The circuit on the right is useful when you need a current source to +ve supply. If you need a current source to the -ve supply then change the PNP transistors to NPN.
Current Sources, Sinks and Mirrors in Audio
Current Sources, Sinks and Mirrors in Audio
Thanks again abraxilato. Ive been scratching my head over your single transistor iv circuit..
does the transistor work like a transimpedance node(?) buffering the i/v resistor that follows?
And the tl circuit is to balance the offset of 1387?
Can i use bc560 instead? P2p will be much easier with to92
And how does it sound subjectively to you? I guess this is the hardest question but seeing as how youve prefered passive iv sound so far i doubt it would be something that sound 'stale' or 'hard'.
does the transistor work like a transimpedance node(?) buffering the i/v resistor that follows?
And the tl circuit is to balance the offset of 1387?
Can i use bc560 instead? P2p will be much easier with to92
And how does it sound subjectively to you? I guess this is the hardest question but seeing as how youve prefered passive iv sound so far i doubt it would be something that sound 'stale' or 'hard'.
Yes you could say that the single transistor is a transimpedance circuit - low input impedance and high output impedance. I suppose it could also be called a 'current conveyor' - it provides no current gain, the output current is the same as the input.
The output current from TDA1387 is always in a positive direction (out of the chip) so to get a bipolar audio signal we need some negative current to balance this to give an audio signal centred on zero. That's assuming the AD844 is being used as the transimpedance amplifier. With a single transistor stage the output is a single ended voltage so always above 0V if the I/V resistor goes to 0V - however it (the I/V R) can go to a -ve voltage reference to centre the signal around 0V.
TO92 transistors of course will work as well as SMT. No problem.
As regards how it sounds - well I have found quite a lot of the sound quality depends on the filtering beyond the I/V, so its hard to separate out the I/V aspects from the filter aspects. But to my ears the TDA1387's bass sounds deeper, more controlled when its given a low impedance active load for its outputs. My DACs that used passive I/V I found benefitted from HUGE amounts of capacitance (up to 1F) on the supplies to get the best bass - with active I/V this is no longer required.
The output current from TDA1387 is always in a positive direction (out of the chip) so to get a bipolar audio signal we need some negative current to balance this to give an audio signal centred on zero. That's assuming the AD844 is being used as the transimpedance amplifier. With a single transistor stage the output is a single ended voltage so always above 0V if the I/V resistor goes to 0V - however it (the I/V R) can go to a -ve voltage reference to centre the signal around 0V.
TO92 transistors of course will work as well as SMT. No problem.
As regards how it sounds - well I have found quite a lot of the sound quality depends on the filtering beyond the I/V, so its hard to separate out the I/V aspects from the filter aspects. But to my ears the TDA1387's bass sounds deeper, more controlled when its given a low impedance active load for its outputs. My DACs that used passive I/V I found benefitted from HUGE amounts of capacitance (up to 1F) on the supplies to get the best bass - with active I/V this is no longer required.
Ive been going over your posts in the 1387 thread and cannot find any details on the two transistor current sink for 1387 output, unless its part of the class a buffer you posted. If so, i have no idea how that would null the current offset at 1387 output and am over my head. ��
If there was no schematic for it, could i ask for the detail of the circuit at least?
If there was no schematic for it, could i ask for the detail of the circuit at least?
The voltage (reference for the base of the transimpedance transistor) was chosen to maximize the possible output voltage swing from the collector whilst still keeping the DAC's output within the allowed compliance voltage range. Its 3.8V when the supply of the TDA1387 is 6V (the maximum allowed). If you run the DAC chip at the more normal 5V then you should reduce the reference to 2.8V.
It doesn't really matter much how you generate the voltage, so long as its not too noisy. I think, with hindsight I should have included some RC filtering on the TL431 output to reduce its noise.
What you found with the AD844 mirrors what I've found. Ideally you'd use no resistor at all when the DAC drives the AD844, seeing as its a transimpedance amplifier. A resistor isn't needed.
The two transistor current sink you need if driving an opamp (or AD844) is an NPN version of the current source I linked to (post 21).
If vref is grounded that would be a zero ohm resistor - not recommended. Vref on the 1543 is a voltage source - by an appropriate R to ground you can choose the current in the pull-up current source. But it sounds best if you set the pull-up source to zero (no resistor at all) and arrange an external current source (or appropriate external Vref for the I/V resistors).
