Bandwidth problem

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

JAY X

Well-known member
Joined
Jan 9, 2009
Messages
683
Hi,

I recently finished building  small  stereo analog processor (not an EQ). External Ac transformer, half wave rectifier psu, +/-17v One layer PCB, Singlepoint ground.. the usual practices. I also designed a metal case to fit the boards, etc... In a normal listening test you wouldn't notice there is a problem... All opamp stages (MC33078, DRV134),have a bandwidth around 100khz, each. But after testing the device with TRUE RTA, (20-20khz bandwidth) it revealed a bandwidth problem...

Many questions arise: ¿is it due to single layer layout? (pcb is flooded with gnd), Some told me  about parasitic capacitances...Maybe the bandwidth filters in the opamps are not correct? ¿ wrong caps values?

In professional audio products the bandwidth is specified within +/- 0.5db  in others  +/- 3db variation ( within ear range..)
¿what factors can contribute to limit the bandwidth of an audio circuit? ¿how can i trooble shoot this?

because all of this, i ask some help to point the cause of the problem.

Thank you for your help!

Jay x.



input test signal  level:  1khz , 775mv @ 0dbu

freq                  dbu

20.000       -3,700
50.397       -2,085
100.794         0.000
201.587         0.647
403.175         0.782
1.015.937 0.138
1.140.350  -0,037
2.031.873         -1,199
3.225.398         -2,470
4.063.747 -2,260
5.120.000  -2,210
6.450.796         -3,560
10.240.000 -3,250
16.254.987 -3,780
20.480.000 -4,660



 

Attachments

  • FREQUENCY RESPONSE CURVE.JPG
    FREQUENCY RESPONSE CURVE.JPG
    112.8 KB · Views: 18
Schematic of the circuit?

Are there transformers involved?

Have you tried measurements at different places of your circuit to find the place where it goes wrong?

Jakob E.
 
Hi!

No trafos involved, just opamps and line drivers. The RTA is calibrated, despite the interface is not as good as they sold to me: a behringer UMC1820. The inputs/outputs are impedance balanced. ASAP i will post a schematic of the stages involved.

Just 2 hours ago i bought some test probes to try and see where is the problem... I will tell you. :)

But generally, even in a one layer design, a nearly flat bandwidth between 20hz and 20khz if well designed, is expectable..¿right?

Jay x
 
You have already been asked to share a schematic.... because that matters more than how many sides your PCB has.  :eek:

Modern opamps will have gain bandwidth many times 20kHz but the devil is in the details. The transfer function is pretty much defined by the negative feedback networks, so share.

JR
 
JAY X said:
But generally, even in a one layer design, a nearly flat bandwidth between 20hz and 20khz if well designed, is expectable..¿right?

Jay x
Generally the PCB is not a limitation until you reach GHz region.

Cheers

Ian
 
Hi,

Here is a schematic..

The inputs are differential inverting because i wanted to allow balanced or unbalanced signals.

Jay x
 

Attachments

  • STEREO INSERT DEVICE.pdf
    200.4 KB · Views: 48
and here is an image capture from the output of the inverting differential amplifier...

The output level is low because the signal input was low....about 400mv RMS


JAy x
 

Attachments

  • INVERTING DIFF AMP OUTPUT.JPG
    INVERTING DIFF AMP OUTPUT.JPG
    148.6 KB · Views: 10
yup rectifiers in PS look wrong..

the differential input has two LPF rolloffs in + input path but only one LPF rolloff in - path which will degrade HF CMRR.

Output buffer with capacitance isolation can actually cause some overshoot or rising HF response before rolling off.

===

I didn't check it rigorously just giving you some drive-by observations

JR
 
If your schematic is correct your power supply is definitely not working. You may just be getting unpowered pass through the circuit. Check PSU wiring and power rails first.

You also have lots of HF roll offs around all the op amps. Most of the caps look too big to me. A few -3dB at 100Kz in a row will be visible at 20KHz.

Cheers

Ian
 
Ther's a total of 3 poles at 100kHz. that should account to about 0.5dB at 20kHz.
I suspect either the test rig or a gross mistake (it happened to me once; in a stereo decoder I used a ceramic cap marked 4.7N and wondered why the multiplex residues were too high; turned out it was a 4.7pF, not 4.7nF)
 
Hi!

I'm checking...there are some mistakes around the rectifier diodes...in my layout.  ::)

Ok..i see. The schematic is wrong!!! And i Will check again my layout...
 
JohnRoberts said:
the differential input has two LPF rolloffs in + input path but only one LPF rolloff in - path which will degrade HF CMRR.

JR

Hi JR,

I did not  understand It too well...¿The problem are the 150pf caps across both 10k?

This input amplifier is from the glenn ballou audio cyclopedia..
And the original circuit show 100pf/6k8 un in the + in leg, and 220pf/6k8 in the -in leg. And 150R output isolator resistor.

Jay x
 
On the other hand, i could eliminate the 10uf caps in the inverting fader amplifiers,as the pot is in between two DC blocking caps...

Or replace the mc33078 by Lm4562, with low input bias current, and eliminate the caps..

Jay x
 
JAY X said:
Hi JR,

I did not  understand It too well...¿The problem are the 150pf caps across both 10k?
For a differential amp to subtract out unwanted common mode signals deeply, they need to be the same... If one is LPF more than the other, this difference will not cancel...

JR
This input amplifier is from the glenn ballou audio cyclopedia..
And the original circuit show 100pf/6k8 un in the + in leg, and 220pf/6k8 in the -in leg. And 150R output isolator resistor.

Jay x
 
Hi!

Merry christmas!!

And the good news...: I think i corrected the frequency response problem. I replaced the 150pf cap at one side of the diff inputs for 68pf, and also replaced the 10uf cap at the the fader wiper for 47uf. (47uf//6k8).

I injected a 1v signal sweep, I lowered a bit the output level with the fader, and measured the output with the Pico scope : flat up to 20khz. Still have to do some more tests, but seems to be fixed.. :)

Jay x
 

Attachments

  • FREQ RESPONSE.JPG
    FREQ RESPONSE.JPG
    202.6 KB · Views: 29
Hi!

It is a log scale, as for noise level, but with different settings that turn It into a loop reading envelope...or so...i don't know how to explain It better...there is a video on YouTube about picoscope. 

Anyway, still i had to make some changes. I eliminated the 10uf at the fader wiper, and the freq response got more linear..

Jay x

 
JAY X said:
It is a log scale, as for noise level, but with different settings that turn It into a loop reading envelope..
It's definitely a linear X-scale on your pic.

there is a video on YouTube about picoscope. 
Yes, they say the scale can be log optionally. You should activate it for better readability.
 

Latest posts

Back
Top