I am trying to understand how to correctly figure the value of a resistor for a purely resistive I/V stage for a current output DAC like the TPA COD.
The BB PCM1794a Dac chip has a 7.8ma p-p output per channel @ 0dfs, 15.6ma of you use the chip in mono mode. so it is easy enough to calculate E=IxR to figure a resistor/voltage but at some point the Dac runs out of voltage!
So, .0078 x say 100 ohms = 0.78v p-p
vs.
say .0078 x 500 ohms = 3.9v p-p
Where is the limit? if the analog side of the DAC chip is running on 5V would the limit be just under 2.5V??
And do you want to size the resistor for max voltage swing? or is there some point you try and aim for?
The datasheet only gives an example I/V calculation for an active I/V stage. nothing about passive I/V stage calc.
Zc
The BB PCM1794a Dac chip has a 7.8ma p-p output per channel @ 0dfs, 15.6ma of you use the chip in mono mode. so it is easy enough to calculate E=IxR to figure a resistor/voltage but at some point the Dac runs out of voltage!
So, .0078 x say 100 ohms = 0.78v p-p
vs.
say .0078 x 500 ohms = 3.9v p-p
Where is the limit? if the analog side of the DAC chip is running on 5V would the limit be just under 2.5V??
And do you want to size the resistor for max voltage swing? or is there some point you try and aim for?
The datasheet only gives an example I/V calculation for an active I/V stage. nothing about passive I/V stage calc.
Zc
Last edited:
The active stage puts a ZERO resistance (less than miliohms, depending of the OpAmp) on the output of the current source. That minimize the distortions and errors due to the finite internal resistance of the DAC current source.
A straight resistor will add distortions and noise. Electronics 101.
A straight resistor will add distortions and noise. Electronics 101.
The limit is set by the DAC's output compliance voltage spec. That's not shown on the 1792's datasheet so you'd have to determine it empirically. TI assumes you'll follow their recommendations (which suck for sound incidentally) and use an NE5534
Active stage is only better in some aspects - myself I've played with active stages but normally end up preferring the sound of passive.
@SoNic - the milliohms does not extend across the full bandwidth because its generated by the opamp's open loop gain. As that falls at 6dB/octave so the input resistance rises by the same beyond the dominant pole frequency of the opamp. Depending on the opamp various amounts of IMD will be produced because DACs have no difficulty in exceeding the slew limits of opamps. You'll see in various application notes the opamp is fitted with a feedback cap to try to mitigate this.
Intermodulation - Wikipedia, the free encyclopedia
Active stage is only better in some aspects - myself I've played with active stages but normally end up preferring the sound of passive.
@SoNic - the milliohms does not extend across the full bandwidth because its generated by the opamp's open loop gain. As that falls at 6dB/octave so the input resistance rises by the same beyond the dominant pole frequency of the opamp. Depending on the opamp various amounts of IMD will be produced because DACs have no difficulty in exceeding the slew limits of opamps. You'll see in various application notes the opamp is fitted with a feedback cap to try to mitigate this.
Intermodulation - Wikipedia, the free encyclopedia
Its probably quite difficult to find an opamp fast enough to give sound quality to beat a passive I/V. I reckon some of the National very fast parts might manage it. If you look at my blog you'll see I'm using AD603 as the buffer/amp stage for my passive I/V. This to my knowledge hasn't been used in a commercial product, but it is a very fast part with GBW in excess of 1GHz.
Hi,
the voltage limit would be as Abraxalito says the so called voltage compliance (roughly +-1v, depending on the type of transistors used for the DACs current sources)....if the outputs were not protected by diodes.
The BB/TI DACs feature protection diodes in their outputs which limit the useful voltage range to <<300mV. As soon as the diodes start conduction the THD figures of the DAC rise. I wouldn´t recommend more than 22Ohms IV-resistor value for the 1792/1794.
This translates to a voltage level that asks for some amplification.
The imho best sounding way to do the IV-conversion is to use a currrent conveyor or current buffer stage like the Jocko and similar. These stages isolate the DACs output and the IV-resistor, presenting the DACs output the low impedance it likes to see and transferring the DACs current into the resistor. The signal voltage develops over the resistor and may be buffered with an output buffer stage.
In its simplest form the IV+Buffer may be build out of just 2 transistors, but a practical form would take up 5 transistors (see the Jocko). The bandwidth and speed of these very simple circuits is amazingly high since they don´t work with global feedback. And we need alot of speed because the DACs feature a fullscale settling time of only 200ns.
