Transistor Vbe Matching Using A Commutating Chopper-Based Circuit

Where we discuss new analog design ideas for Pro Audio and modern spins on vintage ones.
User avatar
mediatechnology
Posts: 5437
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Transistor Vbe Matching Using A Commutating Chopper-Based Circuit

Post by mediatechnology »

The ZTX851 Moving Coil Preamp doesn't necessarily require, but benefits from matched ZTX851 input transistors. Though it's possible to match the ZTX851 on the assembled PC board in-circuit, doing so requires installation of a test socket. Matching the ZTX851 in a test jig is a better solution.

Updated 10/8/20: viewtopic.php?f=6&t=1153&p=15635#p15635

Ian Fritz has come up with a really simple device that allows direct differential measurement and I've used it to match several batches of transistors.

Image
Ian Fritz Transistor Matching Circuit

R1 and R2 require extremely precise matching or trimming. When R1 and R2 are not trimmed, Ian describes a method of swapping the emitter leads and mathematically subtracting out the resistor match error to obtain delta-Vbe.

Even with R1 and R2 carefully matched or trimmed, their tempco, or the tempco of the trimmer becomes significant.

I decided to look for a method that did not require component matching which shared a common resistor and came up with the following:

Image
A Commutating Chopper-Based Transistor Vbe Matcher

This is a proof-of-concept.
In the circuit above the transistors share a common emitter current sink resistor to eliminate the matching requirement.
The reference transistor, Q1 and the test transistor, Q2 are alternately switched on and off. I used 500 Hz.

When transistor Q1 is on and Q2 off, the base of Q1 is grounded and the emitter is at Q1's Vbe.
Let's assume transistor Q1's Ve is -600 mV.
When transistor Q2 is switched on it's Vbe will likely be different.
Q2's Ve may be -603 mV.

As the DG413 CMOS switches at the bases of Q1 and Q2 are toggled, an AC potential will be developed at the emitters whose peak-to-peak voltage will be equal to the delta-Vbe of the pair.
Switching the bases, rather than the emitters, reduces the effects of the DG413's RdsOn mis-match.
The output is read on a high gain AC-coupled oscilloscope.

There are two advantages to this method:

(1) The matching requirement and temperature drift of outboard resistors and trims are eliminated.
(2) The delta-Vbe can be read visually and very rapidly. Reading a moving target on a DVM is more difficult on a DVM than on a scope. As the measurement settles less interpretation is required.

A large mis-match is a high amplitude square wave - a close match has low amplitude.
There is some switching noise between Vbe pedestals which has to be visually ignored.

The 100K resistor, when combined with the 'scope probe and lead capacitance, form an RF low pass filter to prevent AM broadcast ingress and do introduce some measurement error of the peak-to-peak signal with a 1M scope input.

I mated up about 10 pairs of ZTX851 using this jig and also used it to check about half a dozen pairs I matched with Ian's jig.
The correlation between methods is pretty good.

A 'scope with a high vertical sensitivity is required without a preamp stage.
My 'scope has 5 mV/div with a X10 expansion so one division works out to be 500 uV/div.

Interesting observation is that ZTX851 tend to match up within 500 uV pretty easily.
When the Vbe is matched the Hfe tends to be pretty close too.

Updated 12/22/19:

This is the latest schematic for the Commutating Transistor Vbe Matcher.

Image
A Commutating Transistor Vbe Matcher

More here: https://proaudiodesignforum.com/forum/p ... 420#p14420
User avatar
mediatechnology
Posts: 5437
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Transistor Vbe Matching Using A Chopper-Based Circuit

Post by mediatechnology »

This is a scope photo with a pair having a Vbe mismatch of just under 500 µV.

The 100K build-out in the schematic eliminates a lot of RF floating into the image by using the probe capacitance as part of the low pass filter.
A shunt C at the output wasn't particularly effective and actually made the RF ingress worse.
There's probably a more elegant way to do filter the AM radio out.

Image
ZTX851 MC Preamp Commutating Transistor Vbe Matcher Scope Photo

The current used in this test is somewhat lower than the preamps operating current to reduce thermal settling time.
At 5.5 mA the thermal tc is a little long so I went with just over a mA.
emrr
Posts: 569
Joined: Sun Mar 02, 2008 10:27 pm
Location: NC, USA
Contact:

Re: Transistor Vbe Matching Using A Commutating Chopper-Based Circuit

Post by emrr »

Nice. I could have used that back when I was rebuilding a lot of Altec 9470A/9475A push-pull preamps. Those just don't work at all if vbe is out of scale or mismatched.
Best,

Doug Williams
Electromagnetic Radiation Recorders
User avatar
mediatechnology
Posts: 5437
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Transistor Vbe Matching Using A Commutating Chopper-Based Circuit

Post by mediatechnology »

I wanted to preview a way of measuring the Vbe mis-match using a soundcard and FFT.

I modified the circuit in the original post to include a 500 Hz oscillator using a couple of CMOS gates.
I also added an AC-coupled buffer for the output.

It dawned on me that the best way to measure low-level AC signals is by using an FFT so in addition to feeding the oscilloscope I also output the buffered AC component of the mis-match into the Focusrite 2i2's A/D input.

