Roaches: Rolling your own

[stratomaster]

This is regarding the optical tremolo units used in vintage Fender amps.  The preassembled units available from CE Dist and others units seem rather costly for what they are (bulb, ldr, heat shrink), and I tend to find them rather lackluster. This could admittedly be a placebo effect, or just unrealistic expectations.

I've come across a few threads of people using an LED in place of the bulbs and an assortment of LDRs.  I've tried an LED in combination with the vintage LDR with better results than a readily available CdS LDR like the GL5537. It suffered from a choppy waveform and a volume drop at higher speeds. Going with a 5528 could give a deeper trem, but the dark resistance number isn't high enough to retain volume when the light goes dim.

A typical speed range is say 2hz to 10 hz.  At the higher end that means a rise/fall event happening every 50ms or so. A review of the spec sheets shows the response time of the 5537 LDR is ~30ms.  I believe this to be part of the problem with tracking the input and not recovering fast enough to get full volume while the light source swings dim/off. 

Additionally, the LED has a voltage threshold that acts as an abrupt on/off not present with the original bulb. Modern CdS LDRs are also most sensitive to the green range of the visual spectrum. (Note: I used a green LED in my home brew roaches for this reason).

All that to say, has anyone found a good combination that performs as well or better than the stock combination in vintage Fender amps? Are there specific parameters of an LDR to select for?  NE-2H bulbs are available in green--so perhaps that is part of the recipe?

PS. There was a shop here in Austin that claimed to have the magic combination "the surf bug", but then the proprietor of that shop ended up passing away (allegedly of complications from getting shocked working on an amp). Don't know where I was going with that tidbit, but it seemed somewhat related.

[tubeswell]

Subscribed

[stratomaster]

I'll add a few thoughts to this in hopes of sparking discussion.

The arrangement in Fender amps is the Intensity pot is a fixed 50k to ground at all times.

The LDR is attached to the wiper of the pot and ground, and the pot rotation determines how much of the 50k load is in parallel with the LDR. At minimum Intensity the LDR is essentially out of the circuit, and at maximum Intensity the LDR is in parallel with the entire 50k load. Between these two extremes the load is split between a fixed resistance and the remaining fixed resistance in parallel with a fluctuating resistance.

A perfect LDR would have 0 resistance at maximum light input, and infinite resistance at minimum light input, and a near instant response time.  This is impractical, but due to the parallel nature of the circuit, the rule of 10 seems to be a logical place to start. I think it's reasonable to seek LDRs that at least fall to 1/10 the Intensity pot value and rise to 10x during the swing in light intensity from the oscillator/driver.  Any lower than 10x will cause audible drops in volume when tremolo is active.

Unfortunately, the datasheets provide very little insight into how these will behave in this exact scenario. The light and dark resistance values given are after stabilizing under very specific conditions and dwell times. And even at that they are often ranges for the light resistance and just a minimum for the dark.

 Further complicating the matter is the response times and the dwell time used for the resistance specs are not related.  There is no indication of a relationship between response time and stabilization of resistance. (Is response time like a time constant in that 3-5 cycles are necessary to achieve stability?)  Add to this that response time to and from dark can be different, and now all bets are off as to the exact waveform that the LDR will trace out.  It's liable to be clipped at higher oscillator speeds, and asymmetric due to the difference in rise/fall response times. Plus the peaks are unknown.

So what does that mean for component selection?  I don't really know, but I think I want an LDR that has a maximum light resistance spec of 5kΩ, and a response time 1/2 to 1/4 of the minimum time between rise/fall events, and a matching light/dark, dark/light response time. I don't think there's much value in dark resistances much higher than 500k at minimum light input from the bulb, but I have no way of discerning whether a given LDR will reach that value in this circuit based on datasheet specs alone.

