George,
This circuit will work extremely well for PCM1704 with a few simple mods to
'bend' it towards lower current / higher OP impedance.
I have built virtually same circuit with 10k OP Z (load) and the simulated
results were somewhere around 0.0001% at 20kHz depending on the
particular design. There are many variations to the basic theme.
It will need a servo but that is also not too difficult.
cheers
Terry
The output transistors already keep the voltage variations of the current sources at minimum, so there is no need for cascoding the current sources, only a basic ccs is needed. Thanks for the image.
looking for low noise reference voltage, have found this.
http://www.diyaudio.com/forums/parts/35821-some-noise-measurements-leds-zener-
http://www.diyaudio.com/forums/parts/35821-some-noise-measurements-leds-zener-
test with 3 different output configuration. this was made with 1kohm of input impedance to bring up distortion.
first is normal x connection, second is with drivers collector connected to OUT, third is drivers collectors connected to ground.
with high input inpedance the distortion will not be visible in the first 2 images.
This proves that low distortion can be achieve without the x connection.
first is normal x connection, second is with drivers collector connected to OUT, third is drivers collectors connected to ground.
with high input inpedance the distortion will not be visible in the first 2 images.
This proves that low distortion can be achieve without the x connection.
Attachments
Last edited:
The one i have in breadboard is with blue led (bad choice), the output voltage varies no more than +-20mv with temp. But I recommend a output capacitor. Its possible to use a dc-servo, but i dont think it is not necessary, it makes sense in a power amplifier to get rid of the bipolar capacitor in the feedback , but in this case i think that a polypropylene capacitor will be the best choice, but it is open to discussion 🙂
There was no transistor matching, the transistores was chosen at
random, and the resistors are 5%, the output dc offset is set by a 100 ohms trimmer in the upper currente source.
I will use a current source of 6.2ma in the input to cancel the dc offset of dac as suggested by Calvin somewhere in this thread
Thanks.... I think, a good solution... if you make a cascoded Jfet CCS. you have Tera-Ohm impedance for the first more than 100 KHz then dropping with the capacitance of the cascode Jfet. but never lower than a couple of hundred Kohms.
Thanks very much Terry, i will take a closer look.
Do you have any suggestion to improve the circuit ?
MiiB Jfet CCS are the simplest solution for making a CCS but the jfet parameters are not very accurate. but is a possibility. 😉
im sorry but Tera-Ohm impedance is not possible. even 1pf of parasitic capacitance would bring it down even at 1Hz. The cascode serves to stabilize the voltage across the ccs, but in this case is not necessary as the input voltage is 9mv rms maximum.
Have you already read the paper about ccs that ken newton post ? if not read it , there is a lot of good information.
You may be right..., however my LT simulator says different.... I wonder if the models aren't sufficient even to calculate CCS impedance.... hmmm So much for for believing that..
the TL431 has high impedance but starts loosing it after only 1K, but is stiller better at 100KHz than the two other... Can you make one cascoded Jfet... ??
George, whether a 2.35k output impedance is too high or not will mostly depend on the total amount of shunt capacitance presented by the interconnect cable (typically, about 50pF per foot) plus that of the input of the next gain stage. The output impedance forms a low-pass filter in conjunction with that shunt capacitance.
Output impedance in itself does not say anything about the sound quality of a line level gain stage. For example, the Hovland HP-100, for example, was a very highly regarded tube linestage having a 2.4kohm output impedance.
Last edited: