Yeah, I know, I know 
. And I use web search engines. And I know there are good and expensive OpAmps.
Still, why, for example, NE5322 is often despised ? Even though one can find positive opinions on it with some engineering background, like THD, spectrum of harmonics, driving ability.
And even more interesting, Texas Instruments in http://www.ti.com/lit/ds/symlink/pcm1798.pdf recommends the venerable NE5534 and not something fancy-schmancy.
So, what are the engineering style opinions ?
For example, I use NE5532 in my acoustic measurements gear, but ironically in it even 1% THD won't matter - other errors are much bigger. OTOH, NE5532 behaves quite decently.
So, what to pay attention to first ? Open loop gain at 20KHz, for example ? Slew rate ? Ability to drive capacitive load ? Often people do not protect OpAmp outputs from capacitive load by simple resistors.
. And I use web search engines. And I know there are good and expensive OpAmps.Still, why, for example, NE5322 is often despised ? Even though one can find positive opinions on it with some engineering background, like THD, spectrum of harmonics, driving ability.
And even more interesting, Texas Instruments in http://www.ti.com/lit/ds/symlink/pcm1798.pdf recommends the venerable NE5534 and not something fancy-schmancy.
So, what are the engineering style opinions ?
For example, I use NE5532 in my acoustic measurements gear, but ironically in it even 1% THD won't matter - other errors are much bigger. OTOH, NE5532 behaves quite decently.
So, what to pay attention to first ? Open loop gain at 20KHz, for example ? Slew rate ? Ability to drive capacitive load ? Often people do not protect OpAmp outputs from capacitive load by simple resistors.
If you look at the date of the PCM1798 - 2006 - you'll realize that at the time the NE5534 was probably the lowest distortion ('affordable') device for this application in TI's stable. Now there are other options - OPA1611/2, OPA1642, as well as the various National Semiconductor audio op-amps. Generally, such a schematic and/or demo/eval board is designed to give the customer a setup that will match the datasheet specs. One shouldn't infer that the chosen op-amp 'sounds best' or is the best possible part for the job. For example, mating an Analog Devices op-amp to the PCM1798 might give you better objective or subjective performance, but generally you won't see a TI demo board with other brand semiconductors; ditto for ADI's (or anyone else's) boards.
NE5532/4 is perfectly serviceable for audio duty, however DAC I/V is a special case, being more video-like than audio. 5534 achieves the numbers for TI's datasheet (noise, distortion) but doesn't fare too well in terms of SQ due, I think to the high RF environment. For improved sound, I suggest a part which doesn't use an undegenerated LTP for input stage. GBW and slew rate are important factors beyond that.
How about "natively" low impedance circuitry ?
For a different application I have a self-designed "OpAmp" whose input stage is common base amplifier. And it is an I -> V converter - it has negative feedback and resistor in the feedback is the I -> V conversion coefficient.
Because of relatively low open loop gain and local negative feedback the thing is stable without any capacitors.
I also think that because of relatively low open loop gain and local negative feedback the thing is much less prone to temporary internal overload -> IMD. I mean overload due to large step input signal at input.
Of course, I am not talking about OpAmp in DC sense of the word, but since this is audio, we do not care about DC anyway.
...
On THD in general. I doubt that "well behaved" THD is that important. For example, if music is played through speakers, the resulting THD is much higher than, say, THD of NE5532. I.e. I think what is important is IMD appearing at the moments of quick signal transitions at input.
By the way, if I'm correct regarding the IMD, OpAmps with local negative feedback implemented as emitter resistors in the first stage can be better.
Could you please share your low-impedance i/v?
Common base is the way to go, prefferably with a nice behaviour without deep feedback (better transient response).
Take a look on ad844 internals, there is a nice current mirror on input with an output at TZ pin. This opamp is quite good for i/v conversion, but it gets distorted on larger >2ma currents. Input impedance is marginally high too.
Making it with transistors won't help much due to matching and temperature tracking problems.
Common base is the way to go, prefferably with a nice behaviour without deep feedback (better transient response).
