The one spec to look for in particular is the op-amp's slew rate. Higher the better. A fast enough slew rate means the op-amp's output can keep ahead of the feedback loop at all times. Too slow and the output gets distorted while the op-amp's output struggles to get to where the feedback loop wants it. This distortion would look like a ramp clipping the waveform. The old uA741 had a poor slew rate, which led to poor sound.
This issue with op-amps used as IV converters is that of slew rate . Or lack of enough thereof. Current output DACs produce spectral energy upwards to around 30MHz or even higher, depending how you measure it. Most op-amps you find in CD players have a slew rate of around 10V/usec, but the DAC here would need an op-amp that could do 1000V/uSec. The op-amp inside the PCM61 DAC chip has a slew rate of only 12V/uSec. With the slow slew rate op-amp the virtual ground the DAC is feeding stops being at ground until the op-amp slews its output voltage to draw the current thru the feedback resistor to get the virtual ground back to zero volts. You'd see triangular voltage spikes of a few volts during the time the op-amp is slewing. And the slope is pretty constant, so the area of the trangle would vary non-linearly against differences of the size of the signal step out of the DAC. This trangular region you see on the virtual ground is essentially that you find missing from the op-amp's voltage output. A possible solution to this would be to use a high speed op-amp in place of the original op-amp. Op-amps designed for video work would be good for this. You want an op-amp with a very high slew rate specification.
This issue with op-amps used as IV converters is that of slew rate . Or lack of enough thereof. Current output DACs produce spectral energy upwards to around 30MHz or even higher, depending how you measure it. Most op-amps you find in CD players have a slew rate of around 10V/usec, but the DAC here would need an op-amp that could do 1000V/uSec. The op-amp inside the PCM61 DAC chip has a slew rate of only 12V/uSec. With the slow slew rate op-amp the virtual ground the DAC is feeding stops being at ground until the op-amp slews its output voltage to draw the current thru the feedback resistor to get the virtual ground back to zero volts. You'd see triangular voltage spikes of a few volts during the time the op-amp is slewing. And the slope is pretty constant, so the area of the trangle would vary non-linearly against differences of the size of the signal step out of the DAC. This trangular region you see on the virtual ground is essentially that you find missing from the op-amp's voltage output. A possible solution to this would be to use a high speed op-amp in place of the original op-amp. Op-amps designed for video work would be good for this. You want an op-amp with a very high slew rate specification.
Last edited:
My experience trying countless numbers of opamps that use feedback, is that they are all flawed for I/V duty use.
I only found this out after trying the AD844 in open loop mode (no global feedback), which was so much better that all the feedback opamps were to me not suitable for being "great" I/V stages.
It does not take much to implement the AD844 and also use it's inbuilt output buffer if you already have + - 15vdc for your old opamp I/V.
Read the AD844 thread on how to use it and stack only the i/v section for high current output dacs.
http://www.diyaudio.com/forums/digital-source/227677-using-ad844-i-v.html
Cheers george
I only found this out after trying the AD844 in open loop mode (no global feedback), which was so much better that all the feedback opamps were to me not suitable for being "great" I/V stages.
It does not take much to implement the AD844 and also use it's inbuilt output buffer if you already have + - 15vdc for your old opamp I/V.
Read the AD844 thread on how to use it and stack only the i/v section for high current output dacs.
http://www.diyaudio.com/forums/digital-source/227677-using-ad844-i-v.html
Cheers george
Last edited:
How nice of ADI to provide the TZ pin 5 connection!
Yes, the question in part now is what the input voltage burden is. If we knew what the emitter currents are this could be calculated readily. And it could be guessed at to some extent from the quiescent current for the part.
EDIT: They even give that information, 50 ohms typical, 65 ohms max. So this is potentially a little high depending on the DAC for voltage burden when used open loop. However, it is amenable to a treatment similar to what I'm looking at with discrete JFETs and DMOS, since we can wrap some gain around it via the noninverting input, if very low input impedance is desired.
