[sdiy] Testing OTA performance

[Magnus Danielson]

From: jhaible at debitel.net
Subject: Re: [sdiy] Testing OTA performance
Date: 4 May 2006 16:32:53 +0200,Thu,  4 May 2006 16:32:53 +0200
Message-ID: <1146753173.445a10954ece3 at www.debitel.net>

> > > Hey SDIYers,
> > > 	I am testing some parameters for various OTA's.  The OTA's being  
> > > tested are the 3080, LM13700, JRC13700, and JRC13600.  I have  
> > > measured the control current versus gain, and I also want to measure  
> > > the noise output of the OTA's.  I am thinking about hooking a large  
> > > amplifier up to the output of the OTA, with no input into the OTA,  
> > > and seeing how flat that output signal is.
> > 
> > "no input" as in grounded inputs or as in free-floating inputs?
> > I hope you meant grounded.
> 
> And ground them with low impedance (short circuit, or something like
> 100 Ohm), in order not to measure the noise of the resistor, when you
> have a really good OTA.

Exactly. When I say ground them, I mean solid wire, else I would explicitly say
ground them through a 10 k resistor.

Thanks for that clarification! Very good point. I had that in my head, but
forget to say it.

> > > ideas out there?  Also, any ideas on how to measure the distortion of  
> > > these OTA's?
> > 
> > There are three parameters that comes to mind:
> > 1) Input amplitude
> > 2) Iabc current
> > 3) Frequency
> 
> I wouldn't bother too much with distortion measurements.
> If you really have an OTA (and not some other VCA), the
> input differential pair will give you the same distotion
> in practice as the textbook will predict.
> That's without the linearising diodes (if there are any).

Some of his devices have linearising diodes, so then you would see the
importance of them. What I proposed would be an enligthening measurement round,
if the linearisation is utilized contra when it is not.

> > Don't forget the moon-phase and time to next equinoxe!
> 
> I would add another important test: CV feedthru, or variation
> of offset voltage with the control current.

Indeed. I forgot to mention that. There is really two different things,
DC-leakthrough and AC-leakthrough. It may not be the same at higher frequencies
than we see at DC and low frequencies.

> For this, set up a 1-pole VCF (OTA, cap, buffer opamp, feedback)
> and change Iabc over a wide range. (like, 100nA to 1mA)
> (Caution: There are some OTAs that die on 1mA, but these are the
> exception.)

I would not do it like that, I would rather have a resistive load or resistor
over an op-amp. Then I would measure the output DC voltages for the different
Iabc currents.

Cheers,
Magnus