My VAS performance "NEVER LET ME DOWN"

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Ontoaba ...

May be related with DC transient, It is difficult to simulate, my free tina7 user defined signal ability is blocked by free demo version, I can only use PWL macro generator and typing sine step signal in it also impossible.

It is important to use TSTART in the .tran analysis to discard the start up transient which may lasts several usec, depending on the reactive elements within devices and circuit.

Cheers
Arturo
 
Yes, you right. I am using 1mS at that sims. and 2mS simple VAS doing better. The first suspect is capacitive characteristics. I removing compensating capacitor from the VAS, and that flat spectrum distortion is slightly reduced. I remember with my first good sounding VAS that has 100mA bias, high current may reducing its effect. Capacitive characteristics is reasonable than "memory". But that simple VAS has similar bandwidth than mine. This reason need another suspect.
 

Attachments

  • BW.PNG
    BW.PNG
    5.6 KB · Views: 293
Last edited:
I place again the 300pf configuration. This configuration make it better in single tone but worse in multi tone, Flat spectrum is still exist in 100uS (seems that I am using 100uS before, it is close, while as I remember it is 1mS). Without 300pf, simple VAS need 5uS to reach around 10uV flat noise, and my VAS 2uS. When continuing to 4tone, simple VAS need 12uS, my VAS still 2uS. In 5 tone simple VAS noise appeared then dissapeared from big to around 2uV until 80uS, my VAS not. This another things happen.
 
Oh no. It is up to 120uS and continued until milisecond scale. And I also found in my VAS but much smaller (reach 1uV). Simple VAS still reach 5uV. 5uV when integrated is big.
Any clue what this is? I need a way to remove this things. This spectrum appeared then disappeared then appeared again and again until reaching milisecond.
 

Attachments

  • up to 120uS.PNG
    up to 120uS.PNG
    22.9 KB · Views: 304
I am removing all capacitors, no effect, then using different windowing and highest sampling FFT also not help.
It is not capacitive problem, nor FFT problem. Inter modulation? No, nor harmonic distortion it is all freq (like square) and happened also at below 500Hz.
I do check the sum of input generator signal FFT and still undistorted and noise less.
It is pure from VAS faillure. Probably this one that makes the sound differences. Since my VAS has lower in it.
Memory distortion? again may be yes. Is anyone here familiar with memory distortion? how its mechanism?
Is this one that make tubes better in voltage amplifying?. And also people like hbc not trusting transistor to do this job?
Anyone could help me what is this noise?
 
Easier and better, why not?.
It is very difficult one to solve. Better I am using listening in this case, yes easier and better. And about cascoding, I am still using my previous result, That it really increasing linearity both in steady and transient condition. And without cascoding it will less linear caused by its varied Vce, happened both in its steady and transient condition.
When feedback applied, everything become too complex, and listening taking over it.
The two transient characteristics that I got when feedback applied now are capacitive and an unknown.

Note: Transient characteristics are dominating in multi tone test. A crucial thing.
 
Last edited:
I have made a rather different VAS stage.. that has a quite unigue way of suppressing higher order harmonics's....this is an FFT.. with three tones...1 KHz 10KHz and 100 KHz....as frequency rises the circuit starts to develop a little 2. harmonic distortion -90-100 dB's..and that pretty much it...It also seems quite robust to the devices selected...

the circuit attached shows the concept..the one tested is a little more refined...:)

michael
 

Attachments

  • 1,10,100 Khz FFt.pdf
    69.1 KB · Views: 138
  • pre-amplifer.pdf
    28.9 KB · Views: 171
Last edited:
Good PreAmp MIIB, the VAS stage similar with my new VAS, but I split the resistor and signal connected in center, while the input stage using current feedback, with compensator R-C series connected to -inv and +ninv input.
Hi, MIIB. Isn't C5 better parallel with R10, but that pic is already very good result.

I have found that good 3 tone test is 1kHz, 4kHz and 5kHz. And 5 tone test is 1kHz, 4kHz, 4.5kHz and 5kHz. This test is also useful to choose which one is the correct compensator value, I found that my compensator value better 180pf, not 150pf, yes, easy tweaking without listening.
Attached is an example that shows what TMC do in multi tone, and selecting wrong compensator capacitor value also do the same, also happened when compensating capacitor has more than 500ohm (I try it with series resistor and noise showing up). The noise caused by TMC is higher in 3 tone test, 250nV. The things that I concern is the noise, because it is flat.
 

Attachments

  • 4 tones.PNG
    4 tones.PNG
    34.5 KB · Views: 271
Same concept...but extended to be a complete power amplifier. A three tone FFT...and app +/- 60 V output

still 1 K 10 K and 100 K input.. three generators stacked...


