SCRynth: Square Wave and Gate with 9 parts!

[Earthscum]

Well, here ya go folks. I'm  introducing ya to my "baby", the SCRynth.

I've been wanting to get this out for about a year now, but since it was my first unique circuit design, I've been afraid of the typical theft, but oh well. I decided to bite the bullet, and consider this my official publication of this circuit. Thanks to a short talk with R.G. I decided I can either hold onto it and it does nothing except for help me here and there, or I can share it and let others enjoy it as well. I figure the reward of sharing is worth it.



This is an ultra-simple square wave shaper with the added benefit of a sensitive signal gate effect. My original idea was to just drive a pair of SCR's, using the gate as a method for selecting my clipping level. I soon realized that Daniel Lamb had already done this in a TS circuit HERE.

That didn't dampen my spirits much. Searching and searching only brought up variations and revisions of Mr. Lamb's papers, I still had room to create something new and different. I soon realized that I had what essentially amounts to a Schmitt Trigger device, I just had to learn how to drive it.

I experimented with various gain stages to no avail. Some worked, some not so great. Then I figured out that if I match a PNP and an NPN, put the NPN on bottom, as in the circuit, I could get them to self bias at a level that just happened to be a decent bias point for the SCR.

How to drive an SCR:

What we want to use for audio is a "Sensitive Gate SCR". There are many types. Most commonly, they trigger at a gate threshold of somewhere in the range of 200uA @~0.8V. Check the datasheet. I've tried NTE5401, FOR3G, and what I've settled on: CR02

I'll use the CR02 data here. These are 100uA sensitive gate SCR's, available from Tayda. The datasheet states the following information that is useful to us:

On-State Voltage: 1.6V (other parimeters affect this a bit, but not so much to worry about. We are running 9V supplies, we are well covered on this parimeter)
*Gate Trigger Voltage: 0.8V
Gate non-Trigger Voltage: 0.2V (most SCR datasheets don't have this parimeter listed, a bit more on that in a moment)
*Gate Trigger Current: 100uA
*Holding Current: 3mA

The 3 with asterisks are the most important. If you have more than 3mA at the Anode of the SCR, it will not turn off (typically) when you remove the gate trigger.

There are different types of SCR's, some that can be turned off regardless of the current running through the device. These are not that type, so we have to cheat. By using a 10k resistor at the Anode, we limit the current to 0.9mA, well under the holding current. What this means is that the SCR will turn off as soon as the gate trigger is dropped to ground. Using a 10k at the top of the driving pair of transistors does several things: reduces the amount of work the PNP transistor has to do, forces the pair to bias at a diode drop above ground, and limits the current going to the gate of the SCR.

So, now we get to the Gate non-Trigger. This doesn't seem to be listed on most datasheets. This is another thing that brought me to use the driver configuration I did. When the signal goes positive, it causes the transistors turn each other completely on and off. I believe (someone who's good with this can definitely correct me) that the gain of the pair is basically PNP Gain X NPN Gain. Say they both measured at 225 hFE, that would make a total circuit gain of 62500? So, basically, when the NPN is in saturation mode, and the PNP is in cut-off, the gate of the SCR sees no current. Basically, this perimeter can be ignored, unless it's useful to your design req's. We are basically, at one point, causing the current to drop enough that the gate sees X amount of "voltage" and current.

The transistor pair give a really great square wave by themselves, but it's distorted. It's still just another representation of the original. I wanted a square wave that was completely separated from the original tone. I figured I could get an SCR to do this, basically turning a switch on and off to the flapping of your strings. (I play bass, they flap... don't argue!). 

The greatest thing about this is that it has a natural gate to it! If you use a JFet gain stage before this, you can increase the sensitivity a bit. Also, choosing higher gain transistors will increase sensitivity. The gating in this circuit is so much different feeling than any other gates I've tried. It's ultra sensitive, but since you get no appreciable original signal through to the output, it's ultra quiet. It just feels really natural. You don't have to really force it, it just does it. This circuit could easily be modified to run outside the signal chain to control a signal shunt or block to incorporate a gate effect into noisy ultra-gain pedals.

There are a ton of other things we can use SCR's for. I'll try to get write-ups done on some of my ideas, but for now, welcome the SCRynth, a base circuit with MANY possibilities.

[frequencycentral]

I decided I can either hold onto it and it does nothing except for help me here and there, or I can share it and let others enjoy it as well. I figure the reward of sharing is worth it.

