Electro-Harmonix HOG/POG - any idea how they work?

[valhallasound]

Hi all:

I just tried out the Electro-Harmonix POG and MicroPOG, and I am VERY impressed with the pitch shift quality. I am also unsure of the algorithm that they are using. A few observations:

- Obviously not a divide-down octaver (like the Boss OC-2), or a full-wave rectifier octaver (like the Octavia). Much cleaner, no noticable distortion. Probably not a multi-band version of either of these techniques, as the POG offers detuned octaves that would not be obtainable with integer multiplication or division.

- No obvious time-domain windowing or combing artifacts. The transients can sound slightly comb-y for 2 octaves up, but in a weird way. There are pitches, but not one fixed comb pitch. Kinda chirpy, as if high frequencies had less delay than low frequencies.

- I couldn't make it freak out with close intervals on guitar. Most SOLA pitch shifters that I have heard will start putting out garbage on a minor second, major second, or major seventh interval. The POG and MicroPOG sounded very nice - a little bit of intermodulation, but not the randomness you can here with some of the pitch shifters in, say, Logic.

- Latency seemed fairly low. Not instantaneous, but not that big. Certainly not what I have heard with FFT techniques with large windows.

- The POG has 5 mixable pitch shift outputs (octave down, 1 octave up, 1 octave up detuned, 2 octaves up, 2 octaves up detuned).

- The MicroPOG was running on a Freescale DSP56364. This is the same chip that the Line 6 Tonecore pedals use.

So...any idea what algorithm is being used? I initially thought that it was a time-domain pitch shifter, but I could not hear the typical glitches of delay-based pitch shifters with a fixed crossfade rate, and I could not make it glitch like a SOLA pitch shifter. I am thinking that it is either a frequency domain pitch shifter (FFT or wavelet), or a multi-band pitch shifter.

Any ideas welcome. Or, if you have any artifacts that I should listen for, I will give it another try.

Thanks,

Sean Costello

[lalo]

would be great to know!
i'd like to do a very good pitch shifter plug..( only Synthedit, Synthmaker, MaxMSP here...sadly no C++)

did you discover any trick?

[ziptiesispro]

I haven't really tried anything to check the truth, but I remember reading somewhere that the MicroPOG uses an FFT (or DFT I can't remember; however, it would make more sense it's FFT due to DSP efficiency) based phase vocoder with a 4x overlap. I've been wanting to make something like the HOG (but more resembling a harmonizer), so I did some research a while back ago. But, life pulled me away, so I don't remember much. I might be delving back into this soon, so we'll see.

[CinningBao]

I've got a HOG and it sounds and behaves very FFTey - especially the SpectralGate button, which sounds like an FFT noise reduction algo. I need to dig out the manual, but I think it runs at 96Khz with 2048 size blocks.

[valhallasound]

I haven't really tried anything to check the truth, but I remember reading somewhere that the MicroPOG uses an FFT (or DFT I can't remember; however, it would make more sense it's FFT due to DSP efficiency) based phase vocoder with a 4x overlap. I've been wanting to make something like the HOG (but more resembling a harmonizer), so I did some research a while back ago. But, life pulled me away, so I don't remember much. I might be delving back into this soon, so we'll see.

I probably posted that about the MicroPOG in another forum. All speculation on my part. However, I am pretty sure that it is a phase vocoder approach, as opposed to time-domain. It is obvious that other EHX pedals use a frequency domain approach nowadays. The name of the V256 suggests the use of 256 frequency bands, which is a nice power-of-2 you've got right there, and probably uses a 512 point FFT for its work (or 1024 - it's been a while since I did frequency domain work).

Sean Costello