A nice read http://www.diyaudio.com/forums/digital-source/195483-zen-cen-sen-evolution-minimalistic-iv-converter.html
jauu
Calvin
the voltage limit would be as Abraxalito says the so called voltage compliance (roughly +-1v, depending on the type of transistors used for the DACs current sources)....if the outputs were not protected by diodes.
The BB/TI DACs feature protection diodes in their outputs which limit the useful voltage range to <<300mV. As soon as the diodes start conduction the THD figures of the DAC rise. I wouldn´t recommend more than 22Ohms IV-resistor value for the 1792/1794.
This translates to a voltage level that asks for some amplification.
The imho best sounding way to do the IV-conversion is to use a currrent conveyor or current buffer stage like the Jocko and similar. These stages isolate the DACs output and the IV-resistor, presenting the DACs output the low impedance it likes to see and transferring the DACs current into the resistor. The signal voltage develops over the resistor and may be buffered with an output buffer stage.
In its simplest form the IV+Buffer may be build out of just 2 transistors, but a practical form would take up 5 transistors (see the Jocko). The bandwidth and speed of these very simple circuits is amazingly high since they don´t work with global feedback. And we need alot of speed because the DACs feature a fullscale settling time of only 200ns.
A nice read http://www.diyaudio.com/forums/digital-source/195483-zen-cen-sen-evolution-minimalistic-iv-converter.html
jauu
Calvin
Last edited:
@Calvin, That is interesting you say that 22 ohms should be the limit when Twisted Pear recommends 470 ohms and a major audio manufacturer that uses these chips in their products use 150 ohms when the chips are used in stereo and about 100 ohms when the chips are used in mono mode. So this seems to be a very cloudy issue with no clear answers. I am currently running my dac chips in mono mode with about 100 ohms with no issues. even when a 1khz 0dbfs cd is played there is no visible clipping on the scope. and distortion is very low.
Zc
Zc
It very much is a very cloudy issue - partly because TI doesn't tell us what the chip's limits are. Philips/NXP do say for their chips, but they set the AC compliance value extremely low - I think about +/-25mV (from memory). Plenty of NOS passive I/V designs with e.g. TDA1543 violate this spec by more than an order of magnitude without horrendous distortion. I use 47R on the output of my TDA1387s which more or less complies (ha) with the datasheet spec for those parts. I do though measure full scale distortion which is considerably poorer than the spec (around -66dB) - this might be more to do with the distortion in the buffer than from the DAC's compliance limits. I don't myself consider this amount of distortion to be a problem provided its low order and of course it decreases with reducing level.
OpAmps are not THAT bad as you want to make us belive. A decent 55MHz open loop BW and 20V/uS is healthy enough for the 20kHz limit of audio signal. The rest will be filtered by filtes - yes they are needed to achieve low distortion.
Sure, if you belive in fairies with tubes, NOS, no filters and such, nothing can convince you that TI, WM and AD are not trying to "get you" with their "bad" design.
If tubes where a remote alternative to quality OpAmps, they would be present in datasheets. There is no conspiration, just poor sound quality from a I/V based on resitor and tube voltage follower.
Sure, if you belive in fairies with tubes, NOS, no filters and such, nothing can convince you that TI, WM and AD are not trying to "get you" with their "bad" design.
If tubes where a remote alternative to quality OpAmps, they would be present in datasheets. There is no conspiration, just poor sound quality from a I/V based on resitor and tube voltage follower.
If truth be told, I'd prefer you believed nothing. Just try some experiments and work out what works for you.
Yep.
If you have a great RF filter then do please share. I'm working on one now for my NOS DACs and its not trivial. Any time you might be able to save me would be appreciated
an NE5534
What's all this about? TI make the DACs, why wouldn't you take their recommendation?
As far as an RF filter is concerned, a few 10s of pF in shunt will do fine. Of course you can put everything in a cage and use feedthrough caps if you really feel it's necessary.
Simply because I'm aiming for a different result from TI. They sell according to the numbers in the datasheet, I make my DACs pleasant to listen to.
In fact I started off with their recommendation in my Asus Essence ST soundcard - presumably because Asus's engineers thought like you. However I found it sucked sound-wise and haven't looked back.
I find your bravado here oddly unconvincing.
