advertisement


Misc dac output and filtering questions

K

Kit Taylor

Well-Known Member
My sistem suddenly sounding dreadfully thin and raw has prompted some more investigations into my dac project.

I've put a little low pass filter after the coupling cap, 100R 1206 in series and 10nf 1206 NPO to ground. the sound is softer and more delicate and doesn't sound like it's in the room with me. Unusual but not unpleasant.

Would the filter be theoretically better before the coupling cap?

I also want to try a single AD817 as I/V and buffer. The TDA1543 has a Vref resistor that connects to ground, does this need some kind of link to the AD817? What's the miniumum wattage of the feedback resistor?

Also, if I don't use the offset nulling should the two pins that serve this function be grounded?

TIA

Here's dac I'm talking about:

http://pinkfishmedia.net/forum/showthread.php?s=&threadid=4869
 
Couple of questions: did the sound go south after you fitted the filter, and what's the input impedance of the preamp the dac is driving?

10nF is a heck of a load for a passive output, about 70-100 x typical cable capacitance - and that causes enough problems on it's own!. 10nF in parallel with the input impedance of the preamp sets the roll-off point; if the pre is a Naim, usu. c.22K, that means the signal should be -3dB around 800Hz, which I'd expect to sound *extremely* dull and muted. I'd also expect the dynamics of a Trabant in treacle.

If I've guessed correctly so far, I surmise your dac is struggling and you're only hearing an approximation of 'normal' service because of the way the dac is reacting to the heavy capacitative load (it's a current-output dac, no ?)

Much better to buffer the dac output and implement the roll-off in the feedback loop of the buffer as an active low-pass. If you need an I-V stage, do the two things separately; I-V, then a buffer/lowpass output stage to follow.
 
Martin

The dac caught a cold before the filter. I think what happened is that the Vref had become desoldered by heat conducted through the foil groundplane. Soldering the I/V and Vref resistors to a piece of foil onto top of the TDA1543 (yep, current output) has set things right. I think one of the X7R caps I used was under voltage too.

I've ditched the analog filter because it sounds fine with out it, even if there might be some technical merit in noise reduction. I've also got my hands on some gruesomely expensive Stax electrostatic headphones. With a single full range driver it seems worthwhile keeping "phase distortion" to a minimum.

What puzzles me is how to setup the TDA1543 and AD817. My main intent is in going active is to reduce the stress on the TDA1543 caused by the I/V resistors.

I could go for a small value I/V with lots of gain in the AD817, but won't this reduce the dynamic range of the dac?

Or I could ditched the I/V resistors, but the how do I "match" the AD817 to the Vref on the TDA1543?

If you can help, loadsa thanks in advance.
 
Ho-hum, seems like my 1543 is permanently crippled. The bass is flappy like an out of control and it's noisy, one channel being quite a bit noisier than t'other giving the game away. Hope I can grab some more 1543 :^\

Still welcome any of your thoughts on opamps.
 
Hi

If you have the datasheet for the TDA1543, then the circuit for an op-amp IV is on the block diagram resulting in labelled Vouts for each channel. They also have low pass on & it can be done all in one go.
Basically the Vref goes to the noninverting input of the op-amp & the current output goes to the inverting input & then through the IV resistor to the op-amp output.

Ed
 
Thanks Ed. It seems that diagram isn't in all the dataheets floating around on the web. It's in this one though http://www.ampslab.com/PDF/tda1543.pdf

Does a lot of current flow through an opamp feedback resistor, will 0.125w do?

The distortion seems linked to the CS8412 analog power supply, and is worse with an indutor than without. Hmm.
 
You get the full output current of the DAC through the resistor - thats just 2.6mA (max) though so no problem for 0.125W resistor.

I had a quick look at some of your notes - it the spec at the top of your thread current? I think your Iref resistor might be the wrong value. I looked at the calculations & I think you have forgotten the -1.15mA for 'zero' midrange DAC output. This would give a 1k1 Iref resistor giving 3.3V midrange.

Having said that, you'll need to recalculate this for using the op-amps. What supply do you intend to give the op-amps, because there may be a headroom issue (with for example 5V supply & 1k2 IV resistor)? Those op-amps only go to about 1.5V of the rails according to the datasheet.