Take a look on ad844 internals, there is a nice current mirror on input with an output at TZ pin. This opamp is quite good for i/v conversion, but it gets distorted on larger >2ma currents. Input impedance is marginally high too.
Making it with transistors won't help much due to matching and temperature tracking problems.
I'm not sure of your meaning here.
Sounds to me rather like a CFB opamp - IME a good choice for I/V duty. The inverting input is low impedance by local transistor action rather than through depending on the opamp's output.
That's my impression too, based on a very limited dataset
I'm explaining.
Suppose you have an I -> V converter implemented using a classical book-perfect OpAmp.
A perfect OpAmp has infinite input impedance on both inputs. Zero input impedance on the inverting input in I -> V converter is the result of negative feedback and infinite open gain loop - I am still talking about about a perfect OpAmp.
So, in real world, because open loop gain is limited, especially at high frequencies, a fast transition on the inverting input of the I -> V converter first sees high impedance - because the overal negative feedback hasn't yet kicked in.
Because of this two things happen:
1) OpAmp input stage oveloads;
2) the inverting input voltage "jumps" thus somehow modulating the DAC circuitry, i.e. potentials on switches inside DAC jump.
Then negative feedback kicks in and things return to normal.
So, saying "native low impedance circuitry" I mean circuitry having low impedance because of physical reality of the devices involved and not because of overall negative feedback. Such circuitry is common base/gate/grid one.
By the way, in microphones they sometimes use connected in parallel common base transistors in order to further reduce input impedance. This is used with ribbon microphones - for obvious ribbon low impedance reasons.
IIRC, this approach was also used for moving coil amplifiers in vinyl LP era. Jellybean 2N4401/2N4403 transistors are actually pretty low noise - even according to their official specs.
If I understand you correctly, you are talking about OpAmp DC signal characteristics.
If yes, why do we care about them in case of audio DAC ?
I think the OpAmp should be good enough not to electrically screw up the DAC itself, i.e. an OpAmp with 10mA input current would be bad, but with 0.1mA input current is probably OK.
Common base moving coil amplifier: Moving Coil Cartridge Head Amps .
I'd say it's pretty elegant. With the floating power supply there is no input DC current into/from the source.
I'd say it's pretty elegant. With the floating power supply there is no input DC current into/from the source.
if you like the leach, check out EUVL's SEN or CEN from this forum, similar concept ie a current conveyer with floating PSUs; batteries in this case are recommended
An interesting thread - thanks.
And the http://www.diyaudio.com/forums/diyaudio-com-articles/172983-zen-i-v-converter.html article is quite decently written. And PCM63 appears to be a great non-oversampling DAC. I as an engineer trust such a DAC much more than any oversampling DAC.
I'd still go for BJTs and some overall negative feedback. It's easier to design with BJTs
Anyway, has anybody tried such common gate/base circuits here ? Has anybody performed measurements on them ?
One has to remember that when such approached are taken, power supply cleaning (if it's not batteries) is the DIYer's burden.
no problem, if you cant choose a decent dac, I may as well help you choose a decent IV stage =) joking... Nelson, who is something of a patron saint of this place, sure does know how to write an article.
his pcm63 dac, the D1 (A commercial unit at one point) used a common gate IV, and I actually use a variant of this IV without the buffer, which was deemed the NTD1 (New take on the pass D1) I love it!. if you search that you'll find a development thread and a build thread started by user opc. the development thread is long and a bit meandering, but you'll find a couple of banks of comprehensive measurements with an AP system with various mosfets and supply voltages. Such exhaustive measurements are pretty much a hallmark of opc's projects, they are all very well characterised
the design in that thread and PCB used a pretty standard and accessible regulated supply based on 3 pin regs to more easily allow the higher swing, but weve successfully driven it with shunts, it gets pretty toasty though; kind of needs a case designed for a small power amp for heatsinking just for the IV without shunts
We use it with the ESS ES9018, which you may or may not know of yet, its possible to use it NOS, but its not really recommended, as it is a 32 bit DS OS dac. We get some extreme numbers out of such a simple IV using the highest transconductance fet and dropping the resistors right down by increasing the voltage. it can be adapted to numerous dacs, but in the NTD1 format is tuned for the ESS to get -113dB THD+N or thereabouts. there are 2 lots of measurements at least in that thread, the latter is the most detailed and discusses which one was chosen and why for the final form
I prefer fets in general, but actually there was a bipolar version of the CEN I mentioned earlier posted in one of the threads, however it didnt get much love.
