IV single stage inverting Feedback Amp

[ccschua]

Attached below is the IV output stage for the TDA1543. What I found lacking is clarity at high freq and PRAT.

The input voltage to the IV stage is 1.564V Peak to Peak.

I wonder if the calculation for closed loop gain is correct as below (assumed ideal amp, the error will be less than 5% right?)

The output is approx Acl = -Rf / Ra where Ra is the internal output resistance of tube amplification.

Ra = ra || R3

For 12AX7, ra = 62.5k

Ra = 62.5k||100k = 38.5k

Acl = -Rf/Ra = -47k/38.5k = -1.22

Therefore the output voltage is 1.564 * 1.22 = 1.9V peak to peak.

1. I have checked and verify the circuit again and I wonder why isnt the Grid resistor R5 be placed at point A?

2. If i remove the cathode cap, will I improve the dynamic response at the expense of loop gain and how much loop gain is sacrificed?

3. Is the Rload = R3 = 100k too low ?

 

[ccschua]

Or perhaps there should be a New Grid resistor of 470K connected to point A?

Aiken has mentioned removing Ck improves transient reponse, yes / no ?

 

[dsavitsk]

That's an awful lot of work for .336 extra volts, particularly when you can simply increase the supply voltage to the tda1543's slightly and increase the I/V resistors slightly to get the same, or more, voltage. I don't think Ck is your problem.

 

[DougL]

I ran the circuit through the Tube cad simulator.

Open loop gain is about 4.
Vbias = -.63
Voutmax -1/+5.85 V
Ra = 6 V

This circuit is not capable of 2 V rms output.

This looks like it was designed to add color, not provide low distortion output.

If you designed this, lets talk about design goals.

 

[kevinkr]

Try replacing R2 (R2 is also part of the load on the 12AX7A!!) with a 1mA CCS, change R4 to 100K and R9 to 20K and your loop gain will be high enough to actually close the loop and loading on the 12AX7A will be much less severe. A 12AX7A has an rp typically of 68K or higher depending on operating point and for good linearity and gain the load resistance (plate load resistor and external ac coupled load) should be several times higher than rp as a minimum.

This circuit lacks the ability to drive low impedance loads even with the changes recommended.

The other comment I would make is why do you have resistors in series with the dac outputs, I don't believe they need to be or should be there, their currents should just sum.

I will post one of my old designs later that actually worked pretty well, is very simple and requires no feedback to work - also provides a very low input impedance minimizing the effect of non zero voltage on dac linearity.

 

[kevinkr]

Here is the promised circuit. I used this in the early 1990s in my Sony CDP-790 with great results, the unit lives on with a friend and continues to work well.

R6 sets the I/V conversion ratio, with the value shown 4mApp results in almost 2Vrms at the output and that low input resistance makes it look a lot more like a transimpedance amplifier than your 680 ohm resistor and results in near ideal behavior with current output dacs. Connect your dac(s) to R6. (Loose those 100 ohm resistors.)

Effectively DC offset at the input of the parallel 12AX7A is negligible in comparison to the available cathode bias because of the 10 ohm I/V resistor (it's just the baseline offset of a couple of mV or so) so no coupling capacitor is required.

Linearity was very good, and it was extremely quiet too because of the increased transconductance of the two triodes in parallel and the fact that the noise generated by each triode is gaussian in nature so summing them results in 3dB less noise than produced by a single triode.

The CF output allows it to drive >50K loads without problems.

Performance is limited mainly by parts quality. Three 12AX7A are required for stereo. You can use a 5751 for the input pair of each channel if desired, but I recommend a 12AX7A for the CF due to filament insulation concerns, and the filaments should float about +60V above ground for best performance. Power supply should be very quiet, and for best results regulated.

 

[ccschua]

this circuit actually comes from the commercial cdp with the tweaks that is shown, after using exotic caps and using 4xtda1543, the company considers this as 'Reference' series. This is non other than the Oxxxx cdp, costs more than USD 1500.

I believe the tube output is there to add color. Can you post the output waveform of the tubecad simulation. I have been learning tube cad but to no avail.

 

[DougL]

I will post as much of the simulation as I can.

Doug

 

[ccschua]

That's an awful lot of work for .336 extra volts, particularly when you can simply increase the supply voltage to the tda1543's slightly and increase the I/V resistors slightly to get the same, or more, voltage. I don't think Ck is your problem.

The TDA 1543 unit is supplied by Vcc = 8V. I tried direct connection from IV resistor to the output cap and all I can say the tube flavor is there than without.

What I try to establish is where is the f 3db for high pass. By putting a local NFB R4/C3, the gain bandwidth product (GWP) has extend into inaudible region which would need a highpass. This is usually done with a highpass capacitor. I dont seem to see this highpass capacitor.

 

[ccschua]

Oh No. I made a big big mistake.

1. Further examination at the underside of PCB reveals that the cathode resistor is unbypassed (meaning no cathode capacitor, that means no issue with dynamic response?)