If you let me know what supplies you have (I guess the single 6V?) then I can work out a few values that should get it working.

I'm just having a look at the CS8412 datasheet now...
 
Well spotted Ed. I get 1K2 (1.192...).

I think the distortion problem came from a toasted regulator. A replacement part seems to have got it licked.

The dac is fed by a 12V DC source and I don't need a lot of gain as the passive output is plenty loud. I suppose +6v would be a sensible starting point.

Thanks again.
 
Glad you found the problem!

If you use the op-amp I/V then the DAC itself should be fine with 5V supply - it only needs a higher supply if doing the I/V direct. The op-amp will need a greater supply if you want the same voltage swing as you have now on the DAC. I would give it quite a bit more so you have lots of headroom & no chance for clipping. Also do your calculations taking the maximum DAC output current output into account - as they are now it could drive outside the voltage swing.

From an electronics view the op-amp is a better approach, though whether it sounds better or not ...?
 
Vref 1k2 is wrong as it sounds distorted.

880R sounds bright, punchy and very well defined spatially, with the impressive effect of making instruments like they're right there in the room.

1k1 is softer, smother and much bigger, less present and more diffuse. More of an expensive "boutique" type of sound, soothing but not boring. Is there some clever distortion and compression at work here, or is it really more relaxed sounding? Hmm.
 
According to what I can glean from the datasheet, you are almost certainly driving the DAC output outside it's specified range.
With the 880R you drive it under; with the 1K2, over & it appears to distort each in different ways. With the 880R you are asking it to swing voltage peak wise higher than the supply rail :(
I ralise this is the way other people seem to have used it & I haven't tried the same myself, but it doesn't look right.

The datasheet doesn't give that much info, but the important spec. is the output voltage compliance. Using the 'passive' resistor, the current driven through it generates the voltage signal at the output pin. This must stay within the given compliance (1.8V to supply-1.5V) for the device to operate correctly. The only confusing thing in the datasheet here is the quoted 'AC compliance' of +/-25mV. TBH I can't for the life of me figure out exactly what this refers to, though the implication is that the voltage on the pin should be held at a DC level. This is what happens with the op-amp circuit.

I would definately try the op-amp circuit. Just remember that the op-amp inputs will stay at the Vref voltage, and the DAC output current will all go through the feedback resistor (more of a feed-forward in this case) thus giving the oputput voltage swing at the other end of the resistor. Select resistor values (& op-amp supply) to keep this within the op-amp's voltage range & all should be well.
I predict you will get the 'airy' sound, which is probably the most 'correct'. Possibly not what you are after!

HTH!
 
Ed

Did you mean -1.5V or -1.2V?

I think the TDA1543 has abysmal distortion and noise figures even when setup correctly, so any resistor that doesn't sound flat out broken can only alter the balance of existing colorations anyway.

Forgive me for being thick (not good with math), but what resistors do you recommend for +5V on the AD817? It'll have to be that for the time being, as in single supply operation that's all the AD817 accepts.
 
Ah yes - it is 1.2V, & I was wrong about the 880R trying to go beyond the supply :rolleyes: so no-ones being thick - it's easy to make a mistake!!

Firstly, don't be fooled by the AD817 datasheet regarding the supplies - the chip can take any supply voltage from 5V to 30V (effectively +/-15) They only quote dual/single supplies due to convention.

Looking at the TDA datasheet diagram, the feedback resistor sets the output voltage span (2.3mA * Resistance) & the Rbias sets the offset (bit more involved!).
You just need to choose a suitable span to fit within the output swing of the op-amp.

For example, with a 6V supply, a 2V span should be safe, so say a 910R feedback resistor. With no Rbias at all, the midpoint Vout should sit at 2.2V + (1.15mA * 910R) = 3.25V. Near enough midscale to be safe. A 16K Rbias would get that closer to the 3V ideal. Things are a bit tight there, though.

With an 8V supply it would be much happier.
You could use a 1K3 feedback giving 3V output swing with plenty of headroom to avoid clipping.
18K Rbias would trim to midpoint to 4v though this shouldn't really be required (leave the pin open).

