• Ic To Delay Power Update

Please Note: PCBs are available for this project, and the purchase includes the PT2399 ICIntroductionThis is a project that's been waiting for some time, but I finally got my act together, purchased some ICs, and started experimenting. I found quite a few things that no-one else has mentioned (the PT2399 is a very popular delay IC) - some good, some not-so-good. All in all though, this is a capable chip, that does what's needed well and without a great deal of effort. It's also a great deal of fun to play with.One application that really needs a delay is the Surround Sound Processor (Project 18) - a standard Hafler matrix that will provide a passable rear channel signal. Even many so-called home-theatre amps use just this to generate the rear channels!

In order to confuse the brain, we really need to delay the signal, which makes it sound as if it were further away. This principle is used in virtually all commercial units, but the 'real' ones have a tendency to be somewhat more complex than a simple passive unit (although you may be surprised if you analyse some of these 'specialised' ICs).The digital delay presented here is expected to be sufficient for most applications, both for home theatre, music production and PA applications (see below for more on this). It provides a variable delay - the minimum is a little over 30ms - fractionally longer than ideal, but I doubt that anyone will hear it as a problem.

The shortest reliable delay is 50ms, this is the equivalent of being about 17 metres away from the rear speakers, and they will have a suitably 'distant' sound, even when relatively close to the listening position. The delay can be extended to over 1 second, but high frequency response is woeful - to put it kindly.The apparent 'distance' created by a delay can be calculated easily - sound travels at about 345mm/ms so 50ms gives 17 metresMany commercial units include extra features (different delay periods, reverberation, etc), which can be easily applied to this project if desired. Reverb is generally not useful (it simply clutters the sound, and if reverb is part of the original sound source (from the DVD for example), it will be included anyway. Such effects are great to impress your friends, but will become tiresome after a relatively short time, and end up detracting from the program material - you end up listening to the effects instead of the sound.To include the repeat function, it's simply a matter of feeding the output signal (via a pot and resistor) to the point marked 'A' on the circuit. A 20k pot and 10k resistor will work well - this is shown in Figure 4. To disable 'reverb', simply set the pot to minimum.

Other than for guitar effects, I expect that the vast majority of people will simply leave this out.PT2399 Digital Delay Unit PhotosA photo of the two versions is shown above. The one on the left has the minimum filtering, and is suitable for short delays (no more than 100ms), and that on the right shows the full version. You will note that I used Greencaps for the filter - the PCB layout is designed for MKT style caps as always, but I didn't have any 4.7nF MKT caps in stock. You may also note that the regulator ICs (78L05) appear to be installed backwards. This is a small error on the PCB overlay - and the regulators must be installed as shown in the construction article.The Delay CircuitThe delay circuit is shown in Figure 1, and uses the Princeton Technology PT2399 digital delay chip. This replaces the original Mitsubishi device described in. The PT2399 has been around for a while now, and for ages I have planned a project based on it.

Until now, I was unable to get any further - however I have now found a source of ICs, so it is worthwhile.The circuit as described in the data sheet is relatively simple and effective, but I've found that it can be dramatically simplified, with very little loss of performance. The data sheet has complete circuit diagrams, but they are rather messy and most of the pin functions are not explained. Likewise, many of the internal functions of the IC aren't explained either - I've been able to decode some, but it's still guesswork on a couple of functions.The first circuit shown (Figure 1) is (almost) a direct adaptation from the PT2399 data sheet, but with some significant simplifications. As shown it will give good performance over a reasonably wide frequency range. The filters are tuned to around 8.5 kHz (-3dB), and although this could be reduced or increased, there does not seem to be any good reason to do so. This seems to be the optimum response for rear channel speakers, so should be left alone.

The filter circuits use internal opamps, and only require the external components shown in Figure 1.Figure 1 - PT2399 Digital Delay UnitThe above schematic is based on the one in the application note, but is changed in quite a few areas. Component values are simplified, and large value electrolytic caps have been replaced by 10uF.

