Detailed Modelling of a TA7122AP Equivalent Circuit

Detailed Modelling of a
TA7122AP/ECG1085
Equivalent Circuit

• Biasing of Q1 is achieved by linking Pin 5 to Pin 2, typically 100KΩ.
• Voltage-Series Feedback and hence accurate gain is achieved by
  linking a feedback network between Pin 6 and Pin 3.

TA7122/ECG1085 Re-drawn

The TA7122AP or ECG1085 Voltage Amplifier

• For this analysis, Voltage-Series feedback (in blue) is not added to the circuit, but is designed to be employed later to generate an accurate voltage gain at Vo.
  
  
• To the left of Vin is the signal generator, the signal generator's voltage source is Vs, or for this article 𝚫Vo
  
  
• Some circuit component values differ slightly from the original specification, this will be considered in calculations.
  
  
• Parasitic capacitances will be ignored since we are only concerned with audio frequencies.

 

As promised in my last article, a more detailed analysis of the Toshiba TA7122 voltage amplifier now follows. This is not intended to be an electronics tutorial, but more of an attempt to model this old and interesting amplifier, and compare the analysis with actual measured results. Bear in mind both theoretical and actual measurements will have some inherent errors - the latter especially so since readings will be taken from an oscilloscope screen. Nevertheless, a general comparison can be made between theoretical and (near) actual.

I cannot guarantee that the derivation of an expression here is free of errors, but I am reasonably confident of much of the work. Some simplifications were added at the T3 side of the amplifier, where no reference is made to hie, ri, and hfe on that final stage.

It is important to represent the T1 input side of the amplifier accurately, since this is where the bulk of the gain is generated.

I may offer later, a detailed explanation of my thinking as I attempted to mathematically model the TA7122AP IC.

So then, in summary - the purpose of the work below is to establish the voltage gain of the said amplifier as it stands here; that is without Voltage-Series negative feedback that would be later inserted between Vo and the emitter of T1.

The Derivation of Voltage Gain

I will use the delta symbol '𝚫' to highlight what we seek to find.

A prediction of voltage gain 𝚫Vo/𝚫Vs without Voltage-Series feedback now follows.

Much of this is written from an electrical and electronic engineering perspective, and so assumptions of reader knowledge is presumed.

There will be no proofs of Kirchoff's voltage and current laws as this too lengthy to write out. 

Hand-written 'scribbles' and small alterations are included, but this is because this was a 'first run' at collecting all the ideas together. Word processing the work below on a computer would be fraught with difficulties as many readers will know - circuit drawing, finding the correct symbols, lengthy equation writing etc.

 

G1 is the voltage gain for stage 1
G3 could have been written for stage 3, but wasn't.

Setting RL1 = 470KΩ; any higher and the gain
would cause 𝚫Vo to saturate too easily.
With my limited equipment, it can be
difficult to generate a very low input signal 𝚫Vs which
is stable at around 2mV and less.

 

Feedback Network: Above I 'referred' the feedback circuit RF into the input side with an equivalent Thevenin circuit - as if it were all sourced at the input. This then became just one equivalent signal source (but mathematically mixed) with an equivalent input resistance.

 

No additional Thevenin circuit was referred to the T3 side in parallel with Re3, simply because it would have very little effect on calculations. This holds true as Re3 << Rf.

Below should read 'So far then, EQ 6 yields ...'

• K1 and K2 were dimensionless definitions
  that made transposition of formulae
  and computation easier.

Labeling 𝚫VRL2 as 𝚫Vo we have

Matching Theory and Practice
 

• Signal Generator Test Frequencies: 50Hz, 100Hz or 1,000Hz.
• Vcc = 30 volts.
• 𝚫Vs was adjusted so that 𝚫VR2 or 𝚫Vo did not saturate, saturation began a little higher than 25v peak to peak.
• Signal source 𝚫Vs was measured 'offline', ie not connected to the input during 𝚫Vo measurements. Thus effectively creating an accurately measured voltage source.
• Rs is the resistance of the source, and was measured at around 600Ω.
  
