Component Matching Minimizes IM Distortion in the Demrow-Cohen Balanced Front End

Where we discuss new analog design ideas for Pro Audio and modern spins on vintage ones.
Post Reply
User avatar
mediatechnology
Posts: 5443
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Component Matching Minimizes IM Distortion in the Demrow-Cohen Balanced Front End

Post by mediatechnology »

Component Matching Minimizes IM Distortion in the Demrow-Cohen Balanced Front End

With Hans Polak's help I recently redesigned the Flat Moving Coil Preamp PC board.
With highly-mismatched pairs the servo control point in the original design increased IM distortion because it varied collector current.
By moving the servo injection point to the emitters on the new board large mis-matches in transistor Vbe don't increase IM.


Simulation didn't match reality when, in the process of debugging, we found that passive component matching had a huge effect on the 1 kHz IMD produced by 19/20 kHz CCIF IM tests.

The 1 kHz IM, which is produced by even-order nonlinearities, is particularly pertinent for phono preamps because the RIAA curve magnifies the IM product, relative to the stimulus, by about +20 dB.

The following schematic shows the front-end which is fairly typical for most "Cohen" topologies.

Image
Schematic of the KA Electronics FMCP Flat Balanced Input Balanced Output Moving Coil Preamp Using the ZTX851

Large schematic: https://proaudiodesignforum.com/images/ ... amp_18.png

Even-order distortion in the "Cohen" front-end is attenuated by the common mode rejection stage formed by the cross-coupled THAT1240s.

Maintaining high common mode rejection is important not only for hum reduction but also even-order distortion reduction.

Cohen touched on component matching but didn't elaborate as to why.

Image

Graeme, in Op Amp Applications, teaches us the importance of resistor matching to maintain high CMRR.

In the schematic for the Flat Moving Coil Preamp we see the following matched components.

R4 and R5, the input bias resistors, are matched to equalize source loading.
R6 and R7, the collector loads, R8 and R9, R13 and R14 which set gain, and R19 and R20 which inject servo all need to be ratio-matched as pairs to maintain CMRR.

The big surprise was the role of C4 and C5.
Their match is super-critical for HF common mode rejection and it was found they play a huge role in CCIF IM reduction.

When all of the aforementioned components are carefully matched the 1 kHz IM at 61.5 dB gain and +10 dBu output is at the A/D-D/A loopback level of around -120 dBc.
The following FFT is the result of resistors being matched to within 0.05% and capacitors to within 0.3%.
(Note that the spur to the right of 1 kHz is environmental.)

Image

When C4 or C5 is mismatched by just 4.5% the IM increases by +20 dB to -100 dBc:

Image

(R8 and R9, 499Ω, can be made 0Ω with many op amps without a decrease in stability but they provide the option to "tune" HF CMRR when C4 and C5 are not matched.)

All the preamps I build will now have super-matched resistors and compensation caps and I recommend it to anyone building a Cohen-style front end.
emrr
Posts: 570
Joined: Sun Mar 02, 2008 10:27 pm
Location: NC, USA
Contact:

Re: Component Matching Minimizes IM Distortion in the Demrow-Cohen Balanced Front End

Post by emrr »

Wow, really significant!
Best,

Doug Williams
Electromagnetic Radiation Recorders
User avatar
terkio
Posts: 304
Joined: Thu Feb 07, 2019 3:17 am
Contact:

Re: Component Matching Minimizes IM Distortion in the Demrow-Cohen Balanced Front End

Post by terkio »

Very interesting.

What about transistor matching, distortion wise ?
User avatar
mediatechnology
Posts: 5443
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Component Matching Minimizes IM Distortion in the Demrow-Cohen Balanced Front End

Post by mediatechnology »

Hans sim'd transistor mis-match up to 10 mV Vbe difference with no increase in IM.
I don't recall him running sims for gross Hfe mis-match.

I still match them and recommend that others do as well since transistors that are Vbe-matched also have matched Hfe.
Matched Hfe produces a low input offset current, Ios.
User avatar
terkio
Posts: 304
Joined: Thu Feb 07, 2019 3:17 am
Contact:

Re: Component Matching Minimizes IM Distortion in the Demrow-Cohen Balanced Front End

Post by terkio »

I imagine there could be more subtle transistor parameters asking for matching.
Early voltage, capacitance,
User avatar
mediatechnology
Posts: 5443
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Component Matching Minimizes IM Distortion in the Demrow-Cohen Balanced Front End

Post by mediatechnology »

There are though they don't seem to manifest enough to affect IMD2 or they're inherently matched when Vbe-matched as Hfe seems to be.
I just built 6 of them (12 channels) and they were very consistent.

I should scan Graeme's discussion of CMRR in Op Amp Applications where he discusses all the various CM errors.
IIRC he gets into HF models.

When the Cfb are not matched (C4/C5) one of the series resistors (R8/R9) can be "fudged" to make the time constants equal.
The larger C of the two has its' series R lowered.
When doing this it's good to know the loopback IMD2 of the A/D or one can adjust R8/R9 improperly by creating distortion in the DUT which cancels the A/D's IMD2.
Post Reply