why. |
|
|
| Pedja |
| Yes, these are the mono files. Work normally with Nero so I did not expect it can be the problem. The brief check says EAC or even Windows Media Player 8 (XP) can burn them too. |
|
|
| A 8 |
Pedja,
I downloaded NERO and it worked...also seems the data source makes a difference.
OJG,
The pic below shows the passive balanced PCM1704K @ 0dB and a analog attenuator before the soundcard....still looks better then expected. |
|
|
| Bernhard |
| Did you compare left & right channel of your cdp ? |
|
|
| A 8 |
| quote: | | Did you compare left & right channel of your cdp ? |
Yes, they are identical between L&R channels irrespective of the source; Denon DVD 2900, Sony CDP or wav files direct from the computer. |
|
|
| ojg |
| quote: | Originally posted by A 8
PThe pic below shows the passive balanced PCM1704K @ 0dB and a analog attenuator before the soundcard....still looks better then expected. |
It is very good indeed! With your solution you have combined I/V, preamp and volume-control fully balanced in one single potentiometer, quite nifty! I assume the pot is about 1kohm but at normal listening levels it is probably not set at much more than 100ohms, right?
Too bad I did not realize this until Pedja's post, but it is obvious now that your original -60dB signal was un-dithered. |
|
|
| A 8 |
ojg,
That's about it but I use a stepped attenuator.
I had to test 24 bit performance and made a few dumps of -100dB, -110dB, -118dB and digital zero still at the poweramp (2*180w) output.
Pretty impressing as well.......
The small spikes @ 4.3, 14 and 18kHz seems to be from the computer/soundcard combo as they are identical if I just short the input on the soundcard.
First out 24/48@-100dB |
|
|
| A 8 |
| -118dB...I guess the TDA1541 would have a hard time matching this;) |
|
|
| A 8 |
And digital mute just to show I wasn't just lucky getting the -118 reading.
Also shows that spikes orginating from somewhere else but the dac/poweramp combo. |
|
|
| A 8 |
| BTW, it sounds really good too. |
|
|
| Bernhard |
| quote: | Originally posted by A 8
-118dB...I guess the TDA1541 would have a hard time matching this;) |
| quote: | Originally posted by A 8
For those of you who can't resist getting into a ****ing contest I suggest you save your TDA1541/nonos vs everything else for another thread. |
;)
Seems the distortion is hidden under the noise.
Also such low signal level doesn't exist on CD ? :confused:
Want a DAC shoot-out ? :D
I am seriously interested if more bits are better :att'n:
Please post new measurements @ -60dB with clean signal from Pedja.
I got two -60dB signals, one from Pedja and one from another forum member.
One had lower noise and I don't remember which one. :bawling:
This is one of my 1541 with os, don't remember which one:
Worst harmonic is K2 is -65dB |
|
|
| A 8 |
| quote: | | Want a DAC shoot-out ? |
Sorry about doing what I asked you not to....and -no I am not looking for 24 vs 16 bit dac shootout as a 16bit dac is not an option for me.
| quote: | | Also such low signal level doesn't exist on CD ? |
Right, however CD is not the only source availible to us. Concerts on dvd in 20-24/48 LPCM makes a big difference when run through a 24 bit dac vs a 16 bit one and can become very engaging.
DVD-Audio through this dac when mastered right is also VERY nice.
| quote: | | I am seriously interested if more bits are better |
Well at -60 dB there might not be a big difference but at -100dB and below there will be a significant difference..;) |
|
|
|
|
