Tweet
Think changing sample time with the MPP integrator doesn't work for GEB. Varying sample time changes the cutoff frequency not the gain of the integrator.
-
-
Hello Davor, do you still visit the forum ?Comment
-
I'm temporarily back. I'm busy. Thanks for asking.
I'll check again for posts churned by this visit soon.Comment
-
Good to see you back.Comment
-
Hi Davor
I've done a bit more work with micro's since I last looked at your method of ground balance
and I would like to revisit it, as I have long been interested in it, just got to remember where I was up to.
I've had another look at your simulation, and the pulse width of the main sample pulse looks to be about 27 micro seconds wide
I am not worrying about decimal points, but 27 uS looks close enough.
So, if the main sample pulse width is 27 uS, then the ground balance pulse width is 2x main sample width which is 2x27 uS = 54uS
and the delay between the GB pulse and the EF pulse is 4 x main sample width = 108uS, and the EF pulse width = main sample width which is 27uS.
Does this look right ?
Comment
-
Yes, that's about right. Tuning of this method is achieved by varying the delay to the first sample, which is not a micro-friendly approach. Knowing your appreciation for micros, you'd be better off by modifying the scheme that will shift the GB sample left or right. You may design various schemes following the same ratio rule. If you observe the scheme above, you'll notice that the ratio of pulse duration to its end time is exactly the same for both target and GB pulse.
In numbers, the target pulse duration is 27us and the end time is ~54us (+/- tuning variation), or roughly 1/2. For GB pulse it is 54us pulse duration to 108us (+/- tuning variation) or again 1/2.
You may as well go with other ratios like 1/3, but those may lead to ridiculously long GB pulses, and the approach with unity gain integrator may not be the optimal choice.
Unlike the usual PI with EF timing, here the EF is set quite tightly, which allows for higher repetition rates. If you go for a micro, you may make sure that EF is fired tightly prior to the Tx pulse.
Keep in mind that every GB scheme will need some sort of absolute value circuit prior to the audio indication circuitry.Comment
-
Hi Davor thanks for reply
I like numbers rather than ratios ,
with respect to the cd4093 circuit, the ground balance is achieved by varying R2 which will vary the delay to the first sample
and the time base is altered by R1, which will vary the width of the pulses
and a absolute value circuit is basicly a full wave rectifier is that correct.
Comment
-
-
@6666, yes, absolute value circuit is the same as precision full wave rectifier. That way the negative voltages also get their voice. Alternatively you could drive one one tone with positive values, and the other with negative, and have some sort of discrimination, but that's too much to do for little or no benefit.
@eclipse, I can see a rectifier in this schematic, but I did not analyse it in detail to see what exactly it does. First look says it does not perform absolute value function, but instead it combines pitch and loudness for more pronounced effect of positive going amplitudes. But it doesn't fix the negative values problem. Or does it?Comment
-
Originally posted by eclipse View Post
Thanks for the reminder on Micks circuits
he is a clever experimenter and he finds gold.Comment
-
Wondering if anyone has tried GEB by adjusting timing with the MPP integrators and had it work. I've measured ground decay slope between -1 and -1.5. For me, Tx on time having the greatest effect on slope. Including a chart of times I have used for GEB. Not saying what's best, now I adjust delay time and then adjust target sample time for GEB with delay time between target sample and GEB sample time and GEB sample time fixed.Originally posted by green View Post
I take a EF sample also.Comment
-
Just for fun I've rebuilt the CD4093 pulser, and it works, but had to change the value of C2 up to 1nF to get reliable triggering
you can easily see the first pulse , EF pulse, the 2 combined GB pulses, the pulse widths track with the time base pot ,
took a short video, I may upload that later.
Comment
-
Hi 6666
Made a simulator in Excel awhile back that matches my detector fairly close. Like to try it with your timings. What are your timings? target delay_______, target sample_______, GB delay_______, GB sample_______, EF delay_______, EF sample_______Comment
-
An example using some numbers from Davor's reply #36. He didn't include a EF delay so I picked one, shorter than needed to show how the EF sample effects data assuming the slope or TC stays constant. Integral D is ground with a slope of -1. The target sample and GEB sample are very close but the target sample and EF sample add so the output with GEB on reads the EF sample. I get a ground slope of -1.33 instead of -1 with my detector. I modified the Excel program a little from the one I posted awhile back if anyone is interested or maybe someone with more math could come up with something better.Originally posted by green View Post
The starting numbers aren't important, just the change.Comment
-
Originally posted by green View Post
Hi Green from your graph table it looks like you have the numbers ok
in the circuit there is two sets of timings happening, the first target sample delay, and the time base which alters the pulse widths
target delay_______, probably up to 27 uS, I have not put the CRO on that delay part of the circuit
target sample_______,27uS
GB delay_______, Zero
GB sample_______ 54uS 2x target Sample width
EF delay_______, 4x target sample width 108uS
EF sample_______ 27uS, same as target sample width
Video https://youtu.be/x1RGdsnqk_4Comment
Comment