Using the AD844 as an I/V

I/V's

What I found that for I/V duties any opamp no matter how good, if they have feedback networks around them, just can't cut it when compared to the zero global feedback AD844.
Someone please correct me if I'm wrong, but I think that the extreme HF rubbish and glitches that comes out of dacs somehow have adverse effects on I/V stages that have feedback around them, sending them maybe into spurious oscillations or ringing? Where the AD844 without feedback remains more stable with these nasties coming into it.
Sort of like amps tube or solid state which become more stable when feedback is greatly reduced (or removed in tubes).

Cheers George
Hi George, I have a number of I/V solutions here. From simple resistor I/V to transimpedance types, discrete Pedja common base. You know they all have a place. At the moment the Video opamp is performing at a very high level. The GIC is transparent and the sound of the OPA627 makes for a very good buffer. Having said that I am still going to rework the triple stack AD844. It does have a nice sound and I suspect it might actually be best with the PCM1704. I found it was a bit dark sounding at times with the TDA1541. I am sure that is a system specific kind of thing. I still admire the AD844. That is why I am going to put some more effort into it. Being a bit current starved I wonder if putting a 4 mA output into it "even stacked" might overwhelm it? With 1.5 mA from the PCM1704, I think now a triple stack might be very happy. Really looking forward to giving that a try. More to your statement... I did see some very nice sine waves coming out of the "Parallel DAC" while running test tones. That was resistor I/V with a Tube SRPP hybrid jfet stage. I am sure the HF rubbish was well controlled in the circuit. So there could be a lot of merit to your comment. I find DIY to be very educational and I look forward to learning more about I/V's. 😀
 
GIC Filters

That would be truly amazing if true. How have you checked its transparency?
It isn't any worse then anything else I tried might be a better way to say it. To bad I don't have anyway to measure THD. FDNR GIC filters actually shunt the signal path and that way no active device are in the signal path. They simulate a rather large inductor. I like the equal delay at all audio frequencies for impulse response with the Bessel alignment. It was way better than a Sallen key filter it replaced. So in truth, I do make a small assumption there. Your keeping me honest abraxalito. BTW: I am glad your helping Mayday with his regulator project. 😀
 
I've played with GICs in the past, but only as far as simulation. But that 'no active devices in the signal path' mantra is just a saying, it doesn't pan out in the real world. Active devices shunting the signal means they're going to affect it.

Yes Bessel filters are jolly nice for how they maintain an almost perfectly constant group delay. Yet I tend to go for the faster roll-off types and hang the phase issues because out-of-band hash is rather deadly to dynamics. So its horses for courses 🙂

Let's hope Mayday keeps his project ticking over...
 
Agreed. Should I just remove the 330N caps or go back to 47N where I came from?
Funny enough, although people are reporting that the data that SAA7378 is supplying (CD723) is not bit-correct and the DAC is operating at 4xfs, I can hear no flaws. Compared to my also modded Oppo BDP-95 the difference is getting very small.
Regards, Jürgen

Remove 330nF first. Then reconnect. Repeat with 47nF.

Nick

any news? have you tried it yet?

Nick
 
No, I don't believe that its truly an issue of feedback being overwhelmed by HF. At least I've not seen it - I'm working on a feedback based discrete I/V stage right now so I'll keep my eyes open for this and report back if I see anything that confirms your hypothesis.
I hope you are all familiar with Nelson Pass´s explanation why an opamp is not very well suited for I/V. See here: https://www.passdiy.com/gallery/other/zen-i-v-converter
 
FDNR GIC

I've played with GICs in the past, but only as far as simulation. But that 'no active devices in the signal path' mantra is just a saying, it doesn't pan out in the real world. Active devices shunting the signal means they're going to affect it.

Yes Bessel filters are jolly nice for how they maintain an almost perfectly constant group delay. Yet I tend to go for the faster roll-off types and hang the phase issues because out-of-band hash is rather deadly to dynamics. So its horses for courses 🙂

Let's hope Mayday keeps his project ticking over...
Hi abraxalito, I do agree with you. Experimenting with different opamps in the GIC legs I did find some sonic differences. The best opamp was the OPA604. It didn't misbehave in any way I could determine. Guess that is why Burr Brown used it in there datasheets. I anticipate putting the GIC back in my first TDA1541 build. I could of course try a 3rd order on the 1704 however an experiment likely a long way off. Really want to get the AD844 stage optimized for the 1704. Mayday... Has a lot of interest and sometimes strays from his projects. :spin:
 
any news? have you tried it yet?