  The theoretical model above and its computations were performed on a Spreadsheet, with measured comparisons also included, we can finally compare.
  

Click on this image to see the results.

Increasing Amplifier Gain

To obtain easy control over an effective minute input signal,
a simple potential divider circuit was set up.

A claimed '90dB' gain is quoted for this amplifier if limits of its ability are exploited. With that I decided to push the gain of this near equivalent circuit to as high as I could measure, at least for now. Feedback was reduced, to accomplish this RF was increased from 100KΩ to 1MΩ! This effectively reduces feedback current into the emitter circuit of T1.

 

I needed to set up an additional circuit at the input to create a controllable, but minute driving voltage with my old equipment.

 

With Re1 remaining at 0.22KΩ, and utilizing my old (but calibrated) Hameg 20Mhz oscilloscope to measure the outcome, output was pushed to 10v peak-to-peak, and an effective but new input Vs was measured at a little under 0.25*0.00465 peak-to-peak or 1.1625v p2p. The resultant measured gain was then approximately 20·LOG(10/0.0011625) or 78.7dB.

So how did the mathematical modelling compare? With Rs now being effectively replaced by a Thevenin 0.468KΩ, the spreadsheet computation predicted a gain 9348 with RF = 1MΩ, this equates to 20·LOG(9348) or 79.4dB.

Circuit Issues

 

The amplifier circuit wasn't without problems - with the addition of a 1MΩ feedback resistance, the amplifier's 'switch on' time was delayed, and more so if RF was increased further. 

 

The reason being the base-emitter bias voltage at transistor T1 took more time to reach the Vbe 'cut in' voltage at approximately 0.55v - 0.60volts. What creates this scenario is an effective increased Rf⨉Cbe charging capacitor time constant at the input to T1. With Rf at 1MΩ and transistor base-emitter junction capacitance Cbe (at say ~10pF), we have an increased time constant by a factor of 10. This of course means it will take about ten times as long for T1 to switch on. There may be other factors too why 'switch on' was delayed?

 

To overcome this, use a lower feedback value - recommended 100KΩ or less.

Comment

 

Predicted voltage gain modelling results with actual were good to excellent.

 

Analysis shows that altering hfe, ri (hie), and hfe within the developed equations does not effect the outcomes significantly, and indeed this was verified by the spreadsheet program and so possible variations can largely be ignored.

 

However, dynamic output resistance (1/hoe) or ro can and does effect results. Variations in Re1 also had a strong impact on gain as expected, since this also forms part of a internal current-series negative feedback loop.

 

Regardless of the pleasing accuracy of the mathematical model, ultimately the voltage gain of the TA7122AP/ECG1085 (and their emulation above) will be determined solely by the feedback network from Vo (pin 6) to the emitter of T1 (pin 3). 

 

For official configuration suggestions, see the ECG1085 datasheet online.

------------------------------------------------------------------------

Note: Corrections and modifications to the above article may be periodically made without notice.

29/12/2023

cassettedeckman@gmail.com

Modelling of the TA7122/ECG1085

Modelling of the 1970s
TA7122/ECG1085
Three Stage Voltage Amplifier

Actual circuit is to the right of Vin.

Left hand side represents a signal generator.

The Toshiba TP7122 (and ECG1085?) voltage amplifier was commonly used by Sony and others in their cassette decks and reel to reel machines of the 1970s. It was likely used in many other applications. It also resembles Sanyo’s own LA3122.

The TA7122AP or ECG1085 may indeed be difficult to source today, which is why I decided to investigate building an equivalent, effective configurable voltage amplifier.

The above amplifier comprises of three stages:

• Stage 1 T1 – a high gain voltage amplifier,
• Stage 2 T2 – a buffer, unity gain emitter-follower, and finally
• Stage 3 T3 – a final high gain (but less then Stage 1) voltage amplifier.

The pins as labelled by ECG Semiconductors form part of the powering, and user defined network and feedback system.

Current Shunt feedback from pins 5 to pin 2 help stabilize the amplifier’s gain. Additionally, both T1 and T3 circuits are also configured in current-series feedback topologies. The feedback values of Re in both T1 and T3 stages are small, although I suspect all three Re resistances in each of the emitters are mainly employed to prevent thermal runaway?