Last edited:
As I ponder some more, note that the open loop input resistance as stated implies a transistor current of about 250uA. This means that even for a +/- 1mA full scale DAC, the inputs will be driven into class B at moderate levels. Not that that is something spelling disaster, but is still I think a parameter to be considered. OTOH, in analogy to loading output stages of opamps to make them function in class A, one could do something similar with the AD844 input stage.
Yes but this halved (25ohm) if 2 of just the i/v section of the AD844's are stacked, and only one of the output buffers are used, gain stays the same, also in the PCM1704's case of 1.2mA stacking 2 AD844 I/V sections stops any current starvation of the i/v's stage. And stacking 2 sounds far superior than one, probably becase of the class B you mentioned .
Mick Maloney of Supratek stacked 3 for the TD1541 because it has higher output and said it was the best he has heard the TDA1541 sound.
Cheers George
Last edited:
I'm one of the 'some' who hear issues with the LM4562(an LME in all but name) in respect of RF sensitivity. It was I believe Ed Simon who mentioned using an inductor LPF to the input of another of this family.
How are you simming and what metric is telling you your results are promising? I spent a few days two or three years back trying to coax some kind of correlation between what I heard and what LTSpice was telling me (by means of FFTs) and eventually I gave up
Nowadays I do have better ideas about bridging the gap though because I have a hypothesis that's testable about SQ differences in DACs.
Underperforms in sim?
Slew rate is certainly a major issue - the best opamp I've found in I/V so far is the LM6172 with 2,500V/uS. I'm not though convinced its the issue. Lack of GBW seems to be another major one - as you point out the spectrum emanating from the DAC extends to 50MHz. This was demonstrated by Lynn Olson and a buddy of his on an SA. Sticking a 50MHz bandwidth signal into an opamp with even 100MHz GBW leaves hardly any feedback over to do distortion reduction. The distortion that matters here being IMD at lower levels - hardly anything will show up on an audio band THD measurement (not enough discrete tones present) which is why the slow opamps you mention continue to get used..
Transient response, frequency response, harmonic distortion, with particular attention to the input impedances. This is the outgrowth of interest in EUVL's simple current conveyor using JFETs and floating power supplies, which I found intriguing but for which the presentation ( in here and in the Linear Audio article) also ignored the effect of the input impedance variations with current, a distortion mechanism that even with a constant output resistance "current output" DAC was a distortion generator. Since such DACs are anything but current sources this can be a significant source of error.How are you simming and what metric is telling you your results are promising? I spent a few days two or three years back trying to coax some kind of correlation between what I heard and what LTSpice was telling me (by means of FFTs) and eventually I gave up
Underperforms in sim?
I like the idea of beginning with a lowish input impedance open loop for a current conveyor. But I proposed a while back to wrap some gain around the input device to reduce the input impedance further --- not global feedback as such with an amplifier with a lot of gain, but just something to effect a reduction in input Z. I've recently revisited the scheme, and decided to sim the 5534 with an app note's recommendations for feedback R and C, just to get a rough comparison to what I was seeing in sim already. The results are dramatically different. I don't trust the sim model's distortion numbers for the 5534, but I'm confident the discrete circuit's predictions are not wildly off. The big, and probably realistic differences, are in terms of speed and input impedance.
I'll probably present the results pre and post breadboard in a new thread, but I was searching for what people had said about their various preferred solutions with opamps and other circuits. That's what led me to this thread. I did already post some things with a very simple version which merely repurposed some parts that EUVL was using in his Sen circuit, the input Z reduced circuit dubbed InSense iirc. It improved 1kHz distortion by 19dB when compared to a Sen circuit with the same assumed (constant) DAC impedance of 1k. At higher frequencies variable output capacitance becomes more important and the differences between the two are slight --- but cascoding then helps a great deal.