[shooter]

but I have no way of discerning whether a given LDR will reach that value in this circuit based on datasheet specs alone.

and that's why breadboards were designed 


being from the Tesla school of engineering, I do enough math to get started, then I experiment; meters, scopes, seg-gens, spectrum analyzers are far better than theory n calculators

[astronomicum]

but I have no way of discerning whether a given LDR will reach that value in this circuit based on datasheet specs alone.

and that's why breadboards were designed 

I breadboarded a trem circuit back in 2023 and tested 10 of each of 7 LDR models as well as the commercially available VTL5C1. I also looked at the effects of red vs green LEDs. I ran the trem at about 4 and 8Hz and set my meter to look at min ohms, max ohms, and average ohms of the LDRs at each speed. You can really see the performance differences. And there is a marked difference between the min max ranges of the two speeds. I also found that ambient shop light has a large effect on results (on completed roaches) and had to shield the roaches from light during testing. I will spend some time in the next few days to put together a summary of my results.

The key I found to keeping a nice smooth waveform and keeping noise out of the circuit is to design an LED drive circuit that allows you to idle the LED just to it's "on" state using a trimmer to adjust the idle voltage. I do that by trimming the voltage until I see the LDR dark resistance come down to about 1-2M.

[stratomaster]

The key I found to keeping a nice smooth waveform and keeping noise out of the circuit is to design an LED drive circuit that allows you to idle the LED just to it's "on" state using a trimmer to adjust the idle voltage. I do that by trimming the voltage until I see the LDR dark resistance come down to about 1-2M.

That's very clever. I figured the voltage threshold added a step change to the waveform and an element of photodiode behavior.  Do you leave the driver triode mostly intact, or is there significant modification with your design?

This sounds like a lot of work, so thanks for sharing your results.

With respect to light pollution, I found double sleeving the makeshift couplers with the outer sleeve being adhesive heatshrink that you pinch shut while still warm to be rather effective.

[Willabe]

Doug, our host, came up with a good sounding Fender roach. He tried many different combinations until he found a good sounding roach.

I don't remember if he did this because he didn't like the sound of the common replacements or if they were getting hard to find?

I think these are still the same 1's, Doug's store;

https://hoffmanamps.com/MyStore/perlshop.cgi?action=template&thispage=Lamps&ORDER_ID=705235944

And I think Mojo now sells them?

[Willabe]

I don't know that you can get an LED to work/sound like a real incandescent bulb. The turn on/off envelope of the filament in the bulb is not just instant on/instant off.   

That's 1 of the reasons a Uni-vibe sounds the way it does, the bulbs on/off envelope is asymmetrical. 

[stratomaster]

Doug, our host, came up with a good sounding Fender roach. He tried many different combinations until he found a good sounding roach.

I don't remember if he did this because he didn't like the sound of the common replacements or if they were getting hard to find?

I think these are still the same 1's, Doug's store;

https://hoffmanamps.com/MyStore/perlshop.cgi?action=template&thispage=Lamps&ORDER_ID=705235944

And I think Mojo now sells them?

I didn't realize the ones here at Hoffman were different from the CES/Tube Depot/TAD options.

[Willabe]

I think they are.

I just sent Doug a PM, we'll see what he say's.

[Willabe]

My mistake, it was NEW SENSOR he sold the recipe to.

The 1's Doug still sells he buys from them. And they might not be the same exact LDR anymore, Mouser might not sell that LDR anymore.

Doug said he went through 10 LDR's before he found 1 that sounded good to him.

[shooter]

steve's term-o-lator was the "start" for the last trem circuit I build, not sure I documented my changes through, mostly were increasing trem signal, IIRC the waveform was good.

[stratomaster]

steve's term-o-lator was the "start" for the last trem circuit I build, not sure I documented my changes through, mostly were increasing trem signal, IIRC the waveform was good.

I gave that circuit a try, but the lack of availability of the true vactrol that it was designed for put me on this path.  I ended up with an xvive repro that still left much to be desired.

I have a thread documenting the my follies with that circuit. I'll link it here once I find it.