Nick

Good evening Nick,

just disconnected the 330N caps two hours ago. For a first impression I`d say the difference is very marginal. Something gets me thinking: Didn`t Joe say that the cap does very little with dacs like TDA1545 which I am using here? But he is the one that suggested to change to 330N....
Anyhow I´ll keep testing, knowing very well that it is easy to fool ears and brain...
And maybe important to know that due to my age (55) my ears cannot detect frequencies close to ultrasonic...
Seems that not many People here are doing a comparable test with their equipment.
Cheers, Jürgen
 
I hope you are all familiar with Nelson Pass´s explanation why an opamp is not very well suited for I/V. See here: https://www.passdiy.com/gallery/other/zen-i-v-converter

Thanks for that Bach.

Looks like Nelson thought the same about opamps that use feeback for I/V duties, and the HF rubbish and glitches that comes out of dacs that could upset them.


Part of the quote from Nelson Pass:

"Later I noticed that I was hearing differences depending on the relative gain (and feedback) of op amp stages even when there was no oscillation. It appeared as if simply approaching the point of instability created issues that didn't show up on the distortion analyzer. Since then, I have learned to give potential feedback issues a wide margin.

On another occasion I designed an experimental amplifier that proposed to use a high frequency bias (similar to tape recorders) in an amplifier output stage as a means of reducing crossover distortion. The result measured reasonably well, but the sound was awful. I concluded that mixing high frequency tones with audio and feedback is a bad idea.

I encountered another such instance in a vinyl cutting session with Telarc where an ultrasonic tone of low amplitude was accidentally introduced into the system. The resulting record had issues that the producers could hear plainly on the pressing (Yes, Robert Woods and Jack Renner had excellent ears) and no one could guess at the problem, but I played the record into a spectrum analyzer and identified the tone. They remastered it from scratch.

And that's why I think these things.

Given the idea that the I/V converter above is inviting trouble, what can we do?
."




Cheers George
 
Last edited:
I hope you are all familiar with Nelson Pass´s explanation why an opamp is not very well suited for I/V. See here: https://www.passdiy.com/gallery/other/zen-i-v-converter

Barrie Gilbert has written in a similar vein about opamps too, there's an article 'Are opamps really linear?'. Here's a link to that article - http://linearaudio.nl/sites/linearaudio.net/files/Are Op Amps Really Linear.pdf

That 'Zen' I/V looks to have rather highish input impedance (quoted around 15ohms) - I prefer lower as it gives the music more 'drive' (or 'snap') somehow, at least with the DAC chip I use (TDA1387).
 
Looks like Nelson thought the same about opamps that use feeback for I/V duties, and the HF rubbish and glitches that comes out of dacs that could upset them.

I agree that opamps in general are poorly suited to I/V - there are a couple of reasons for that ISTM. Firstly the input stage gets regularly overloaded when its an LTP, as its subjected to very fast step functions. Hence Scott Wurcer's proposed solution (in his AD797 DS) of fitting a capacitor directly to the DAC's output. The second reason is that the DAC's output current (normally at least 1mA) goes beyond the classA region of the opamp's output stage operation and this introduces supply line noise.

The first problem is solved by choosing a CFB opamp such as AD844 - no LTP is present. There are also certain VFB opamps which avoid the LTP - LM6171/2 is an example, LT1354 is another. I've used LM6172 as I/V and it sounds good. Another option would be one of ADI's opamps where the LTP has been 'hardened' i.e. made more linear.

My point in the earlier reply to you George was that the woes of feedback can be separated from the woes of opamps. I'll probably put up the schematic of my CFB discrete stage sometime soon on my blog - I listened to it for the first time last night. This is using feedback to obtain a very low input impedance but has no LTP and its output stage runs pure classA.
 
Both issues you stated in the first paragraph could be alleviated by implementing AD811, but without any capacitor(s) across DAC output.

I also use LM6172 - after 811 and only if I need to sum differential outputs to single-ended... this combo is the best I can do with OP's and what's available.

1704 requires active amplification - 811 provides very low impedance, low distortions and gain at the same time. Anything else that can achieve the same combined benefit will also have the flaws.