Harmonic Frequency Fourier Normalized Phase Normalized
Number [Hz] Component Component [degree] Phase [deg]
1 1.000e+03 1.855e+01 1.000e+00 -0.05° 0.00°
2 2.000e+03 1.433e-04 7.722e-06 -89.66° -89.61°
3 3.000e+03 3.310e-05 1.784e-07 52.52° 52.57°
4 4.000e+03 1.038e-05 5.593e-07 -131.93° -131.88°
5 5.000e+03 1.977e-05 1.066e-07 -143.20° -143.16°
6 6.000e+03 2.207e-05 1.190e-07 -80.12° -80.07°
7 7.000e+03 1.140e-05 6.143e-07 -60.39° -60.34°
8 8.000e+03 1.343e-04 7.237e-07 43.52° 43.56°
9 9.000e+03 2.861e-04 1.542e-05 104.33° 104.37°
Total Harmonic Distortion: 0.001887%
 

Attachments

  • output.pdf
    41 KB · Views: 137
  • 1,10,100 Khz FFt-1.pdf
    71.6 KB · Views: 85
Yes, that CE capacitor is better if removed, and if it need compensator, you could try at other place like between input output in J2. Ofcourse without compensator and keep in low NFB is better, since all local V->I and I->V transfer has good linearity and transient response (if noise from my previous post exist) is minimum.

I am also try 1k-10k-50k-100k, I have problem with 1k-10k-100k, because my sim only support 64 number harmonics, so I using 2k-10k-50k-100k to show spectrum around 100k.
I couldn't found JFET advantage here, but I ever built a mixer with tl071 and c4570, and TL071 has better in clarity than my mixer that using c4570, at their datasheet, technically c4570 much superior than tl071.

In this test I also couldn't find TMC creating noise. From the spectrum previous post, the noise looks like small square spike that may repeated when transient condition restarted due to random signal (like crossover dist, but much smaller and more often repeated by random, instead of crossing).
 

Attachments

  • 1k-10k-50k-100k.PNG
    1k-10k-50k-100k.PNG
    51.2 KB · Views: 395
This is other PCB that usable for VA270 with 2sc3355(BEC) and bc557(CBE) without swap its legs. Just add 150pf or 180pf Cc and it will be the same. It using all 2sc3355 replacing 2sc1815 and VA270 also work with it(all 2sc3355).

VA290 has much higher bandwidth, but cable feedback easily happen. It need good input-output cable separation and need additional input HF filter. Not recommended for DIY, except PCB layout, it is usable for VA270 with Cc added on PCB back side.
 

Attachments

  • VA290 (faster version of VA270).PNG
    VA290 (faster version of VA270).PNG
    21.7 KB · Views: 338
Nice to hear that, nelson ,you also comparing quality by its sounds .
It is long story. I am using some different input stages and VAS in the past.
My first version is simple differential pair with 100mA bias VA to drive single EF directly. This one is my first good sounding amplifier. Then some people very interested in my amps. But the later version is worse with the same EF, this is first time I realize that VAS has the differences too, not just same function to amplify the voltages.
Until I have this VAS, and 2SC3355 or SS9018H/I really help in stability and its performance. Before using them, I have many compensator circuitry to prevent oscillation, and that 150pf is 470pf that time.

I have using cascoded VAS before it because my amps need high voltages, and also using video CRT transistors, some people says that video transistor not good for audio, but I found it nothing wrong.

I am using cfp input when start my V+I amplifiers. That I think it need more linear input stages. I cascoding it to get constant Vce at around 7V, and I see some amplifier using this too. I am impressed in first test, and improving it until now.

If that my VAS really that good (as low memory distortion is true), may be my new VAS will not better than this.

Here is my example if anyone want to hear it. Beware may be it is lttle exciting:).

EDIT:
I forgot it, any changes from A872E to BC557B/C or C3355 to S9018H need some different TO-92 legs arrangement in PCB. 2SA/2SC has ECB and BC/SS CBE, see the datasheet for detail. IC hole matrix PCB is usable.

Hi Ontoaba
in LTSpice your circuit is oscillating using Cordell modells
or have I made some mistakes?
 

Attachments

  • Obraz1.png
    Obraz1.png
    124 KB · Views: 198
  • perf vas.asc
    10.6 KB · Views: 61
Q1 and Q2 need to be replaced with RF transistor (ft>800Mz) it work with 2sc3355 or S9018.
BTW I am no longer using it, because some simple VAS with low performance and much distortions are sounded better, don't know why:confused: I am still try to know what just happen, but two of my amps still using them (this VAS).
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.