Found it. But I think I was wrong about the effect of the plate resistor value.  Been too long to remember accurately. https://el34world.com/Forum/index.php?topic=30909.0

[Merlin]

Tbh I'm kinda surprised you've not had good results with the 5528. At <20Hz the reaction time really shouldn't be an issue, since we're dealing with sine modulation, not square. And an LED has a less abrupt threshold than a neon bulb!

[stratomaster]

Tbh I'm kinda surprised you've not had good results with the 5528. At <20Hz the reaction time really shouldn't be an issue, since we're dealing with sine modulation, not square. And an LED has a less abrupt threshold than a neon bulb!

I was surprised too, Wizard. Fender's published waveform isn't sinusoidal though.  I've yet to scope it for myself to confirm however. 

I think this is part of the problem.   

[Merlin]

I think the anode waveform is an outcome, rather than something you need to aim for. This is the waveform I got from that circuit, using a regular neon bulb

[Wardy]

So I recently completed a channel switching hybrid blackface / marshall combo based using a vibroverb as the base. I used the original roach and blackface circuit and wasn't at all impressed with the result. After a lot of hunting around and trial and error, I settled on SLuckeys Trem-o-nator circuit posted by shooter. I changed the oscillator caps from right to left with the values. 0.02, 0.02, 0.033. Also to make it work properly had to change the max speed resistor to 120k.

now sounds great.   

[astronomicum]

I took some time yesterday to put together a spread sheet of my LDR data.

The circuit used for testing consisted of two triodes: the first as the LFO, the second to drive the LED using the cathode voltage. The LFO output was attenuated to about a 6-7V Peak to Peak signal used to drive the driving triode. For testing, the driving triode was biased at approximately 1.5V, adjusted just at the Red diode On state. Minimum test speed was about 2.5Hz, and Maximum Speed was about 7.1Hz.
A Fluke 177 was set to read Minimum Ohms, then Maximum Ohms, and lastly, Average Ohms for each LDR and speed setting.  Testing was done with each roach protected from light since it was not sealed on the LDR end to facilitate LDR changes.

Testing was done on the GL5506, GL5516, GL5528, GL5537-1, GL5537-2, GL5539, GL5549; the Light and Darkness light and dark resistance specifications increasing with each model in the order listed. The commercially available VTL5C1 was also tested.

The document Roach Data 1 shows the raw data and data sorted by Maximum Ohms at Minimum Speed, and Roach Data 2 shows the raw data and data sorted by Minimum Ohms at Minimum speed.

General Observations:

•   It must be noted that the data collected is only representative of the specific parameters which the LDRs were tested, therefore changes in LFO speed, the driver bias, grid voltage, gain, and driving wave form, among others, will affect results. The data does, however, allow for a general comparison of dynamic performance among the various models.

•   Each model group does show performance groupings consistent with its increasing Light and Dark resistance specifications, but the data does show some wide variations. Some of these variations could be from supplier issues. Consider purchasing directly from the manufacturer.

•   The VTL5C1 data demonstrates similar performance to the GL5506 and GL5516 models.

I am still working on the Red and Green LED data.

Edit 250808: Corrected and Replaced Roach Data Files

[stratomaster]

I took some time yesterday to put together a spread sheet of my LDR data.

The circuit used for testing consisted of two triodes: the first as the LFO, the second to drive the LED using the cathode voltage. The LFO output was attenuated to about a 6-7V Peak to Peak signal used to drive the driving triode. For testing, the driving triode was biased at approximately 1.5V, adjusted just at the Red diode On state. Minimum test speed was about 2.5Hz, and Maximum Speed was about 7.1Hz.
A Fluke 177 was set to read Minimum Ohms, then Maximum Ohms, and lastly, Average Ohms for each LDR and speed setting.  Testing was done with each roach protected from light since it was not sealed on the LDR end to facilitate LDR changes.

Testing was done on the GL5506, GL5516, GL5528, GL5537-1, GL5537-2, GL5539, GL5549; the Light and Darkness light and dark resistance specifications increasing with each model in the order listed. The commercially available VTL5C1 was also tested.