To further stabilise the amplifier and thus set a usable gain and bandwidth, voltage-series feedback network from pin 6 to pin 3 is required; marked in blue.


Analysis

Circuit analysis can often be difficult, and so the model offered here is slightly compromised. That is – it is assumed that the operation of this amplifier is restricted to audio frequencies, and no analysis has been made regarding the effects of parasitic capacitances and inductances, resulting in possible, but unlikely self oscillation. Often we see small ceramic capacitors of several pF placed strategically to avoid full phase inversion feeding back to the input – hence self oscillation. 

Stage1 T1: Very High Gain Voltage Amplification

With reference to the circuit above, if we consider just the first stage, and without any feedback from pin 5 to pin 2 (but independently biased), we observe a single stage amplifier with current-series feedback in the emitter leg.

Neglecting parasitic capacitances the voltage gain for Stage 1 can be approximated with reference to the (ac only) small signal h-parameter model ...

Simplified Hybrid Parameter Equivalent
Modelling Only ac Signals.

hie = dynamic input resistance, labelled as ri
1/hoe = dynamic output resistance, labelled as 'ro' below.

Considering the effects of hie, and on the application of Kirchoff's voltage and current laws we arrive at the following Vo vs Vin model -

or alternatively, if the effects of ri and Re1 are ignored (in the left side of the denominator) a simplified model will be ...

In a low loaded state, we can expect the gain to be very high.

Observing the ECG1085 data sheet, several examples are given. With R_L1 commonly set to R_L1=820KΩ, R_E1=0.220KΩ, T1 output dynamic resistance (usually labelled as 1/h_oe, now labelled here as ‘ro’) could be anywhere between 20KΩ and 300KΩ, and so we can expect the gain to be of the order -

(a) ro = 20KΩ

Av = 820*20/(820+20)*(1/0.22) = 88

(b) ro = 300KΩ

Av = 820*300/(820+300)*(1/0.22) = 998

I expect the first stage transistor to be a sensitive high current gain (high h_fe) type, similar to that of the modern and popular KSC1845-FTA.  With such a dramatically variable voltage gain, there will be consequences – the main being saturation and bandwidth. However, more negative feedback will desensitise the amplifier to a more stable manageable system as we will see later.

 

Stage 1: T1 Circuit in Isolation

 

 

This circuit was replicated as indicated in the main figure above, and so was independently configured and biased, with the input test frequency set to just 50Hz.

For any amplifier there is always a gain × bandwidth constant. Since the gain was anticipated to be so high at this stage, clearly the bandwidth was going to be severely compromised. Even at just 500Hz, the gain of this single stage was dropping, so it was decided to set the testing at 50Hz and under, where gain was close to its maximum.

Without presenting biasing data here, setting up the circuit I have chosen RL1=470KΩ (not 820KΩ), and with RE1=0.22KΩ (as before), the voltage gain for this single stage was determined experimentally.

Observing traces on a 10MΩ-configured-input oscilloscope as accurately as possible, I arbitrarily set the input and got: V_RL1=-1.45v, and V_in=2.75mV (not V_S), the computed gain was - 

Av = -1.45v/0.00275 = -527

Again, compare to the theory where Av = -RL1*ro/(RL1+ro)*(1/Re1), then for several estimations of ro, 

 

• ro=50KΩ, Av ~ 205 

• ro=100KΩ, Av ~ 375 

• ro=150KΩ, Av ~ 517 

• ro=200KΩ, Av ~ 638 

 

Then under these biasing conditions, ro (1/hoe) was estimated at around 150kΩ.

Stages 1,2,3: Full Circuit Without External Feedback Network

This time, Vs is measured in an open-circuit state to avoid internal resistances (R_S ~600Ω) interfering with the results. Again, using the oscilloscope visually to compute A_V,

V_S = 7.4mV, and V_RE3 = 20.8v

(peak-to-peak measurements on both accounts)

gives

Av = 2810

This compares with a theoretical approximation of

A_V(CL) = (R_F/R_E)*(R_L3/R_S) = 3833.

