6 ms·

From the petition: > Licensed use of 2-25 MHz Band frequencies as proposed in this Petition would enable financial firms, including firms like the SMC members

by ke88y 3y ago

From the petition: > Licensed use of 2-25 MHz Band frequencies as proposed in this Petition would enable financial firms, including firms like the SMC members that act as market makers, to improve efficiency, increase liquidity, and reduce transaction costs in the financial markets in which they participate, thereby benefiting all market participants. I am generally sympathetic to commercial use of spectrum, but allocating this band to prop firms doing high-frequency trading is fucking nuts. Can we please use this spectrum for something that actually contributes anything real to the world?

mlyle 3y ago
Kinda wonder what they're going to send, too. I mean, sure: more direct paths, but there's not that much bandwidth, so the symbol rate is not that high (and the time it takes you to send your message is latency, too).
- eterevsky 3y ago
  Using more power directly increases bandwidth according to Shannon's theorem.
  - cozzyd 3y ago
    and also increases the probability of receiving the signal, as HF is finicky.
    - mlyle 3y ago
      These are the "same thing" -- energy per bit / noise density chooses your bit error rate (and not enough energy per bit goes off a cliff to 50% errors where you don't transfer any information).
      - cozzyd 3y ago
        yes that's fair! Not being able to receive the signal at all is a gross form of lack of information density.
        WJW 3y ago
        Technically there's not really such a thing as "Not being able to receive the signal at all", the signal is there it is just (many thousands of times) lower than the noise floor and almost impossible to recover. There are information theoretic tricks you can do with the shape of the signal to still fit a few bits/second in there but it is going to be pretty low. GPS and other satellite navigation systems are probably the most prevalent systems that do this; the satellites don't have all that much power but receivers make it work even with very small antennas.
        cozzyd 3y ago
        There is surely a quantum limit here. For example, try a THz signal through a wall...
        WJW 3y ago
        I would think even a Thz signal through a wall is still trillions of times above any quantum limit. Still, now you've got me curious and I'd be grateful for some proper boffins to elucidate us further: if you have an transmitter emitting single photons in an isotropic antenna pattern, could they still be received by a receiver even though there is no "wave" to speak of? Even if part of the information is going in another direction? There is some proper double slit-type weirdness going on there.
        mlyle 3y ago
        As to detection: thermal limits dominate. Optically we make photon counters out of cooled avalanche photodiodes or photomultiplier tubes that will repeatedly trip for each incoming photon (at high probability). But they also randomly trip at a rate because of internal thermal noise. As to emission, yes, you can make (close to) omnidirectional single photon emitters. But the odds of it interacting with your antenna or detector are low.
  - mlyle 3y ago
    More power increases throughput for a given bandwidth (not proportionally), but we're talking about 20kW, which is not so much (the path loss is high).
  - CamperBob2 3y ago
    It increases the channel capacity. Bandwidth (spectral occupancy) remains the same. (Apart from increased distortion associated with higher-power transmitters in the real world, of course.)
- RF_Enthusiast 3y ago
  I've seen what this is a counter proposal to on a wed-based SDR, and there were ten carriers within a 5 kHz space. The 10 carriers were switching on and off. It seemed to me like it was using a much more basic definition of digital, meaning off was "0" and on was "1". This would/could be a zero latency system, if the extended information was sent by legacy IP/internet, but the commands that require zero latency (i.e., "buy", "sell") were sent as a carrier on/off signal. This proposal is for a 50 kHz system, so if they used bandwidth in the same way, that would be 100 carriers sending on/off commands and/or combinations. This is for a Part 90 (of the FCC rules) operation, so I would guess they'd install a Part 90 UHF transmitter at the place business is being conducted to control or retransmit to the Part 90 shortwave transmitter, which would take up too much physical space in most buiness environments.