+1 on that bro, thanks for sharing, looks really cute!

[Taylor]

Cool! I did a quick sim, can't get it working, but I know absolutely nothing about SCRs so if anybody wants to play around with the parameters for the SCR and maybe the transistors, we can get a more sensible result and see what's really happening inside the circuit.

[Earthscum]

I'm not quite sure how the sims work, but the SCR may be a normal one. Check the properties for the gate and holding current. If the holding current listed for the device is lower than the current actually supplied to it, it will latch. This is what makes it useful as a controlled clipper... it won't turn off until the voltage through it drops enough. As far as the trigger current, I'm guessing the sim isn't using a sensitive gate SCR. A standard SCR has an extremely difficult time working within the ranges of voltage and current we use. The gate takes much more current to trigger, and many of them trigger within the mA range (10mA, 12mA... even 80mA). Obviously it's gonna be hard to trigger that and keep juice in a battery.

[Taylor]

If you right click on the scr, go to edit, you can change the trigger current, holding current, and gate-cathode resistance. I upped the trigger current to 20mA and G-CR to 100k and it starts making more sense.

The output seems really small, in the microvolts. Also, with the 100n the output is highpassed so the waveform is not a square. 1uf gives a more square wave.

But the good thing about it is that it does seem to maintain 50/50 duty cycle regardless of dc bias at the input or an asymmetrical input.

[Earthscum]

For some reason I can never get that site to work, just says I don't have a Java enabled browser (I've tried a couple Java plugins... I think it's just a Linux thing).

I wish I had a couple probe sets for my new o-scope, I'd just show ya what's going on. I can visualize it, myself.  Basically, what's happening  is the output from the pair swings to V+ (minus .6V and the voltage division caused by the 'resistance' of the three devices). This passes the .8V that the SCR's gate needs to trigger. When the signal swings to ground (+0.6V), it cuts off all current to the gate of the SCR. Since we're keeping the actual device current well below the Holding Current, it will shut itself off. The voltage swing I observe on the output is around .2- ~V+. I don't quite understand why it drops to .2V, though.

To limit the output swing, I use an LED parallel to the SCR. When the SCR is conducting, the voltage is lower than the LED's forward voltage. When the SCR closes, the LED limits the voltage swing to 1.6V. I've also used series diodes to tailor the output swing.

*Correction. I just measured the output from the CR02's. These things are swinging from 4.5mV to 9V!!!

[Earthscum]

Wow... been a year, lol.

Here's a physical diagram that I made up that helped another member out.



So, now that I know a bit more of what is going on with this thing...

If you right click on the scr, go to edit, you can change the trigger current, holding current, and gate-cathode resistance. I upped the trigger current to 20mA and G-CR to 100k and it starts making more sense.

The output seems really small, in the microvolts. Also, with the 100n the output is highpassed so the waveform is not a square. 1uf gives a more square wave.

But the good thing about it is that it does seem to maintain 50/50 duty cycle regardless of dc bias at the input or an asymmetrical input.

The input of this is REALLY low impedance. The bigger the input cap, the less the signal gets lost. Not surprised, though. The output is severely affected by the HP filtering, for sure. When I scope it (yep... got probes!) the Anode, it is really square, but after the cap it starts to slope instead of keep square. Also, it's near perfect flat top DC, but after the cap/pot the harmonics start to show themselves, and the tops look like someone grabbed them at one corner and ripped them at an angle.

The CR02 is sensitive gate in that it has no internal G-K resistance. Typically, there is a G-K resistor that bleeds off a certain amount of current. It doesn't take much current at all to turn it on.

With 9V across a 10k resistor, the Anode only sees .9mA in it's On-State. According to the datasheet, well... the chart doesn't go down that far, it stops at 20mA and an on-state voltage of just over 1V, 100mA sets it just over 1.2V. If we ignore any kind of curve and look at it linear, since it's close enough, it would be about .7V at 2mA, .6V at .9mA. That's about par with the 4.5mV reading I got a year ago.  

[PRR]

But, but.... how does the SCR ever turn off?

> *Holding Current: 3mA
> If you have more than 3mA at the Anode of the SCR, it will not turn off (typically) when you remove the gate trigger.

Ah. I had not realized that this "mis-use" is "reliable". But it probably is.

> I believe ...that the gain of the pair --

Voltage Gain or Current Gain? A common-emitter transistor has both.

> is basically PNP Gain X NPN Gain. Say they both measured at 225 hFE, that would make a total circuit gain of 62500?