Nick
 
Do you have any idea where the AD811 output stage leaves classAB? I know its a hot-running chip so might be well biassed.

Another chip to consider would be AD8016 or even the older AD815 for I/V duty. AD8016 lets you play with the bias via a couple of programming pins.

For myself I'd not use LM6172 for summing duty - I prefer JFET opamps for that where I can use high value resistors (upwards of 100k). LM6172 has rather high input bias current so would need fairly low Rs in its feedback network.
 
Hi,

M.Hawksford had written about the difficulties of (differential) Input stages and global feedback loops when used as IVs and he came up with the grounded base circuit, but added with parts for error correction.
He prooved in his paper that a for a OPAmp not to suffer from an overdriven Input stage it'd either require a very fast video OPAmp with hundreds of MHz if not GHz bandwidths, or a cap at the DACs output to slow down the slew of the current steps.
Built a I/V after his documentation and wasn't overly impressed with its performance.
I doubt that if You can achieve better than -90dBfs of THD with a nice spectrum with a simple Jocko grounded base circuit, that even lower THD means no improvement to the sound, but the additional parts required for EC are actually spoiling things.
Besides one could simply add a second transistor in Darlington fashion, preferrably a CFP Darlington to reduce THD and to achieve very low input impedance and to allow for higher idle currents.
This means of course introducing feedback, but of the local, degenerative type and not of the problems loaden, several stages spanning looping.
The typical small signal JFETs like used in the Zen won't achieve input impedances of less than 10Ohms without massive paralleling of devices.
A CFP could help here also, preserving the JFETs low distortion behaviour due to the missing base current.

jauu
Calvin
 
Anyone???
Is it possible to duplicate these HF nasties and glitches that come out of these dacs.?
So one can see/measure what's going on with feedback type I/V opamps when they are subjected to them (especially while the dac is doing the conversion process of the digital music signal)?

Cheers George

You can look at that rubbish with any decent 100MHz oscilloscope. We are talking mostly NOS and current out DAC's, right?

You'll find that capacitors across these DAC outputs are absolutely NOT needed. Even in a standard non - NOS implementation they will cause more harm than good. This is the reason why all DAC (CD player) manufacturers who used 1704's left a provision on the PCB for these capacitors, but the capacitors are always missing, i.e. they are not installed.

Nick
 
Last edited:
Hi,

M.Hawksford had written about the difficulties of (differential) Input stages and global feedback loops when used as IVs and he came up with the grounded base circuit, but added with parts for error correction.
He proved in his paper that a for a OPAmp not to suffer from an overdriven Input stage it'd either require a very fast video OPAmp with hundreds of MHz if not GHz bandwidths, or a cap at the DACs output to slow down the slew of the current steps.
Built a I/V after his documentation and wasn't overly impressed with its performance.


jauu
Calvin


So.... did you use the fast video amplifier? AD811 is a current feedback OP specifically designed for video amplification, it comes in a "standard" DIL8 package + it has static impedance of few tens of ohms. Once properly implemented, a dynamic impedance seen by 1704 is well below 1 ohm (!!)

The problem with this amplifier is in its implementation!! It must not oscillate. Many people try "plugging it in" which of course will overheat them to destruction. So, careful planning is required, with a bit of measurements to ensure they don't oscillate.

811's come in a ceramic package as well - sounds even better!

Nick
 
Do you have any idea where the AD811 output stage leaves classAB? I know its a hot-running chip so might be well biassed.

They draw a lot of current. I asked myself (because there's no info available with regard to bias) - what could possible be drawing that amount of current...? I concluded that input stage and "VAS" could not be the only sections to blame for such draw... which left me thinking that it must be the output stage...What else if not that??? They are by far the best OP's I used for I/V duties. But, they don't like the capacitors (apart from decoupling of course, which needs to be done properly). They don't even like the Rhopint feedback resistors - because they are slightly inductive (!!).

Nick
 
You can look at that rubbish with any decent 100MHz oscilloscope. We are talking mostly NOS and current out DAC's, right? Nick

The steady HF rubbish yes can be seen, as I have seen with my 400mhz Tek scope.
But it's the glitches that can't be studied, and how to duplicate them. To see what they do an I/V opamp with feedback, while music is involved.

Texas Instruments/Burr Brown PCM1704 that is suppose to be "glitch free" in reality is not, so the digital geniuses over at the other place say.

Cheers George