2. Rl = output load resistor = R3 = 220K (for lower frequency response)

Please comment if the following make sense.

Rewriting the gain equation.

A (Open load) = mu Ral / (rp + Ral + Rk (mu + 1))

where

mu = amplification factor = 100 for 12AX7

Ral = parallel resistance of Ra (=4.7k) and Rl (= 220k) = 4.6k

Rk = cathode resistor

A = 100 x 4.6 / (62.5 + 4.6 + (100+1)x0.47) => 4

From Aiken,

Acl = A*Rf / (Ri + Rf - Ri*A)

where

Rf = Feedback resistor, R4 = 47k

Ri = input resistor, R9 = 10k

Acl = 4 * 47/(10+47 - (-4)10) => 1.938

therefore the voltage swing (peak to peak)is

= 1.938 * 1.564 * 2 = 6V

V rms = 2.14V

1. Given the zero cathode grid bias is 2.15V, I wonder if the output voltage is distorted again ?

2. The output voltage depends on A, Rf, Ri

3. Since cathode bypass cap is not there, the advantage is lower distortion at the expense of lower gain, right no. The distortion in this context is phase distortion reduction ?

 

[ccschua]

Try replacing R2 (R2 is also part of the load on the 12AX7A!!) with a 1mA CCS, change R4 to 100K and R9 to 20K and your loop gain will be high enough to actually close the loop and loading on the 12AX7A will be much less severe. A 12AX7A has an rp typically of 68K or higher depending on operating point and for good linearity and gain the load resistance (plate load resistor and external ac coupled load) should be several times higher than rp as a minimum.

This circuit lacks the ability to drive low impedance loads even with the changes recommended.

The other comment I would make is why do you have resistors in series with the dac outputs, I don't believe they need to be or should be there, their currents should just sum.

I will post one of my old designs later that actually worked pretty well, is very simple and requires no feedback to work - also provides a very low input impedance minimizing the effect of non zero voltage on dac linearity.

Hi Kevin,

Noted your comment. In fact the recommended plate resistor with 2xra is commonly used. However in this case, I am not sure if the manufacturer intention is to 'emphasize' the tube warm character (with the 2nd harmonics?)

I am not sure of the DAC series 100R resistors. I guess it could be there to attenuate noise or protection ?

How do I provide CCS at R2 with 4.7mA using the PNP BJT ? This may seem more complex to implement.

Douglas has mentioned the clipping at the negative cycle. I will recheck into VB+.

I love the circuit provided. it keeps the IV voltage below the 25mV compliance.

 

[ccschua]

OH MY GOD, please accept my apologies.

after checking what Dougl mention, I called the owner and was told he replaced the original E88CC with ECC82. The marking on the tube resemble ECC83.

I need to jump into the river now. Before this, I downloaded the specs of the 6922 and the 12au7.

 

[DougL]

Here is the Tube Cad simulation for an ECC82.

It looks more reasonable. An active load would reduce distortion.
Substituting a 6FQ7/6CG7 or 6SN7 might be a good change, because these tend to have lower 3'rd order compared to a 12AU7.

I am very impressed with Kevin's 12AX7 I/V.

After you dry yourself off (river), we are still happy to help.

You should buy a copy of Morgan Jones Valve Amplifiers.

Doug

 

[DougL]

Here is the 12AX7 Simultion from last night.

 

[kevinkr]

OH MY GOD, please accept my apologies.

after checking what Dougl mention, I called the owner and was told he replaced the original E88CC with ECC82. The marking on the tube resemble ECC83.

I need to jump into the river now. Before this, I downloaded the specs of the 6922 and the 12au7.

Nah, I wouldn't recommend jumping into the river over this, and I would say it takes an honest man to admit such a mistake. FWIW the linearity of the ECC82/12AU7A is not going to be all that wonderful into a 4.7K plate resistor either, and the current required for reasonable linearity would be 5mA or greater so the recommendation for a CCS still stands.

That simple dac I/V circuit I posted works really well and should better any of the implementations discussed so far. I've used it in the past with a number of BB, AD and Philips current output dacs.

I believe, particularly with my circuit that all you need to do is connect both TDA1543 current outputs directly with no resistors to the input as discussed in that post.

I also strongly recommend Morgan Jones' book. (Valve Amplifiers 3rd Edition)

 

[ccschua]

Yes. the unit is running at around 5mA and I will try without series resistors .

I manage to replace the 12AU7 tube (Mullard BVA) with the original E88cc tube Tesla. At zero signal, the bias voltage is more or less the same. What I found is better clarity but lacking in bass. Isnt that E88cc is more linear and hence better clarity ?

 

[revintage]

If it could be changed to ECC88/6DJ8 things would be a lot easier to fix.

Hey ccschua,
So, what did I tell you two weeks ago ?

 

[ccschua]

Yes. the jury has pass the sentense.

Now I wonder how to improve on te clearity and dynamics of this tube output. Is by the active load and reducing the negative feedback ?

 

[ccschua]

deleted.