No. They are parallel, not cascaded. The current gain is basically 225.

Trigger current is 100uA. So input current must be 100uA/225 or about a half a microAmp.

Figure a guitar can put nearly a half-Volt into less than 1 Meg. That's a half a uA. Is in a ballpark.

The Voltage Gain is hard to say. It approximates the Mu of the transistor, or hOE/hIE, which is rarely cited because it is over 1,000 and usually other circuit parameters swamp it. Here they may not. So the Voltage Gain _could_ be many hundreds, even a thousand.

If turn-on takes 0.8V, turn-off takes 0.4V, 0.4V peak-peak sensitivity, and you have gain of 400, then it will fuzz down to 1mV, which is a very low level, far out on typical decay/sustain. And when signal drops below sensitivity, it may just "drop silent" instead of hissing like other excess-gain fuzzes. The transistors' self-hiss is probably not enough to trigger. You do need to be careful about hum and RF pickup being amplified and triggering.

What interests me is the Input Impedance. We need the bias current. It idles the upper emitter near 1.2V so the 10K resistor drops 9V-1.2V or 7.8V, so the current is 0.8mA. The hIE will be nearly 30 ohms. For the lower transistor this is multiplied by hFE, 225, for 6,750 ohms. For the upper transistor we have 10K+hIE times hFE or a couple Megs. So the input impedance is 6K or so. This is a good Power match for most of a e-guitar's bass and midrange, but a poor match for the higher impedance of guitar pickup in the high harmonics. That's probably desirable. Squaring the highest partials would just make atonal tizzz. A heavy load is a ton simpler/cheaper than some elaborate filter. (However this high-cut is lost if buffered.)

[Earthscum]

Thanks a bunch, Paul. Every assessment you made is exactly on par with the reaction I get with a guitar. If you dare drudge through part of the video, I posted a 14min clip of some of my pedals, and the Scrynth (plugged into the breadboard). I demo that right at the beginning, and then play with it through the wah and Nurse Quacky a bit later.

The decay is the best I've come across, honestly. With Bass, it will occasionally spit once or twice if the tone is turned up, or the string has a good amount of harmonics/phasing going on, but for the most part when it cuts, it's done. Basically a complete gate on itself. With schmitt triggers, I get the sputtering in and out because of obvious things like signal gain and how much low frequency you can/need to cut to keep a stable signal at the higher frequencies. The Scrynth doesn't "float", even when buffered.

When you buffer it, though, LP filtering is almost a must or it will start picking up every time you brush a string. I used a J201/10k buffer with a 2k2(470n)2k2 LP T-filter and it played pretty nice. With the ~6k impedance figure, that actually makes sense. I tried 10x impedance values, and I need to still turn down some gain (500k pot at the gate of the fet).

I think I halfway figured out the transistor gain thing the other day on my own when I realized that I had to split the signal current in half, and then I looked at the pair like resistors and realized maybe it should be treated like a voltage divider. I was kind of confused after that because I was figuring each transistor has a measured gain of 225... I just got it, I don't know how I was thinking half of 225. I would have 550/2=225 (each transistor amplifies half of the supplied signal). If that's correct, then if I had transistors with different gains, I would figure an average of 2 gains, right?

This whole circuit, honestly, is kind of a blind luck thing the more I look back on it.   Almost like I peed off the edge of the Grand Canyon and somehow wrote my name in the snow below.  

eta: I remember how I figured out the inverter gain thing the other day. As the NPN is closing, it reduces the amount of current available to the collector of the PNP, which is trying to open. They limit each other.

[PRR]

> each transistor has a measured gain of 225...

With reasonable current and collector voltage.

Here, because of the high Voltage gain, most of the time one or the other transistor runs near zero collector voltage; else at near-zero current. Starved! Transistor action comes apart. The effective current gain is, most of the time, very low. In a sim with hFE nominal 200 and 400, one or the other transistor has working hFE under 40, as low as 10.

For 800mV p-p input through 5K resistor, I see 8mV p-p at the bases. The PNP-NPN pair's input impedance is near 100 ohms!! (I can't get that result counting on thumbs; not sure what is really going on. Probably the base-collector junction becomes forward-biased.)

This suggests the 0.1uFd cap into them may be too small, cutting off low bass.

> kind of a blind luck thing

I think so too. It happened to work-out useful. When I try to "improve" it in sim, it quits working or gets real complicated real fast.