Using the AD844 as an I/V

Joe, it just occured to me:
Could be a good tweak to bypass V+ to V- directly, so the HF portion of the current doesn´t go through ground.
Path would be: bypass cap / chip V+ / out + / LPF filter cap / out - / chip V- / bypass cap, am I right?

Same would apply to I/V converters in case of current out.

In tda1543 it´s the opposite: I yesterday removed the MKP 2.2nF to ground and got a BIG improvement (passive out)! Love it.
 
Joe, it just occured to me:
Could be a good tweak to bypass V+ to V- directly, so the HF portion of the current doesn´t go through ground.
Path would be: bypass cap / chip V+ / out + / LPF filter cap / out - / chip V- / bypass cap, am I right?

Same would apply to I/V converters in case of current out.

In tda1543 it´s the opposite: I yesterday removed the MKP 2.2nF to ground and got a BIG improvement (passive out)! Love it.

I am trying to picture what you have done. Any chance you could sketch something out. That way I could surmise what the filter does, if that is what you have done, and may comment on how to optimise it. You help me and I will help me. Even sketch it out on paper and everybody can take a digital photo, and then post it here.

Cheers, Joe
 
Hi guys,
About bypassing, why no use inductors ahead of the chain instead?

The use of supply rail inductors is not uncommon. However, one has to be careful in designing the resultant LC filter. Being a resonant circuit, it must be properly dampened. Usually, the inductors for such application have ferrite bead cores, which does provide self-dampening. Even so, ferrite bead inductors are effective at very high frequencies, but not of much use at audio frequencies. Inductors large enough to effectively filter at audio frequencies are quite cumbersome. Particularly so, on an otherwise SMT PCB. At audio frequencies, it's much more practical to simply bypass filter utilizing relatively large amounts of shunt capacitance, than to LC filter.

One other factor influencing this design choice is that bypass filtering is as much, or more, about effectively providing an low impedance A.C. current supply loop very close to each load, than it is about filtering noise generated away from that load. Shunt capacitance meets this particular need. Series inductance does not.
 
Hi Joe,
What I have done, in the standard tda1543 passive circuit, is to remove a LPF cap that was in parallel with the I/V resistor. I confirmed what Richard (Abraxalito) had hinted: removing the cap improves the sound.

Anyway my experience is limited to the tda1543. The rest is theoretical


I am trying to picture what you have done. Any chance you could sketch something out. That way I could surmise what the filter does, if that is what you have done, and may comment on how to optimise it. You help me and I will help me. Even sketch it out on paper and everybody can take a digital photo, and then post it here.

Cheers, Joe
 
Hi Joe,
What I have done, in the standard tda1543 passive circuit, is to remove a LPF cap that was in parallel with the I/V resistor. I confirmed what Richard (Abraxalito) had hinted: removing the cap improves the sound.

It seems a curious thing, but what you say confirms it, that delta-sigma DACs can be improved with a post-DAC filter and ladder DACs without it or to a less degree. Mind you, the output of TDA1543 simply might not like driving the capacitance. I have a NOS-DAC here and the post-DAC filtering is a QuadFilar 1:1 transformer, and it works very well - the only downside is that it needs a series cap to block the DC from the tx, but it also gives galvanic isolation and is 100% fully balanced outs. In this case, no 844 needed. But I have multiple 1543 dumping current into a resistor and the output is then adjusted by varying that single R. But using a single 1543, then yes, use 844.

Cheers, Joe
 
Last edited:
I've found that if I use a common-base connected transistor on the output of an I-out DAC, its almost certain to oscillate if that transistor's output (collector) goes to a capacitor. Not sure why this is but I'll always feed it to an inductor first, problem solved. The I/V resistor's in circuit prior to the inductor of course.
 
I've found that if I use a common-base connected transistor on the output of an I-out DAC, its almost certain to oscillate if that transistor's output (collector) goes to a capacitor. Not sure why this is but I'll always feed it to an inductor first, problem solved. The I/V resistor's in circuit prior to the inductor of course.

WRT oscillation of GBS, I've never had this problem. And I've built a few of them.

T
 
I've found that if I use a common-base connected transistor on the output of an I-out DAC, its almost certain to oscillate if that transistor's output (collector) goes to a capacitor. Not sure why this is but I'll always feed it to an inductor first, problem solved. The I/V resistor's in circuit prior to the inductor of course.

could be combination of poor layout + unfiltered rubbish coming out the DAC hitting the emitter.

Nick
 
You want the subjective or the 'objective' answer of what its doing? I reckon its reducing the high frequency supply noise which is generated by the classB (output stage generally) circuits switching.

I don't recommend a 47uF - its very unlikely indeed to be of a high enough voltage rating. Get yourself some TDK 1206 10uF/35V (or 50V) X5R or X7R caps and start paralleling. The 1206 case size is about the sweetspot for bang/buck. Bigger cases are considerably more expensive than paralleled 1206s for similar capacitances and 1206s give lower ESR.
 
Just did your suggestion abraxalito, and put them on the AD844's and also on the BUF03's, and yes it seems a touch cleaner/sweeter and maybe blacker background, worth doing I think. To the ear it's not a massive difference, but detectable.
Without doing a scope scan of before and after I have no idea if the noise figures changed.

I used these

TDK Ceramic Capacitor 10uF 50V X7R 2220 C5750X7R1H106M 10MFD Rohs NEW 5pcs | eBay

Cheers George
 
Wow, expensive caps. Once you have one or more of these soldered in its worth putting a modicum of inductance on the supplies feeding them - perhaps a low impedance (power, not signal) ferrite bead.

How many are you using? I found that a single cap doesn't make too much difference but a dozen or more is much more obvious in sweetening the top. However at $1.79 a pop this suggestion quickly gets rather dear.... I use the 1206 types which nowadays here cost about $12/1k.
 
I'm just finishing up this circuit (see attached) and notice that this thread started when George was looking at the PCM1704 from the datasheet circuit. In terms of swapping out the OPA627 (IV gain) and the OPA2134 (LPF) - would the AD844 actually be a decent choice, or are serious modifications required first?
 

Attachments

  • pcm1704_sch.png
    pcm1704_sch.png
    54 KB · Views: 527
Left channel only done, using the 844's own buffer for the output. Don't forget to stack 2 for the PCM1704, read previous posts for stacking the iv stage only not the output buffer. You may have to use the dc offset nulling circuit on pins 1 and 8, depending on how close to zero dc offset your 1704's are.

Cheers George
 

Attachments

  • pcm1704_sch.jpg
    pcm1704_sch.jpg
    224.9 KB · Views: 525
Last edited: