[sdiy] saw/ramp from square?

[Magnus Danielson]

Ingo Debus skrev:
> 
> Am 01.01.2009 um 13:02 schrieb Magnus Danielson:
> 
>>>>
>>> Yes, Elektor had a square-to-saw converter that worked independent of 
>>> frequency. I could dig out the schematic if you like.
>>
>> You could use a feedback loop in combination with a peak-detector such 
>> that previous cycles saw-peak is sampled and controlling the 
>> charge-current. This will cause some amplitude modulation but it could 
>> be reduced to some degree, and as always with peak detectors this 
>> would be most significant at low frequencies. If the peak lowpass 
>> filtering was adapted such modulation could be reduced to a minimum.
> 
> Yes, I think it worked that way. I'll send the article to Magnus via 
> private mail.

Having had the article Babel-fish converted into English I can now make 
some comments on it. While the circuit is fairly simple (4 transistors, 
5 caps, 3 resistors and a diode excluding the buffer), it is really a 
feed-forward design rather than feed-back design. It detects the period 
length in one branch and then use that to create a charge current in the 
other to compensate for variable amplitude and maintain a fair sawtooth 
curve. There is something mentioned about a patent for it, but I am 
unable to locate said patent.

Anyway, the frequency to voltage conversion method is pretty straight 
forward, it uses the fact that the squarewave edge looks the same 
regardless of frequency, so a simple highpass filter makes a transistor 
load some current for a fixed period into a large condensator. Each pump 
time will not change the voltage much, so the DC level is more or less 
linear with frequency. Crude but effective.

This voltage is then used to charge the integrator cap, which is reset 
periodically by a separate transistor. As the charge current changes 
with frequency it will (mostly) cancel the amplitude variations that a 
constant current would produce.

It is clear that minor modernization can be performed to this circuit to 
improve on it. A feedback scheme could improve the response as a quick 
frequency change occur, something which the origial circuit is less able 
to handle. This is also brings up a potenitial problem with this feed 
forward design, at least in this implementation, that it can be slow on 
quick changes. A feed-forward/backward design only using the previous 
cycle would be much quicker to handle such changes.

Cheers,
Magnus