Its working, and i love it!!
Sounds no different from passive i/v, only stronger and more open i was worried the active would be invasive, changing the sound characteristics but this is just the ticket for me, so thank you very much.
going back to my old dac, the flawed ad844 setup, i do miss the bass tho. is there any way to improve your iv with minimum part count?
I used resistor v divider and have a b1 buffer right after the i/v
Sounds no different from passive i/v, only stronger and more open
i was worried the active would be invasive, changing the sound characteristics but this is just the ticket for me, so thank you very much. going back to my old dac, the flawed ad844 setup, i do miss the bass tho. is there any way to improve your iv with minimum part count?
I used resistor v divider and have a b1 buffer right after the i/v
The reason for the apparent lack of bass is probably that the single transistor doesn't present a very low impedance to the DAC. At 1mA (the maximum) its going to have an input impedance about 25ohms. How to improve this is to run it at higher bias - on my modded Taobao 1387 DAC I am running about 5mA extra bias through mine - this brings the input impedance down to 5ohms.
To do this mod you need two extra current sources per I/V transistor - so a total of 8 extra transistors to build the CCSs. Two of the CCSs are built out of PNPs and source current from the +ve rail, the other two are NPN versions and sink current to GND. The PNP CCSs feed into the I/V transistor's emitter (the DAC outputs) and the NPN CCSs sink current from the I/V transistor's collector (go in parallel with the I/V resistor).
To do this mod you need two extra current sources per I/V transistor - so a total of 8 extra transistors to build the CCSs. Two of the CCSs are built out of PNPs and source current from the +ve rail, the other two are NPN versions and sink current to GND. The PNP CCSs feed into the I/V transistor's emitter (the DAC outputs) and the NPN CCSs sink current from the I/V transistor's collector (go in parallel with the I/V resistor).
Thanks again richard. I compared your iv to zen and i find yours to be superior in many aspects. Still havent even installed ccs..
Does 1387 benefit from i2s attenuation much the same way previous tda chips do? i remember trying ecdesign's i2s attenuator with 1543 and hearing big improvements.
Does 1387 benefit from i2s attenuation much the same way previous tda chips do? i remember trying ecdesign's i2s attenuator with 1543 and hearing big improvements.
Last edited:
Thanks very much for the feedback, I really appreciate it. What is 'zen' - do you have a link to it so I can see a schematic?
As to I2S attenuation, its not something I've tried in isolation - I used to do it on the DAC-AHs I modified. I would swap TDA1543 for TDA1387 or TDA1545 and attenuate the I2S lines at the same time. Perhaps I should revisit it. One thing I did find made a small improvement a couple of years back was changing the BCK to run at 1.4MHz - in other words only 32X 44k1. I have a schematic on my blog for doing this although I think it may have a small error in it. Its on my list of to do things to revisit that and try it out again. Since I did that test I've made many changes to the analog parts of my DACs.
As to I2S attenuation, its not something I've tried in isolation - I used to do it on the DAC-AHs I modified. I would swap TDA1543 for TDA1387 or TDA1545 and attenuate the I2S lines at the same time. Perhaps I should revisit it. One thing I did find made a small improvement a couple of years back was changing the BCK to run at 1.4MHz - in other words only 32X 44k1. I have a schematic on my blog for doing this although I think it may have a small error in it. Its on my list of to do things to revisit that and try it out again. Since I did that test I've made many changes to the analog parts of my DACs.
The gratitude is mine
Its nelson pass' zen i/v. It has more romance compared to yours but loses out in dynamics and transparency. No doubt the coupling caps in that circuit contributes to a great degree.
I cant find ecdesign's i2s.circuit.that i uses but roughly remember it as 220ohm series resistors per pin grounded with several diodes to.give 500ma attenuation.
Id try your circuit but if its active it will be a while before i breadboard another.dac along it.
Do u find external clocks make much difference with tda chips?
Its nelson pass' zen i/v. It has more romance compared to yours but loses out in dynamics and transparency. No doubt the coupling caps in that circuit contributes to a great degree.
I cant find ecdesign's i2s.circuit.that i uses but roughly remember it as 220ohm series resistors per pin grounded with several diodes to.give 500ma attenuation.
Id try your circuit but if its active it will be a while before i breadboard another.dac along it.
Do u find external clocks make much difference with tda chips?
