Variable duty cycle sawtooth generator

[Smuki]

Hello!

I have been trying to build a generator, which will output a triangle/sawtooth wave with variable frequency and duty cycle. I came up with the circuit attached to the post. It works quite nice up to around 500kHz (my aim is 1Mhz) but then the problems begin. Actually the most disturbing problem is the distortion seen on the attached images. I have traced the problem to the junction capacitance of the two diodes (decreasing the value of this parameter in the simulation reduces the distortion). I've been searching the web for diodes with small junction capacitance, but was unable to find an adequate one (according to the simulation Cj would have to be lower than 300fF - the lowest I could find has 1,5pF). One solution is adding diodes in series and lowering the capacitance that way, but this solution is by no means an elegant one and brings with it other problems (bigger voltage drops->smaller current and frequency, larger variation in the combined voltage drop which is dependent on the current,...).
My question therefore is: how to solve this problem? Any ideas for circuit improvement  which resolves the problem, or a completely different design which satisfies the mentioned requirements (variable frequency up to 1Mhz, variable duty cycle-preferably from 1% to 99%) would be greatly appreciated. I would however like to avoid the use of OTAs as in this idea: http://www.till.com/articles/VariableSaw/index.html

The second problem I encountered presents itself already at frequencies at which the above described problem does not yet cause noticeable distortion. This second problem is the changing of frequency with the change of duty cycle (rotation of potentiometer P2). At 50% duty cycle the frequency is largest. Then at 1% and 99% duty cycle, the frequency drops.For example: at 113kHz with 50% DC, it drops to about 110kHz at 1% and 99% DC.
So my second question is: where is the reason for that? When tossing around the equations I came to the conclusion, that the frequency should stay the same for the given circuit. Maybe some parasitic effects are causing the problem again? One reason might be the exponent-like changing of the diodes forward voltage with current, but I also simulated the circuit with ideal diodes and the problem remains. Any ideas?

I am fairly new to electronics so I am hoping that the solutions will be quite obvious to some of you more experienced members of the forum and wil therefore not take too much of your time.

I thank you in advance for all your answers and help.

Cheers,
Nejc

[audioguru]

The datasheet for the old NE5534 says that is is internally frequency compensated for a minimum voltage gain of 3. But you have one as a follower with a voltage gain of 1 so it will probably oscillate or have wild overshoots.

Its gain is only about 20 at 1MHz so a triangle-wave will be very curved.

[Smuki]

Thank you for the reply.

I've missed the minimum voltage gain 3, but in the simulation the output of the follower outputs a nice square wave (at least in the spice simulation) despite the gain lower than three. Another thing which indicates, that the opamp is not to blame is, that the problem (the curved triangle wave) disappears by only lowering the diodes junction capacitance. After that correction the output triangle is as it should be, with the ne5534s in the circuit.

I will however try a different opamp when I get home and see what happens. Does anyone have any recommendations which one to use? As already mentioned, I am fairly new to electronics and do not yet have my own  list of favorite opamps. Therefore any opamp candidates for such a list will be very welcome.

cheers

[audioguru]

I think your simulation program did not read all the details on the datasheet for your old opamps.