Smaller value electrolytic caps have been replaced by 10uF as well. This arrangement has been tested, and there is no apparent noise penalty, despite the reduced filter capacitor values. Having an on-board regulator helps a lot, because it presents the delay IC with a low impedance supply, without the need for large value electrolytic caps.The 5 Volt supply must be regulated, as anything over 6 Volts will destroy the delay chip.

Because of the regulator, it can handle any voltage from 9V up to a maximum of 15V or so. If a 5V supply is available, the 5V regulator may be omitted and the existing 5V supply used instead. The PT2399 draws a maximum of about 30mA (supply current falls as the clock frequency is reduced and delay is increased).

Any existing supply must have enough reserve to supply the extra current. It is essential that the 5V supply has a low impedance, or the internal clock oscillator may not even start, meaning that you get no output at all.Something that isn't explained at all in the data sheet is the filter topology.

The filters are low pass, multiple feedback types, and have a rolloff of 12dB/ octave. The final filter (using R5 and C8) adds an extra pole, making the output filter 18dB/ octave. The actual filters are shown below, so it makes a bit more sense. Showing parts hanging off an IC is not useful to determine exactly what's going on.Figure 2 - PT2399 Filter CircuitsThe filters shown above are using the PT2399's internal opamps, and are based on (but not identical to) those shown in the data sheet.

Ic To Delay Power Update

The simplification I used here is to make all filter resistors 10k, rather than a mix of 10k and 15k. Because the version shown above actually works better, and reduces the chance of errors caused by component mix-ups.While the data sheet provides a value of 100nF (or inexplicably 82nF for the echo circuit) for C4 and C9, it's hard to justify such a high value.

I recommend that you use 33nF caps in these locations, but you can go as low as 10nF.This is one thing that is rates barely a mention, even though it is covered briefly in one version of the data sheet. C4 and C9 (in Figure 1 & 3) appear to be used for switched capacitor filters, which track (more or less) the clock frequency. The second data sheet I located describes the function of these as 'modulated integrator' and 'demodulated integrator' by adding a capacitor, but has zero details of how this is done in the IC. The block diagram provided in the (slightly) more complete data sheet is not useful (although it does specify that the power supply should be nominally 5V.

It also claims that the maximum output is 2V RMS. In someone's dreams.With the recommended values of 100nF, the IC starts slew-rate limiting from about 1kHz and above. This means that the signal level at any frequency above 1kHz has to be below the slew rate threshold, so it falls at 6dB/octave from 1kHz. I've tested the delay with these caps reduced to 10nF, and while there is certainly a bit more HF noise, the delay also shows an improved high frequency response and far less slew-rate limiting. 33nF is not a bad compromise, but feel free to experiment.As it transpires, the circuit can be simplified quite dramatically. So much so, that I've tested the delay with no filters at all, and there is no audible noise. There is quite a bit of visible noise, but it's well outside the audio spectrum and is easily removed with the most rudimentary filter.

This is not so effective with very long delays, but the maximum recommended is around 300ms (I've extended it to over 1 second, but high frequency response is severely affected when the delay exceeds a few hundred milliseconds.Although you may see the input and output multiple feedback filters referred to elsewhere as 'anti-aliasing' filters, they're not. I tested the circuit with the highest audio signals that the delay IC would pass, then beyond. There was no aliasing at all. Once the signal passes a frequency determined by the clock (and presumably C4 and C9), the signal first goes into slew-rate limiting, then stops completely at a somewhat higher frequency.At 2kHz and with the suggested 100nF caps, the maximum signal level is about 500mV RMS, at 4kHz it's 250mV, and so on. If you plan to use the circuit for short delays only, the value of C4 and C9 can be reduced - tests with 10nF gave good results and low noise (without any additional filtering) with delay times up to about 50ms.

For longer delays - as used for an echo pedal for example - I recommend that C4 and C9 should not be reduced below 33nF.Also not mentioned anywhere. If total value of R10 and the optional pot is less than around 2k, the oscillator probably won't start, and you'll get no signal. If you need less than 50ms delay, it might prove necessary to include a relay across an external resistor (say 3.3k or thereabouts), so the oscillator will start reliably. The relay shorts out the external resistor after a short delay. Once the oscillator is running, it seems to be quite stable. There is also a transistor solution that I've tested, and it works fine - this is shown in Figure 4.