Clocks are one of the things I used to obsess about quite a lot before I starting building DACs. I'd pore over phase noise plots...
Then since I've built DACs using Philips multibit chips (not including TDA1541) I've never found any degradation with any clock I've used - whether that's a recovered one from a DIR9001 or WM8805 or a crystal in my QA550 or Taobao-sourced media player. So jitter is not something that's keeping me awake at night, not at all.
Then since I've built DACs using Philips multibit chips (not including TDA1541) I've never found any degradation with any clock I've used - whether that's a recovered one from a DIR9001 or WM8805 or a crystal in my QA550 or Taobao-sourced media player. So jitter is not something that's keeping me awake at night, not at all.
Does 1387 lock on with 192khz signal from dir9001? Im noticing that receivers have their own sound. It is distictively different going from 8412 to 8414, and now im waiting to try 9001.
Also how high of a gain can your i/v support? I did notice the sound changing from raising the resistor value, rather similar to the effect i was getting from ad844
Also how high of a gain can your i/v support? I did notice the sound changing from raising the resistor value, rather similar to the effect i was getting from ad844
I've not tried 1387 with 192kHz so I have no idea. I did try it at 88.2kHz (2X OS) but it sounded greyer than at 44.1kHz so I abandoned my oversampling experiments.
I'm not clear what you mean by 'how high gain' - do you mean what's the largest resistor value you can use? That's set by how large an amplitude can be swung at the I/V transistor's collector, which in turn depends on the reference voltage you set at its base. If you run the 1387 at 6V and have 3.8V at the base of the I/V transistor then you could maybe go up to 3k. 3k3 would definitely be too high.
I'm not clear what you mean by 'how high gain' - do you mean what's the largest resistor value you can use? That's set by how large an amplitude can be swung at the I/V transistor's collector, which in turn depends on the reference voltage you set at its base. If you run the 1387 at 6V and have 3.8V at the base of the I/V transistor then you could maybe go up to 3k. 3k3 would definitely be too high.
Its nice to know that someone as expert as you has the same preference that i do. I also find oversampling to be less vivid sounding than nos, so yours is a path that i can safely follow
Foobar has a new os algorithm thats suppose to be superior so i still want to try that.
Soon i will be breadboarding your passive lc filters.
How much of a compromise is your 1 inductor filter you posted on 1387 thread compared to your earlier ones?
Im on my way back from the parts shop where i already picked up 10mh inductors- theyre little vertical tube thru hole types coated in green ceramic, 10cents each.
Foobar has a new os algorithm thats suppose to be superior so i still want to try that.
Soon i will be breadboarding your passive lc filters.
How much of a compromise is your 1 inductor filter you posted on 1387 thread compared to your earlier ones?
Im on my way back from the parts shop where i already picked up 10mh inductors- theyre little vertical tube thru hole types coated in green ceramic, 10cents each.
Ha thanks for the compliment - I rather like the definition of 'expert' that goes like this 'someone who has made all the mistakes there are to be made in a particularly narrow field'. In which case I don't yet qualify
When I tried oversampling I was careful to use a minimum phase filter of my own design, not a standard off-the-shelf chip, as I didn't want to have the result skewed by implementation details of the standard filter chips. So I confidently predict that no 'improved algorithm' is going to make OS sound better than NOS - the greying effect is I believe due to glitching and settling errors from the DAC chip itself.
The single inductor filter isn't designed as a 'compromise' solution, rather its primarily designed to fix up the NOS droop, not to provide anti-imaging filtering. It can be used with any other filter or with none. Your 10mH inductors don't sound to me to be too promising - are you sure they're not 10uH ones in disguise? Check the DCR to be sure, there won't be any 10mH under 5ohms but 10uH will be lower than this.
When I tried oversampling I was careful to use a minimum phase filter of my own design, not a standard off-the-shelf chip, as I didn't want to have the result skewed by implementation details of the standard filter chips. So I confidently predict that no 'improved algorithm' is going to make OS sound better than NOS - the greying effect is I believe due to glitching and settling errors from the DAC chip itself.
The single inductor filter isn't designed as a 'compromise' solution, rather its primarily designed to fix up the NOS droop, not to provide anti-imaging filtering. It can be used with any other filter or with none. Your 10mH inductors don't sound to me to be too promising - are you sure they're not 10uH ones in disguise? Check the DCR to be sure, there won't be any 10mH under 5ohms but 10uH will be lower than this.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.