Measuring CCS impedance

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi All,
I've been playing with a test setup similar to Gary Pimm's shown here: CCS performance measurments .

When I tested a CCS I've got good results but was limited by the noise floor of my soundcard adapter. Here is a summary of my tests: More on CCS | Bartola Valves

CCS-test-1024x656.jpg


Actually the 1Meg resistor is only 47k to speed up charging time of the coupling capacitor

I built recently a 60dB differential preamplifier which I used with success in measuring complex impedances. However, when I try to use same setup is not possible. The CCS will develop a voltage across the 10ohm sensing resistor which obviously is way too much for a 60dB preamp and it saturates the stage.
I may be blind and not see an obvious way of fixing this, but how can I use my differential preamplifier to capture the low level signal without getting full clip at the output? I thought about returning the CCS DC before the sensing resistor but this puts the HV supply plus the 4K7 resistor in parallel with the CCS and will distort the measurement of the impedance as the CCS Zo is in parallel with the 4K7 resistor. I thought about adding an extra CCS between the power supply and the CCS DUT, but this again puts two CCS in parallel when measuring impedance.

Am I missing something really obvious here?
Thanks for the help
Ale
 
I wanted to avoid a reactive component straight into the measuring point. At least I'd need a 10uF connected between R1 and the positive input of the differential preamplifier. Also a 10K shunt resistor will be needed between the + AND - inputs of the differential preamplifier to allow the coupling capacitor to charge. Should settle within a second and shouldn't impact much the measurement as being in parallel with the 10ohm resistor in AC.
In theory should work but wouldn't be the most accurate way of doing this...
 
When I modeled every 2-device topology I could think of, a feature became clear you won't see on the bench. The stiffness varies with frequency. So before doing the bench verification, I recommend doing a investigation in SPICE. On the bench, do what Jung said.

I settled on using feedback pairs. With good devise selection, they have less drop so better for VAS, and are "good enough", far better than a LED/BJT but not the super stiff JFET/BJT cascode. They hold up past 20K pretty well, but a self biased FET is the champ for that.

After about a year of fiddling, I concluded that one has to make a real amp with a socket to swap then in and out as what the models tell you may not correlate with sound. Is consistent better than absolute stiffness? If your rails are regulated, is a resistor sonically better?
 
Disabled Account
Joined 2012
That reminds me --- back in the 1970-ish era, there were current diodes like 1N5314 etc. I used them because the higher voltage ones (50v I think) were made with cascoded FETs for higher Zo. simple and small and just a drop-in part.
They are still around, I'm sure.

Thx-RNMarsh
 
Well-designed circuits measure sufficiently good but, to some, sound bad. Different people draw different conclusions from this.

It seems to me that if you want a CCS then you should have a good one: high impedance, linear impedance, not too frequency dependent apart from the unavoidable stray capacitance. Device parasitic capacitances are the thing to watch, as they are very non-linear.
 
Hmm. Every time I mentioned a depletion mode FET, I got sneered at. They make sense to me.

Anyway the question I keep bring up is which is more important to the sound, maintain a consistent impedance through the audio range, or being as stiff as possible? Where a self-biased JFET is not as stiff as the standard BJT-LED, it has much wider BW. The JFET/BJT cascode is many times stiffer, but falls off even faster. At some point, there is always a diminishing return from every parameter.

So, is the slighter stiffer feedback pair that holds up a little longer SONICALLY better than a single FET? How much better than a resistor is really needed? 5 times? The FET can do that. 1000 times? You need the cascode. A cascode has a larger minimum drop which may cause issues in a VAS, but be fine in the IPS. What happens to all of the choices when at clipping? Some get ugly.

DF is quite correct, if you play with different devices, you can get very different results. For any topology, you need the correct component selection. Green LED's are more linear with temperature than RED, but than again, the CCS should be in a very stable environment by definition so it should not matter.
 
tvrgeek said:
Anyway the question I keep bring up is which is more important to the sound, maintain a consistent impedance through the audio range, or being as stiff as possible?
Stray capacitance puts these two objectives at odds with each other. Better to allow for the capacitance, and ensure that it and the CCS impedance are as linear as reasonably possible.

Fortunately, linearity and high impedance are not necessarily enemies as being really good at one means you don't have to be quite so good at the other.
 
Ale: While the "Noise Inspector" is a nice implementation by TI -- the circuit noise with the OPA37 front end is 30nV/RtHz -- you can do better with a single chip design like the SSM2019 or THAT1510 orTHAT1512. Which are differential in, single out. With A=1,000 the SSM2019 is spec'd at 1nV/RtHz. The measurement amplifier I used for the regulator shootout used a pair of back-to-back electrolytics and clamp diodes with the SSM2019 -- (WJ used the SSM2017 in his 1995 articles, but it has been supplanted by the 2019). Response for the SSM2019 was flat from 10Hz to 100kHz.

Next time I will include the clamp diodes, but will also have a switch so that the coupling capacitor can be "pre-charged". (Even with AC Coupling I have blown right through the protection). It will be balanced in-balanced out with BF862 and OPA1632 and a servo.
 
Last edited:
Stray capacitance puts these two objectives at odds with each other. Better to allow for the capacitance, and ensure that it and the CCS impedance are as linear as reasonably possible.

Fortunately, linearity and high impedance are not necessarily enemies as being really good at one means you don't have to be quite so good at the other.

I modeled a dozen or so two-device CCSs and I would say this is not so. Yes, the device selection is important as you can make the best circuit poor with a bad choice, but the topology seemed to have a large effect on the impedance over frequency and it was the primary driver of basic stiffness.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.