The document Roach Data 1 shows the raw data and data sorted by Maximum Ohms at Minimum Speed, and Roach Data 2 shows the raw data and data sorted by Minimum Ohms at Minimum speed.

General Observations:

•   It must be noted that the data collected is only representative of the specific parameters which the LDRs were tested, therefore changes in LFO speed, the driver bias, grid voltage, gain, and driving wave form, among others, will affect results. The data does, however, allow for a general comparison of dynamic performance among the various models.

•   Each model group does show performance groupings consistent with its increasing Light and Dark resistance specifications, but the data does show some wide variations. Some of these variations could be from supplier issues. Consider purchasing directly from the manufacturer.

•   The VTL5C1 data demonstrates similar performance to the GL5506 and GL5516 models.

I am still working on the Red and Green LED data.

Thanks for this. Quick observation, the column titles for the fast speed data are incorrect.  It's easy to interpret what's happening, so not a big deal.

This seems to confirm the 5k to 500k being the "action" zone, and the significant drop in maximum resistance at higher speeds is indicative of the observed volume drop.

I am curious about the sampling speed of the meter used, as it's possible that the meter is missing the peaks at faster speed and exaggerating the drop in max resistance.

There is also vast unit-unit variation.  So there's really isn't a single best model to buy. This may also explain why the commercially available roaches vary in performance--and why people experimenting with say the Tremonator are having different degrees of success.

Thanks again. This is great to have.

[astronomicum]

Thanks for this. Quick observation, the column titles for the fast speed data are incorrect.  It's easy to interpret what's happening, so not a big deal.

This seems to confirm the 5k to 500k being the "action" zone, and the significant drop in maximum resistance at higher speeds is indicative of the observed volume drop.

I am curious about the sampling speed of the meter used, as it's possible that the meter is missing the peaks at faster speed and exaggerating the drop in max resistance.

There is also vast unit-unit variation.  So there's really isn't a single best model to buy. This may also explain why the commercially available roaches vary in performance--and why people experimenting with say the Tremonator are having different degrees of success.

Thanks again. This is great to have.

I corrected the files and reposted them.

Even down to 250K still puts you in a good effective range.

I attached the Fluke 177 manual page on Min, Max, Ave accuracy.

Variation in LDR performance could very well explain some of the issues.

Biasing of the LED could also explain others. Referring to the attenuation curve, attenuation from 1dB and up starts at a resistance of about 250K ohms, increasing exponentially as resistance from the signal path to ground decreases. It is important to note that a wider LDR dynamic resistance range would generally be thought to give the greatest dB reduction, which is true, but only if that range has a low Light resistance, AND has been adjusted to its optimum position of the curve (unlike a potentiometer which has a fixed position, the position of the LDR range will change based on the bias voltage). The smaller the LDR dynamic range, the more critical the adjustment. Getting the range properly positioned can be done by adjusting the bias voltage so the LED is in some level of LOW On state. More On pushes the dynamic range to the right on the curve, less On or even Off, pushes the range to the left. Put another way, too much On, then not enough Off; too much Off, then not enough On. Properly adjusted, I believe any of the LDRs up through the model GL-5537-1 can be used successfully IF the bias is adjusted (I use a 1K trimmer on the cathode along with a tail resistor) to get it's position in the most effective area of the curve. I do this by lifting a leg of the LDR in an otherwise completed circuit with a meter across the LDR, the circuit energized, and the LFO off, then adjust the bias so the LDR is showing a reduction in maximum resistance indicating the LED is just starting to illuminate (about 1-2M for the GL5537-1). I would expect targets to be a bit lower for the smaller dynamic range LDRs. Once complete, measure, as accurately as possible, the cathode voltage and keep it in the amp records. As the tube ages or if you change a tube, just adjust the bias voltage back to what your records indicate. If you end up changing a roach, repeat the procedure. I use the GL5537-1 due to its low Light resistance, its wide dynamic range in low speed (my preference) and its higher max range at high speed to minimize attenuation due to LDR lag.