  - mlyle 3y ago
    > This would/could be a zero latency system No, this is just discrete multitone. Leaving aside the issue of errors, it still takes you time to know the carrier is gone. You'd have much better choices about how to do this that are more spectrally efficient and equivalent in latency. Reductio ad absurdum: if it were zero latency, you could turn on and off the signal infinitely fast, and then send an infinite amount of information across a narrow channel. > This proposal is for a 50 kHz system, so if they used bandwidth in the same way, that would be 100 carriers sending on/off commands and/or combinations. Yah, or broadly: it's like OFDM. You don't have to constrain yourself to one bit per carrier, and you can use error coding across multiple carriers, etc. Also important to consider authentication...
    - RF_Enthusiast 3y ago
      Yes, those are both good points. Also, a bad actor could straight up jam the signal. If you want to check out the signal of the competing system, it's just above or just below 15.735 MHz, using the sidebands of a shortwave broadcaster operating in Digital Radio Mondiale. 5 kHz is usually the DRMondiale signal, and 5 kHz is the sideband transmission.
      - 3y ago
- toomuchtodo 3y ago
  https://sniperinmahwah.wordpress.com/2018/05/07/shortwave-trading-part-i-the-west-chicago-tower-mystery/ https://sniperinmahwah.wordpress.com/2018/05/07/shortwave-tr... https://news.ycombinator.com/item?id=17028892 https://news.ycombinator.com/item?id=17028892
  - mlyle 3y ago
    Yes, I'm familiar... it doesn't answer my question (the past discussion had some slight speculation in the same direction).
- cycomanic 3y ago
  You're correct that data rates are not consistently high (although you will get bursts of traffic with significant rate), but a higher bandwidth channel allows you to reduce your latency as you can transmit your ethernet frame faster (even if it contains little data, the rest might be packed with random data).
  - mlyle 3y ago
    You misunderstand what I'm saying-- I'm saying that the throughput of a channel that's a few kilohertz wide isn't high, so it must necessarily add some delay to messages (making the advantage from HF over optical fiber smaller).
metacritic12 3y ago
Truly an allocation with zero or negative net value. Wouldn't they relatively be equal off if the latency for all firms was increased by say 100ms.
- ke88y 3y ago
  I'm fairly convinced that trading venues should be substantially slowed such that HFT isn't a thing. I am unconvinced that the benefits of liquidity are worth the resources allocated. I'm also skeptical of the Religion of Liquidity -- I think maybe after a certain point increasing liquidity is not necessarily a net good.
  - kmeisthax 3y ago
    There's actually an exchange that did this. In order to avoid software problems that would negate the slowing, they actually plugged their servers into a giant box full of fiber optic cable. If someone manages to get past that, then they don't have an HFT firm, they have a time machine. https://www.youtube.com/watch?v=d8BcCLLX4N4 https://www.youtube.com/watch?v=d8BcCLLX4N4
  - nickysielicki 3y ago
    actively preventing price discovery only makes the market more unfair and brutal.
    - ke88y 3y ago
      I've spent a loooong time going down this particular rabbit hole. I just don't buy it. We can spill a lot of ink going back and forth, but honestly? At the end of the day, "this seems like useless bullshit" is actually a really good instinct. I know that sounds a tad anti-intellectual, but I think HFT for market making is one of those cases where very smart people delude themselves out of an obviously correct instinctive understanding of the truth. But we can dive down the rabbit hole if you want.

