IGSL

[simonbaker]

Now comes the solution that does not change much in the existing design. Actually it will require some changes further on, but lets play with this for the time being.
First of, the D2 polarity is wrong. When reversed the output swing is allowed to go beyond the diode barrier voltage on positive side, while the previously offending saturation is kept well under 0.6V. That's a start.
Next I wish to enable as much gain as possible, while maintaining a log relationship at the output. To do that I reverse a diode in a first stage. To avoid floating with rising signal (this is a motion compensation design) I need to introduce an additional diode in counter-polarity to the existing one. I checked it with various offsets and it always works good. Now the response signal shape looks good, and the output swing is ~6V, with no saturation, no shifting left or right, no shrinking of active response, and the best of all - it is logarithmically compressed. See how octave increase maps into linear output?

I'd say that whoever designed this stage had an inspiration, but somewhere in a process he/she made a small mistake.

And a completely unimportant one: 100k in series with the first stage is doing nothing but increasing noise. At least with LF347. Just get rid of it.

Please feel free to play with these designs. I'd say they are pretty elaborate, but for some unexplainable quirk in a process of implementation they just got wrong.

Yes, I see, those are good points. Some of us talked about parallel opposite diodes to improve response time as well, but the effect on the pulse width is a good observation. However, ultra-strong signals are usually the least of our problems -- the current circuit is amply good with shallow targets.

I'm not sure why you don't want the parallel opposite diodes on the last stage also, unless you want to control audio volume with the signal. In fact, if you look at the original TGS circuit, the diode is omitted from the LM308 stage on the GB channel, and there is a "feed forward" resistor from the LM308 output to the audio stage, which I interpret as an attempt to provide some proportional audio volume response. Of course diodes are noise sources so the fewer the better -- but negligible on final stage I'd think.

On second thought -- my sim shows two diodes on the last stage is not as fat a pulse as one, so your diode arrangement looks better for shallow signals. For very, very deep targets, the diodes barely come into play.

I have always wondered why there is a 100K resistor going into the LM308 non-inverting input too - lightening arrestor ? The less parts the better.

-SB

[ivconic]

[Davor]

I'm not sure why you don't want the parallel opposite diodes on the last stage also, unless you want to control audio volume with the signal.

Yes, I want to control audio loudness in accordance with target response, I also want a straightforward control of the comparators triggering level, and I don't want negative swing to hit the lower rail and cause saturation problems, so one diode there is.
In a way loudness works as well with the original design, but due to the floating problem of an unsymmetrical signal supplied by the first stage it has a rather compressed response. Instead of piiiiIIIIIIIiiiiiip you get piiiiiiiiiip, and I really wish to get piiiiIIIIIIIiiiiiip.

[Aziz]

Hi all,

clipping (saturation of the signal path) is the most evil thing in analog hardware processing. Therefore I like the approach of Davor translating the exponential behaviour into the linear behaviour.

At least, the "log-amp" isn't much critical at this stage.
Aziz

[Aziz]

I happen to like analog anyway - so I'm thinking a digital model would be handy to decide what analog filter you want. Could actually use a laptop with sound card a la Aziz's experiments and just probe into the TGSL and do the tinkering in software.

But If I live long enough I'll probably try some digital version also, trying to use the lowest end, low power processor possible.

-SB

Hi SB,

why don't you use the latest sound card detector controller platform for this? You don't need the TGSL or any of the other platforms.

But you should be able to code with C/C++ and Win32 (SDK).
Cheers,
Aziz

[Davor]

Next issue would be power consumption. 60-70mA is way over i expected. Not bad but could be much better.

You can reduce power consumption right the way by replacing both LF347's with TL064's, or even LM324. In case of LM324 It will go a bit below zero at second output because of input bias currents, but TL064 will do just fine - as is. It will save you ~15mA of supply current. In case you have DIL sockets on your board, just try it and see.

As I said, a touch here and a touch there...

[Jerry]

Hi Davor, have been reading your comments with great interest. I have a TGSL that I use on a daily basis and I am in process of building the TGSL-EDU and also a TGSL of modular design which will be used mainly for experimentation. Your suggestions will be very easy to implement and I will give them a try as soon as I prove that everything is correct in the boards as I now have them.

Thank you,

Jerry

[Davor]

I'm expecting my PCB any time now. So in order to make it just right I'll try to optimise as much as possible, without changing much on the board. So far I'm just browsing through the design and noticing possible quirks. As you can see, the gain blocks were an easy one. just a child play with diodes. To wrap this up I'll need to see about the implications of this newly recovered amplitude, and will post my suggestions as they resolve.

[Davor]

It is terribly late, but I have to say that channel comparators are even more interesting to fix than the channel gain stages. There is a collection of oddly placed components, LM358 forced to work as a comparator ... total mess. OK, it works ... kind of.

More to come.

[golfnut]

Its around U8 I would like to see some changes.

I like the change in volume you get with target proximity on some dets,, even the surf pi does this a bit.

With IGSL all targets crossing the Threshold come in the same.


It would be good to have a log range - rather than binary.

Plus on Igsl often its impossible to advance the thrshold pot close enough to quiesent level to max sens to small objects.

A small target doesnt trip sound (if you scope pin 3 you could see a volt of detected signal with no notion of a Hit)

I wired my threshold pot between plus and minus rails to get the two inputs closer together for more sens to small targets. But yes I still dont get a range of volume ..


S

[ivconic]

You both are right; treshold and audio stage are weakest part of IGSL. More than any other stage; those need improving for sure.
Davor, using real comparator instead LM358 will not do.
Digital output will only switch transistor fast and shut it down even faster. That's what we don't want at audio driver.
LM358 is just temporary compromise... at least i thought that at the time!

[ivconic]

Come to think: comparator can be used, but will need more electronics around it. I had Musketeer solution on my mind, but than overall design will be even more complicated.
And i don't want exact copy of Musketeer audio. I planned vco's on both blocks later...

[Davor]

As I said, it is a total mess. I have something cooking to fix it for real, but for the time being I have a quick fix that will improve it a lot, at least in sense that audio does not lag behind the actual target detection. It is very simple: just get rid of all the capacitors around LM358 and the comparators.
Let me show you...

The first picture is a current state of the art (!?) and the corresponding signal plots. Observe how LM358 hits the rails. It is that bad. It is configured as a level shifter, but poor thing is just banging up and down.

The second picture is the same thing, but without all the offending capacitors. See how picture instantly beautifies. OK, it may look a bit more beautiful, but it is still a saturating crap. I'll fix it soon.

Also you'll find attached the corresponding artwork (!?) in LTspice. Again, It must have been completely OK before someone "improved" it. There are some misplaced components laying oddly here and there, and that is fixable - even without PCB rework.

[ivconic]

I guess you implied symmetrical supply there? (according to pics)
If yes; than it is not the case with initial IGSL setup.

[Davor]

It is +8V and -6.2V or just as per the schematic in ExpressSCH and elsewhere. I guess it is not +8V and 0V because it would not work well at all.