The circuit now involves DMOS cascode devices and BF862s, for both the Sen-style current conveyor and the input Z reducing amplifier. It's supposed that ~7V photovoltaic optos are used variously to do floating biases. Cumbersome, but practical since they are only used to drive gates, so when bypassed appropriately they should work just fine. Or, primary batteries could be used, and their lifetime would be that of shelf life.
That's about the state of affairs at this point.
just filter it! adding massive slewrate and bandwidth without proper filtering is asking for trouble; it will happily receive/rectify all of the VHF that isnt coming from the DAC as well.
Brad, interesting you bring this up, i've been thinking of ways to reduce the input impedance of a jfet FE for ESS by doing exactly what you propose, but i'm unsure of where to start with sims.
Brad, interesting you bring this up, i've been thinking of ways to reduce the input impedance of a jfet FE for ESS by doing exactly what you propose, but i'm unsure of where to start with sims.
Last edited:
In my setup, driving a 7k5 attenuator, it sounds even better if its output is taken directly from pin 5, bypassing the buffer altogether.
Tim.
The trick is to filter if necessary without adding voltage burden. As mentioned in the Curl thread the manufacturers rarely tell us the effects of developing much of a voltage, and how much the effects may involve the codes. Philips/NXP used to mention a max voltage on some converters, I seem to recall 25mV.
Yes, the 844 can certainly be used as a current conveyor and does not need to use the output buffer. And the stuff I'm working on can benefit from a buffer following, although if what it feeds is close by there is no need.
In the 844 there is probably some emitter ballasting in the Wilson mirrors, in order not to degrade the S/N by much. This is good and probably adequate for most DACs. The JFET versions along the lines of EUVL's have intrinsically rather low noise, but when the amplifier is used to reduce the input impedance that becomes the determinant of equivalent input voltage noise, but can be of the order of 1nV/sq rt Hz without too much trouble.
Yes, the 844 can certainly be used as a current conveyor and does not need to use the output buffer. And the stuff I'm working on can benefit from a buffer following, although if what it feeds is close by there is no need.
In the 844 there is probably some emitter ballasting in the Wilson mirrors, in order not to degrade the S/N by much. This is good and probably adequate for most DACs. The JFET versions along the lines of EUVL's have intrinsically rather low noise, but when the amplifier is used to reduce the input impedance that becomes the determinant of equivalent input voltage noise, but can be of the order of 1nV/sq rt Hz without too much trouble.
I tried that and yes it was very good also, but the bass was not tight enough for me, it is a very high output impedance at TZ pin 5, even a 100kohm load droped it's output quite a bit, so it needs to see a very high input impedance in the order of a few meghoms not to suffer any loading problems.
The output buffer is very good, but to get it to sound good the I/V should not be current starved, and it was I found out, so by stacking two 844 i/v's they current shared which has a bonus of lower (half) the input impedance and then the buffer stage of only one was used, and only then it's full potential was finaly realized.
I found that with a PCM1704 (1.2mA output) you need two 844's and Mick Meloney found at least three need to be used with the TDA1541 (4mA output), but I feel it would even need 4 844's.
Cheers george
This is excellent ... I used it to D/A Wolfson 192 khz 64 x ...
The voice was my dream
and I used Lme 49720 after dac ... and used oscon cap for digital PSU ... and Elna silmic II for filtering and LME psu...
for psu dac I used Nichicon KZ ...
thanks
The voice was my dream
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
and I used Lme 49720 after dac ... and used oscon cap for digital PSU ... and Elna silmic II for filtering and LME psu...
for psu dac I used Nichicon KZ ...
thanks
Last edited:
What RF choke do you use? Any recommendations? Is this just on pin 2 to stabilize the CFB opamp? Have you tried ferrite beads?
Just about any RF choke will help, I use single layer ones for better high frequency loss. I have also used beads with a few loops of magnet wire thru them. This is all to lower the amount of RF going into the op-amp. I feel it is very important to roll off the input to any op-amp used for I/V, all that high frequency output from the DAC is not needed and makes life harder for any I/V device, so lets lower it. Play around with any RF choke you might have on hand.