Hi,
after our measurements the excellent THD-Values of the PCM1794 are kept only with small I/V resistor values. In fact the measured values were slightly better than the datasheet values. Now if You accept a slight increase in THD figures, larger values of R may be used and may be sensitive if it easens Your design. The probability is high that a buffer stage following the I/V-resistor increases THD values by a more significant amount. In praxis there´s also to keep in mind, that a 0dB fullscale signal rather seldomly occurs. If You design for -3dB or -6dB as typical maximum signal than You can safely double the resistor values. With a simple current conveyor circuit (Jocko et al)You omit the problem of low resistance values by buffering the DAC from the I/V resistor and You are quite free in choosing a sufficiently large R-I/V value. A small cap in parallel to the I/V-resistor may be used as bandwidth limiting filter (aka anti-aliasing). No other filtering may be needed.
The signal voltage generating from the I/V-resistor may then be buffered to the analog output.
The advantage of the IV-resistor and the current conveyor is their inherent speed. If the DAC output settles in 200ns fullscale the input of the following stage needs to be able to follow these steep flanks. A resistor or a common-base/common gate-stage can do this. A OPamp with a closed loop (not openloop!) bandwidth (or rather Gain-Bandwidth-product) of 55MHz may not be fast enough. Rather look for video-OPamps or slow down the DAC with some capacitance to allow for slower OPs. Even fast OPamps feature OL-bandwidth values of less than 1k. The falling OL-response means less correction by feedback action above the OL-bandwidth limit, hence increasing THD with rising frequency. Another side effect is the non-constant input impedance. The input impedance rises with increasing frequency and can reach unacceptably high values. The DACs output is clocked at a couple of MHz. Input impedance needs to be constantly low also at these high frequencies. Its to understand, that we are rather not dealing with audio at the DACs output, but a HF-signal. So what applies to audio design, doesn´t apply here in first instance.
I understand App-notes and Datasheets as a service of the manufacturers to their customers. This does not necessarily mean that those circuits are the non-plus-ultra solution but rather a cost effective solution and an oversight over how things may be done.
That chip manufacturers nearly fully rely on circuits with chips seems logical to me. I wouldn´t expect to see a tube stage from TI or ADI since this is not their field of expertise and means a mix of different manufacturing technologies for the resulting device.
jauu
Calvin
after our measurements the excellent THD-Values of the PCM1794 are kept only with small I/V resistor values. In fact the measured values were slightly better than the datasheet values. Now if You accept a slight increase in THD figures, larger values of R may be used and may be sensitive if it easens Your design. The probability is high that a buffer stage following the I/V-resistor increases THD values by a more significant amount. In praxis there´s also to keep in mind, that a 0dB fullscale signal rather seldomly occurs. If You design for -3dB or -6dB as typical maximum signal than You can safely double the resistor values. With a simple current conveyor circuit (Jocko et al)You omit the problem of low resistance values by buffering the DAC from the I/V resistor and You are quite free in choosing a sufficiently large R-I/V value. A small cap in parallel to the I/V-resistor may be used as bandwidth limiting filter (aka anti-aliasing). No other filtering may be needed.
The signal voltage generating from the I/V-resistor may then be buffered to the analog output.
The advantage of the IV-resistor and the current conveyor is their inherent speed. If the DAC output settles in 200ns fullscale the input of the following stage needs to be able to follow these steep flanks. A resistor or a common-base/common gate-stage can do this. A OPamp with a closed loop (not openloop!) bandwidth (or rather Gain-Bandwidth-product) of 55MHz may not be fast enough. Rather look for video-OPamps or slow down the DAC with some capacitance to allow for slower OPs. Even fast OPamps feature OL-bandwidth values of less than 1k. The falling OL-response means less correction by feedback action above the OL-bandwidth limit, hence increasing THD with rising frequency. Another side effect is the non-constant input impedance. The input impedance rises with increasing frequency and can reach unacceptably high values. The DACs output is clocked at a couple of MHz. Input impedance needs to be constantly low also at these high frequencies. Its to understand, that we are rather not dealing with audio at the DACs output, but a HF-signal. So what applies to audio design, doesn´t apply here in first instance.
I understand App-notes and Datasheets as a service of the manufacturers to their customers. This does not necessarily mean that those circuits are the non-plus-ultra solution but rather a cost effective solution and an oversight over how things may be done.
That chip manufacturers nearly fully rely on circuits with chips seems logical to me. I wouldn´t expect to see a tube stage from TI or ADI since this is not their field of expertise and means a mix of different manufacturing technologies for the resulting device.
jauu
Calvin
You need to capture the settling time of 200ns? For what, you want to hear 5MHz?
Even so, how is 55 MHz BW not enough for those 5MHz?
The filters that follow an OS design are at some 40-50kHz. You have 1000 time more BW - that's all you need.
Measurements prove that 114dB THD is achieved with OpAmps. I would like to see a tube that can get close of that. The rest is just... urban legends.