Note that this may or may not be a problem, depending on the batch of ICs you get.Figure 3 - Super-Simple PT2399 Delay CircuitThe version shown above is suitable for use as a fixed short delay, and makes a surprisingly good account of itself. As you can see, the circuit is dramatically simplified, with most of the filter components simply deleted. Because it's expected to be operating at a clock frequency of around 15MHz, the simplifications do not cause any loss of audio quality - if anything, it's better than with the fully configured filters. Response extends down to about 10Hz - it can go even lower, but there's no point.High frequency response is considerably better than the full version shown in Figure 1, and easily extends to 12kHz. Believe it or not, but it's even possible to leave out C3 (1nF) and C8 (10nF). I have tested the circuit extensively with no filters at all (apart from what I assume to be the switched capacitor filters using C4 and C9), and with short delays there is almost no visible noise, and no audible noise. With long delays, if there is considerable HF content in the audio signal, you will hear distorted sibilances and some other artefacts.

In general though, I'd be perfectly happy to use it just as shown. For guitar (or other musical instrument applications), it would be better to increase the value of R5 back to 2.7k. This will roll off the higher frequencies, but the echo/reverb effect is more natural, Real echoes usually have a rapidly diminishing HF content, as these signals are absorbed easily - even by the air itself.I included the 'repeat' pot and connections in this, and it's done in exactly the same for the version shown in Figure 1, except it's not shown. This is the basis for a reverb or echo effect, but as noted below, a single delay does not make a nice sounding reverb or echo. Also not shown is a method of modulating the clock frequency and hence delay time. This can become the basis for a tremolo circuit, but if used for that there will be a delay between striking a note and hearing it - this is surprisingly easy to hear (and makes it hard to play, even with only 50ms delay).

Try standing 17 metres from the speakers and trying to play normally - it's not easy!I'd love to be able to claim that it can be used for flanging effects as well, but the minimum delay is simply too great, and it won't work in this role. This is the one major complaint I have with the PT2399 - the minimum delay is too long, precluding it from many applications.Figure 4 - Simple Transistor Timing Delay CircuitIf you really want to use the minimum possible delay time (and your IC refuses to start with a low value timing resistor), you can add this simple transistor delay to the timing circuit. The 47uF cap is charged via R1 (10k resistor), and the base is supplied from the R2 (also 10k). Interestingly, the oscillator won't run if the timing resistor is open circuit, but it will start once a connection is made. This is in contrast to operation with a value below 2k - it not only won't start, but it cannot start itself even if the resistance is increased.

A lockout condition such as this is unacceptable, so the above circuit was devised to get around it. The delay is about 25ms - long enough for the PT2399 to get its act together. This has been tested fairly extensively, and I'm satisfied that it will solve startup problems with very low values for R10.Input And Output ConsiderationsThe circuit as shown needs to be driven from a low impedance - no more than perhaps 1k. Output impedance is fairly low, but the 2.7k (or 1k) final filter resistor means that the load should be at least 10k.

There is one thing that has considerable nuisance value, and that's the limited voltage that the IC can handle. In many cases, it will be necessary to attenuate the signal before it gets to the delay circuit, and boost it again afterwards.There is no specified maximum level in the data sheet, but testing reveals that it needs to be kept to no more than around 1 Volt RMS. The level must be below the maximum or distortion will be plainly evident. The output from a CD/DVD player is generally about 2V, so you may need to attenuate the input signal by around 10dB, and provide 10dB gain afterwards to get the level back again.

It can be reduced (and amplified) more, but at the expense of noise.The necessary gain (and optional mixing) stages can use basic opamp circuits, or ('universal' preamp/ mixer) can be employed for everything.Delay TimingThe following table is taken from the data sheet, and shows the delay time for various values of the timing resistor. I remain unconvinced that the figures shown are necessarily very accurate, and it's certain that there will be variations from one IC to the next. Delay times in excess of 200ms cause audible distortion, shown shaded in the table. You can certainly use the delay with even greater delays than shown, but you will hear the distortion through any halfway decent speakers.