The actual waveform of the Vbe mismatch is a square wave with switching artifacts.
The amplitude of the 500 Hz component (about 490 Hz) is directly related to the delta-Vbe.
I first used the analyzer with dBV scaling: A -60 dB 500 Hz reading is about a 1 mV mis-match.
I later calibrated the system with a 500 mV squarewave so that the 500 Hz peak directly reads in mV.

This is what it looks like measuring two transistor pairs.
One pair has a 100 µV mis-match the other 450 µV.
By displaying the error visually its possible to follow large changes after device insertion using the dBv scale and find small differences using direct mV readout after the pair has stabilized.

Image
ZTX851 transistor matching using a soundcard and FFT to read delta-Vbe.

Measurements with the FFT correlate closely with the 'scope.

So to briefly summarize I use a commutating switcher for the emitter load using the DUT's themselves so there are no resistor mis-match errors.
This makes an easier-to-measure and easier-to-track-changes AC component equal to the delta-Vbe.
The level of the delta-Vbe is then measured with a sound card to provide measurement resolution below 1 mV.

I'll post an updated schematic as soon as I try a few more tricks.
User avatar
terkio
Posts: 304
Joined: Thu Feb 07, 2019 3:17 am
Contact:

Re: Transistor Vbe Matching Using A Commutating Chopper-Based Circuit

Post by terkio »

Error from DG413 switches.
Delta Ron max 3 Ohm.
So with Ic 1mA, Hfe 100. Error max delta Vb is 30μV.
Ron max 35 Ohm.
So with Delta Hfe 10%. Error max delta Vb is 35 μV.

Overall measurement error from the DG314 is 65μV max @1mA Ic.
User avatar
mediatechnology
Posts: 5437
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Transistor Vbe Matching Using A Commutating Chopper-Based Circuit

Post by mediatechnology »

I think you worry too much.

1) The static rds On errors are well-balanced between sides.
2) The delta-rds from 0 to +/-600mV with 15V supplies is <<1Ω. Try matching a a pair of 10K emitter resistors that close even with a trim.
3) If it were actually 65 uV that is only about 1.6% of the error of the up to 4 mV error you might have if you didn't select them. (ZTX851).
4) In the real world you're never going to get a consistent match in a test jig below 100 uV because the short term thermal drift of an uncoupled test pair is easily going to be more than the switch error.
5) You'll spend all day looking for that 25 uV pair and, when you do find one, there will be moments when it looks like a 100 uV pair.

Image

I don't see a delta-rds of 3Ω under the conditions these switches operate.
Where is it?

During the measurement period the switches ground the bases which is the center of the curve.
The switches only supply -600 mV (NPN) or +600 mV (PNP) when the device is in cut-off.
The delta-rds when the devices are in cutoff is de minimis.

The delta-rds between switch sections at 0V when the DUT transistors are in conduction would be the dominate error source.
I submit that value to be small.
User avatar
terkio
Posts: 304
Joined: Thu Feb 07, 2019 3:17 am
Contact:

Re: Transistor Vbe Matching Using A Commutating Chopper-Based Circuit

Post by terkio »

The figures I used are from Maxim.
https://www.maximintegrated.com/en/prod ... DG413.html
They are general worst case figures, giving a wide upper bound about the measurement accuracy.

Your analysis is far more accurate. Actual accuracy is much better than 65μV.

This is for Ic 1mA and Hfe 100.
Accuracy is scaled by Ic and Hfe.
May be, it should be looked at in case of higher Ic
It looks good for higher Hfe like 500 of the promising low noise ZTX689B
User avatar
mediatechnology
Posts: 5437
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Transistor Vbe Matching Using A Commutating Chopper-Based Circuit

Post by mediatechnology »

I do however think that a lower test current is justified since the thermal settling time is quicker.
If I use 100 uA rds-On errors are 1/10th.

Duty cycle of the test waveform is also critical: I'm adding a flip flop to the simple astable I was using.
At 1 mA test currents the unequal self-heating is a factor.
User avatar
mediatechnology
Posts: 5437
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Transistor Vbe Matching Using A Commutating Chopper-Based Circuit

Post by mediatechnology »

Another factor that affects consistency are the emitter contact resistance of the test sockets.
That error isn't divided by hFE.
I have a 14 pin Textool ZIF on order.
I already have a 28 pin Aries but I'm reluctant to use it since it seems wasteful.

I did modify the switches to invert the control to one and rewired the DG413 connections to make both switches "On" with the same polarity internal CMOS transistor thinking their Rds On (at 0V) might be more matched.
It didn't make much if any difference.

Just for grins I increased Re to 100KΩ for an approximate 150 uA test current.
The decrease in thermal settling time is significant.
Then I inserted a 47Ω in the base of one transistor.
The hFE of this pair is about 130.
I could see a slight change which should be about 55 uV.
47Ω would be an extreme Rds On imbalance.

I still have a lot of work to do.
The main thing is getting an exact 50% duty cycle so I need to add a flip-flop.
I'm way past the schematic in the OP and this one's fluid.
User avatar
mediatechnology
Posts: 5437
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Transistor Vbe Matching Using A Commutating Chopper-Based Circuit

Post by mediatechnology »

I added a 4013 D flip flop to provide an accurate 50% duty cycle.
This made a big difference in consistency.
The thermal offset, from self-heating, is significant with unequal duty cycle even at 150 uA Ie.
Post Reply