Except here the people that belive in NOS and cry that they don't have "good filters".
Even so, how is 55 MHz BW not enough for those 5MHz?
The filters that follow an OS design are at some 40-50kHz. You have 1000 time more BW - that's all you need.
Measurements prove that 114dB THD is achieved with OpAmps. I would like to see a tube that can get close of that. The rest is just... urban legends.
Except here the people that belive in NOS and cry that they don't have "good filters".
Last edited:
Right. You want the best sound regardless of cost but TI are only concerned with keeping costs down.
That still begs the question, why choose their DAC?
You think that their DAC is the best available, but they don't know how to take best advantage of its performance? That doesn't seem likely.
You think that their DAC is good value for money, but their judgement slipped when it comes to recommending an opamp? Nah, good value for money doesn't really chime with best sound regardless of cost, and anyway, chips is chips. You'd think that a regime that can evaluate a DAC could evaluate an opamp.
I mean, you can't build a DAC if you can't evaluate it.
OK, you got me, what is your philosophy here? Your hearing is better than everybody at TI put together when it comes to opamps but they still make good DACs, but that's a fluke? You're stretching my capacity to believe here.
Yes, but then you're beset with doubts and uncertainties on all sides, aren't you.
If you take a relaxed view of these problems you will find their solution easier. It's worth remembering that the control of RF ingress has been solved many times in the past in environments much more demanding than any audio DAC is likely to inhabit.
Perhaps if you tell us exactly how the problem manifests itself we could assist with some counselling. All problems are emotionally based in the end. After all, if you didn't care, it wouldn't be a problem.
What gave you that impression?
Are you asking me or telling me? I'd say no, they want the best figures within reasonable costs. That's what I've already said, but you weren't listening?
I chose their DAC back in the days when I considered figures to be the best guide to the sound of a DAC. I realize now that was far too simplistic a view to take.
Nor is that likely to me. Try again as both of us are unimpressed.
Irrespective of what I think, its a fairly good set of figures for the money.
Building on your original premise but you've yet to substantiate that. So you really need to put more work into shoring up your original premise.
Already stated.
Yeah mine too. Must mean you took a wrong turn somewhere don't ya think?
Interesting claim - any grounds for it?
Examples please? In particular control of RF where the desired SNRs are better than 127dB.
What problem are you referring to here? The problems arising from your unsubstantiated assumptions?
You lost me.
Alright while this has been entertaining we are getting wildly off track!
Obviously there are strong opponents for various ways for doing things.
THIS thread is NOT about how to use an opamp or an active I/V stage in a DAC design. you can create your own thread to discuss such a topic if you choose. just don't do it here please.
So lets stay on topic and discuss all things related to choosing resistors for a passive I/V stage in DAC designs.
We started discussing using low value resistors for lowest distortion. this sounds interesting. I think i need to dig out some R-boxes and experiment with Resistor values.
Obviously there are strong opponents for various ways for doing things.
THIS thread is NOT about how to use an opamp or an active I/V stage in a DAC design. you can create your own thread to discuss such a topic if you choose. just don't do it here please.
So lets stay on topic and discuss all things related to choosing resistors for a passive I/V stage in DAC designs.
We started discussing using low value resistors for lowest distortion. this sounds interesting. I think i need to dig out some R-boxes and experiment with Resistor values.
OK so now we're back on topic I'll mention a couple of other points about resistor values in passive I/V stages.
Firstly, DACs have code-dependent output resistance. This means they're not perfect current sources (which would have infinite output impedance) - rather they have a finite output resistance which is not constant. It depends on the binary value sent to the DAC. Having a lower value of I/V resistor means this effect is minimized.
A second point is related - there's also an output capacitance a DAC exhibits, again its generally not a constant. A lower I/V resistor value pushes the time constant of this capacitance further up in frequency resulting in lower distortion.
Firstly, DACs have code-dependent output resistance. This means they're not perfect current sources (which would have infinite output impedance) - rather they have a finite output resistance which is not constant. It depends on the binary value sent to the DAC. Having a lower value of I/V resistor means this effect is minimized.
A second point is related - there's also an output capacitance a DAC exhibits, again its generally not a constant. A lower I/V resistor value pushes the time constant of this capacitance further up in frequency resulting in lower distortion.
I did reply EXACTLY to the topic.
The only "calculation" that is to do to a passive I/V stage is to put the lowest resistance possible. There is no other "magical" calculation or resistance number that will work better. Any increase in resistence there increases the distortion.
A passive resistor will degrade the noise by diminishing the output voltage level.
Good luck looking for unicorns.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.