(The expression above was derived by modelling the circuit where R_E1=0. Modelling with R_E1 >0 is more complex, I may later write an article on how both were derived!)

Of interest, R_E1 was later short-circuited, and so the small amount of feedback offered by R_E1 was theoretically zeroed, here the gain was examined: V_S = 7.2mV, and V_RE3 = 22v, and so

Av = 3056

A comparison of errors based on the mathematical model were the theoretical was taken as the reference, then for R_E1=0.22KΩ, and R_E1=0KΩ.

R_E1=0.22KΩ, error = |3833-2810|/3833 ~ 26.7%

R_E1=0Ω, error = |3833-3056|/3833 ~ 20.3%.

So, as a stand-alone amplifier, in this present form we can expect a large gain of somewhere between 2500 and 4000, but still with a restricted bandwidth. However, further negative feedback will desensitise the amplifier, and impresses gain and bandwidth stability.

Full Circuit with Final Feedback Network β

According to feedback theory,

A_VCL = A_OL/(1+A_OL·β),
(A_VCL: closed loop gain,  A_OL :open loop gain)

and if A_OL·β>>1 then

A_VCL ~ 1/β,

where β is a feedback ratio (β≡V_E1/Vo) and can be shown via potential divider action to be 

β = R_E1/(R_F2+R_E1)

Applying this to the T1/T2/T3 full circuit -
 

A_VCL= 1/β = (R_F2+R_E1)/R_E1

To drive the amplifier into a stable amplification environment, and much broader bandwidth, we can arbitrarily set R_F2 = 10KΩ, and with R_E1=0.22KΩ,

A_VCL= (10+0.22)/0.22 = 46.5

Comparing this to actual measurement, V_O = 4v (p2p) with V_S = 91.6mV (p2p), we have -

A_VCL = 4/0.0916 = 43.7

Square Wave Response

As a quick indication of the amplifier's stability, the input was subjected square wave excitation at both 1Khz, and 10Khz with a 10KΩ load across Vo. 

Prior to this, I did notice that the amplifier had a very good frequency response, but there was an obvious, but small rise in output at around 100Khz-200Khz. Was this the signal generator?

Recalling amplifier expression A_OL/(1+A_OL·β) again, a rise in output can be attributed to the denominator exhibiting some pronounced phase shifting at high frequencies. The gradual but apparent higher frequency phase shifting suggests that the term |1+A_OL·β| is behaving as |1-A_OL·β|, and thus lowering its value to increase overall gain. Should this amplifier have become very unstable (very unlikely), then we'd presume that the denominator term move closer to zero, ie |1-A_OL·β| → 0.
 
Shown below are traces of two square wave input signals - both at 10Khz. Note the small overshoot on the first, and then compare to the next when a compensating capacitor Cf is added.

 

Please note - The oscilloscope probe leads were compensation calibrated. Indeed, some of this overshoot can be found at the output of the signal generator - possibly a little internal line-capacitance/inductance, and part-signal reflection from generator to input of amplifier, ie impedance mismatch? (NB: this point has not been clarified yet)

On excitation of 10Khz input.
Observe a small amount of overshoot; the amplifier is relatively stable.

With the addition of a compensating 68pF ceramic capacitor across RL2 (Cf), the amplifier is further stabilised, or dampened. This inclusion of Cf is very effective as this revised feedback attenuates the 'lift' around 100Khz-200Khz.

Very stable amplification,
but possibly slightly over-dampened?

In the present setup, the amplifier with a Avcl gain of approximately 44, is delivering a -3dB-0dB-3dB bandwidth which is slightly better than 5Hz - 200Khz.

Amplifier Input and Output Impedance

Not yet measured, but according to feedback theory - for voltage series feedback, Zin increases by a factor of (1+A_OL·β), and Zout by a factor of 1/(1+A_OL·β), ie, a decrease.

Amplifier Distortion

Not yet measured.

*This article is subject to corrections and additions without notice. 18/11/2023

 

cassettedeckman@gmail.com