just a tip though, you cant swing a cat without hitting an engineer around here, moreso than most audio forums; so maybe pick something else to differentiate yourself with i'm not, just an avid diyer, but there are all types here, easier to learn to live with them all.
i'm off to bed, if you have any problem finding the references for the designs let me know i'll chase up some links in the afternoon (AEST)
his pcm63 dac, the D1 (A commercial unit at one point) used a common gate IV, and I actually use a variant of this IV without the buffer, which was deemed the NTD1 (New take on the pass D1) I love it!. if you search that you'll find a development thread and a build thread started by user opc. the development thread is long and a bit meandering, but you'll find a couple of banks of comprehensive measurements with an AP system with various mosfets and supply voltages. Such exhaustive measurements are pretty much a hallmark of opc's projects, they are all very well characterised
the design in that thread and PCB used a pretty standard and accessible regulated supply based on 3 pin regs to more easily allow the higher swing, but weve successfully driven it with shunts, it gets pretty toasty though; kind of needs a case designed for a small power amp for heatsinking just for the IV without shunts
We use it with the ESS ES9018, which you may or may not know of yet, its possible to use it NOS, but its not really recommended, as it is a 32 bit DS OS dac. We get some extreme numbers out of such a simple IV using the highest transconductance fet and dropping the resistors right down by increasing the voltage. it can be adapted to numerous dacs, but in the NTD1 format is tuned for the ESS to get -113dB THD+N or thereabouts. there are 2 lots of measurements at least in that thread, the latter is the most detailed and discusses which one was chosen and why for the final form
I prefer fets in general, but actually there was a bipolar version of the CEN I mentioned earlier posted in one of the threads, however it didnt get much love.
just a tip though, you cant swing a cat without hitting an engineer around here, moreso than most audio forums; so maybe pick something else to differentiate yourself with
i'm not, just an avid diyer, but there are all types here, easier to learn to live with them all.i'm off to bed, if you have any problem finding the references for the designs let me know i'll chase up some links in the afternoon (AEST)
Last edited:
no problem, if you cant choose a decent dac, I may as well help you choose a decent IV stage =) joking... Nelson, who is something of a patron saint of this place, sure does know how to write an article.
his pcm63 dac, the D1 (A commercial unit at one point) used a common gate IV, and I actually use a variant of this IV without the buffer, which was deemed the NTD1 (New take on the pass D1) I love it!. if you search that you'll find a development thread and a build thread started by user opc. the development thread is long and a bit meandering, but you'll find a couple of banks of comprehensive measurements with an AP system with various mosfets and supply voltages. Such exhaustive measurements are pretty much a hallmark of opc's projects, they are all very well characterised
the design in that thread and PCB used a pretty standard and accessible regulated supply based on 3 pin regs to more easily allow the higher swing, but weve successfully driven it with shunts, it gets pretty toasty though; kind of needs a case designed for a small power amp for heatsinking just for the IV without shunts
We use it with the ESS ES9018, which you may or may not know of yet, its possible to use it NOS, but its not really recommended, as it is a 32 bit DS OS dac. We get some extreme numbers out of such a simple IV using the highest transconductance fet and dropping the resistors right down by increasing the voltage. it can be adapted to numerous dacs, but in the NTD1 format is tuned for the ESS to get -113dB THD+N or thereabouts. there are 2 lots of measurements at least in that thread, the latter is the most detailed and discusses which one was chosen and why for the final form
I prefer fets in general, but actually there was a bipolar version of the CEN I mentioned earlier posted in one of the threads, however it didnt get much love.
just a tip though, you cant swing a cat without hitting an engineer around here, moreso than most audio forums; so maybe pick something else to differentiate yourself with
i'm off to bed, if you have any problem finding the references for the designs let me know i'll chase up some links in the afternoon (AEST)
Thanks for the tips.