Adding the capacitor in parallel with the feedback resistor filters & prevents oscillation (possible due to fast op-amp). Cutoff freq. is 1 / (2 PI R C). The value in the datasheet should be a good start.

Vout then goes to your blocking (coupling) cap.
 
Tanks Ed. Not quite sure what this means

"With an 8V supply it would be much happier.
You could use a 1K3 feedback giving 3V output swing with plenty of headroom to avoid clipping.
18K Rbias would trim to midpoint to 4v though this shouldn't really be required (leave the pin open"

Leave Rbias out altogether or just don't hook it to the AD817? should AD817 the negative input then be grounded?

TIA
 
I'll try & explain the operation (as I understand it)

The Vref pin serves two functions - it gives a fixed 2.2V Vref, and can also be used to generate an internal Ibais by putting a resistor to ground.

The 2.2V Vref is used to offset the op-amp inputs to 2.2V. They will always sit at this voltage.

The current output of the DAC is a sink - the current always flows into the pin between 0mA and 2.3mA. As this current flows through the feedback resistor, Vout is generated at the op-amp output, in this case (with 1K3) giving a voltage swing of 0V to 2.99V.
As the op-amp is offset by the Vref of 2.2V, the output will actually swing between 2.2V and 5.19V - well within the limits of the device.

By adding an Ibias resistor you can add a fixed positive output current to the DAC thus giving an extra fixed offset to the generated Vout. As we use the Vref offset, this isn't really necessary. It would only be worth using if you used a lower supply voltage & wanted to maximise the Vout swing - in that case you would want to trim the Vout to be right in the centre of the op-amps available Vout swing.

Just seen your reply - so yes, always connect the Vref to the positive input as on the datasheet, but an Rbias shouldn't be required.
 
Very pleased with the AD817. Warm, quite soft and open with a lot of dynamic power and solidity. Perhaps a touch of "overdrive" to the sound, so perhaps a little less feedback is in order.

The set up is an 8V LM317 with 4.7uF X7R 1206 on the ADJ and input, 4.7uF and 100nF 0805 decoupling on the opamp itself, and 1k3 in parallel with 5pF NPO as per the datasheet. A groundplane and (I assume) RFI shield is provided by a 3M embossed copper tape wrap.

All hooked together with lashings of SCSI ribbon cable, so my poor dac now looks like a schoolgirl from one of those especially naughty japanese cartoons.
 
Thanks to 'Ed for your help.

One trick I learned is to bias the opamp into class A with a Siliconix J5XX diode across the output and V+. I'm curious about how to select the best current and impedance. J511 (4.7mA, 50K) and J508 (2.4mA, 130K) are the recommended options.

Hmm.
 
The AD817 is now biased with a J511. Wonder how thisinteracts with the bias current from the TDA1543?

Sounds good. Open, good timing, very visceral and dynamic, clear and slightly dry, with an agreeable, somewhat soft tonality that I haven't heard in off the shelf gear.

The sound reminds me of some the descriptions I've read of horns. Punchy but sort of light, with perhaps not the purest and most refined presentation but still very insightful.
 
Interesting!

As the circuit is, the op-amp is effectively only operating in 'class A' as the DAC current output is unipolar (only goes into the DAC output), but only just. You aren't using the DAC output bias, as it doesn't really need it, so this doesn't complicate things, & it's probably quieter for it.
That op-amp has plenty of current headroom (50mA?), so you could add a fair bit of biasing if you wanted (but watch overall power dissipation!) to see how it sounds.
The only thing to be wary of in that case is introducing gain errors and noise if the bias current source isn't a perfect current source - you might throw away the benefit of the op-amps low distortion.

I would say with both these chips that the power supplies were very important to get as quiet as possible - especially the DAC. It would also be worth trying a little smoothing on the Vref from the DAC.

Good to hear you are hearing interesting results, anyway! It's quite a different approach to the 'passive' method, but reassuring to know the pudding tastes different, at least!
 
You must log in or register to reply here.


advertisement


Back
Top