Regarding measurements - is this: http://www.diyaudio.com/forums/digi...ssic-pass-labs-d1-ess-dac-41.html#post2716414 post meant ?
What equipment was used for measurements ? A really good (like, say,. Lynx) soundcard ?
Regarding the article. Saying that it's nicely written I mean that I as engineer feel a lot of sense in it. For example, if increase of OpAmp supply voltages improves sound, and there was no visible overload of the OpAmp with lower voltages, it is quite possibly an indication that some parts of the OpAmp enter cutoff (I mean current) mode thus producing distortions for some time - until negative feedback kicks in.
I also think I am/was thinking about more or less the same phenomena the author was thinking.
...
Regarding BJTs. I think it's easier (at least, for me) to design with them circuits which are less sensitive the the active elements (i.e. BJTs) parameters variation. Conceptually BJT contains (kind of) voltage reference - I mean the typical 0.64V base emitter open BJT voltage, and IMO this "voltage reference" helps a lot. For a lot of BJT circuits I simply do calculations in my head - I'm talking about DC stuff.
I am not sure about noise in this application. My overall impression is that in current driven circuitry (which common base/gate circuitry is) BJTs might have better noise. Or, rather, I should put it differently: FETs definitely have better noise when source impedance is big and we are talking about common emitter/source circuit.
In this case we have to compare what ? Current noise of BJT base multiplied by BETA vs voltage noise of FET gate multiplied by S ?
I actually take this seriously. I.e. the joke is good
.Anyway, regarding choices. I think I've realized I have two needs:
1) something decent which is significantly better than my laptop on-bord DAC and which has 24 bits (probably true 20 bit DAC would be OK) and which just works. This is because the goal of it is to further advance my headphones related DSP project;
2) something "for my soul". I.e. something of artisan nature with better controllable by me electronics, something an electronic engineer can get pleasure from by tweaking it
.I started this thread meaning to talk about ready-made OpAmps. The DAC I think I'll buy comes with OPA2134. Which appears to be quite decent.
OTOH, the board I think I'll buy is USB powered, so there should be 5V -> -5V converter (the OpAmp requires IIRC 10V between rails). Since it will be powered by minimum allowed rail to rail voltage there is not much headroom in it. And USB +5V is not clean - that's why in another thread I asked about USB +5V regeneration.
Maybe I'll become tempted to improve the board by creating as much as I can clean +/- power lines out of USB +5V. Something like +5V -> +/-12V -> transistor filter in each rail and separate +5V -> +10V -> linear zener + RC filters + transistor voltage regulator for the DAC. All this power supply stuff will be implemented on a separate board which will be connected to the DAC one.
OK, reading the forum threads I stumbled upon http://www.linearaudio.net/userfiles/file/Didden LA Vol 2 EUVL.pdf .
The "matched Idss" is not for me. I am spoiled by working in the industry.
And I'm glad I could read about the measurements.
The "matched Idss" is not for me. I am spoiled by working in the industry
.And I'm glad I could read about the measurements.
I also found this: http://www.diyaudio.com/forums/digi...-minimalistic-iv-converter-5.html#post2694418 post.
And this: http://www.diyaudio.com/forums/digi...-minimalistic-iv-converter-7.html#post2698468 .
And from this: http://www.diyaudio.com/forums/digi...-minimalistic-iv-converter-8.html#post2699650 post:
I would add another stage and "global" negative feedback, so the distortions should become lower.
And this: http://www.diyaudio.com/forums/digi...-minimalistic-iv-converter-7.html#post2698468 .
And from this: http://www.diyaudio.com/forums/digi...-minimalistic-iv-converter-8.html#post2699650 post:
I would add another stage and "global" negative feedback, so the distortions should become lower.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.