by ke88y 3y ago

mlyle 3y ago

Kinda wonder what they're going to send, too. I mean, sure: more direct paths, but there's not that much bandwidth, so the symbol rate is not that high (and the time it takes you to send your message is latency, too).

eterevsky 3y ago
Using more power directly increases bandwidth according to Shannon's theorem.
- cozzyd 3y ago
  and also increases the probability of receiving the signal, as HF is finicky.
  - mlyle 3y ago
    These are the "same thing" -- energy per bit / noise density chooses your bit error rate (and not enough energy per bit goes off a cliff to 50% errors where you don't transfer any information).
    - cozzyd 3y ago
      yes that's fair! Not being able to receive the signal at all is a gross form of lack of information density.
      - WJW 3y ago
        Technically there's not really such a thing as "Not being able to receive the signal at all", the signal is there it is just (many thousands of times) lower than the noise floor and almost impossible to recover. There are information theoretic tricks you can do with the shape of the signal to still fit a few bits/second in there but it is going to be pretty low. GPS and other satellite navigation systems are probably the most prevalent systems that do this; the satellites don't have all that much power but receivers make it work even with very small antennas.
        cozzyd 3y ago
        There is surely a quantum limit here. For example, try a THz signal through a wall...
        WJW 3y ago
        I would think even a Thz signal through a wall is still trillions of times above any quantum limit. Still, now you've got me curious and I'd be grateful for some proper boffins to elucidate us further: if you have an transmitter emitting single photons in an isotropic antenna pattern, could they still be received by a receiver even though there is no "wave" to speak of? Even if part of the information is going in another direction? There is some proper double slit-type weirdness going on there.
        mlyle 3y ago
        As to detection: thermal limits dominate. Optically we make photon counters out of cooled avalanche photodiodes or photomultiplier tubes that will repeatedly trip for each incoming photon (at high probability). But they also randomly trip at a rate because of internal thermal noise. As to emission, yes, you can make (close to) omnidirectional single photon emitters. But the odds of it interacting with your antenna or detector are low.
- mlyle 3y ago
  More power increases throughput for a given bandwidth (not proportionally), but we're talking about 20kW, which is not so much (the path loss is high).
- CamperBob2 3y ago
  It increases the channel capacity. Bandwidth (spectral occupancy) remains the same. (Apart from increased distortion associated with higher-power transmitters in the real world, of course.)
RF_Enthusiast 3y ago
I've seen what this is a counter proposal to on a wed-based SDR, and there were ten carriers within a 5 kHz space. The 10 carriers were switching on and off. It seemed to me like it was using a much more basic definition of digital, meaning off was "0" and on was "1". This would/could be a zero latency system, if the extended information was sent by legacy IP/internet, but the commands that require zero latency (i.e., "buy", "sell") were sent as a carrier on/off signal. This proposal is for a 50 kHz system, so if they used bandwidth in the same way, that would be 100 carriers sending on/off commands and/or combinations. This is for a Part 90 (of the FCC rules) operation, so I would guess they'd install a Part 90 UHF transmitter at the place business is being conducted to control or retransmit to the Part 90 shortwave transmitter, which would take up too much physical space in most buiness environments.
- mlyle 3y ago
  > This would/could be a zero latency system No, this is just discrete multitone. Leaving aside the issue of errors, it still takes you time to know the carrier is gone. You'd have much better choices about how to do this that are more spectrally efficient and equivalent in latency. Reductio ad absurdum: if it were zero latency, you could turn on and off the signal infinitely fast, and then send an infinite amount of information across a narrow channel. > This proposal is for a 50 kHz system, so if they used bandwidth in the same way, that would be 100 carriers sending on/off commands and/or combinations. Yah, or broadly: it's like OFDM. You don't have to constrain yourself to one bit per carrier, and you can use error coding across multiple carriers, etc. Also important to consider authentication...
  - RF_Enthusiast 3y ago
    Yes, those are both good points. Also, a bad actor could straight up jam the signal. If you want to check out the signal of the competing system, it's just above or just below 15.735 MHz, using the sidebands of a shortwave broadcaster operating in Digital Radio Mondiale. 5 kHz is usually the DRMondiale signal, and 5 kHz is the sideband transmission.
    - 3y ago
toomuchtodo 3y ago
https://sniperinmahwah.wordpress.com/2018/05/07/shortwave-trading-part-i-the-west-chicago-tower-mystery/ https://sniperinmahwah.wordpress.com/2018/05/07/shortwave-tr... https://news.ycombinator.com/item?id=17028892 https://news.ycombinator.com/item?id=17028892
- mlyle 3y ago
  Yes, I'm familiar... it doesn't answer my question (the past discussion had some slight speculation in the same direction).
cycomanic 3y ago
You're correct that data rates are not consistently high (although you will get bursts of traffic with significant rate), but a higher bandwidth channel allows you to reduce your latency as you can transmit your ethernet frame faster (even if it contains little data, the rest might be packed with random data).
- mlyle 3y ago
  You misunderstand what I'm saying-- I'm saying that the throughput of a channel that's a few kilohertz wide isn't high, so it must necessarily add some delay to messages (making the advantage from HF over optical fiber smaller).

metacritic12 3y ago

Truly an allocation with zero or negative net value. Wouldn't they relatively be equal off if the latency for all firms was increased by say 100ms.

ke88y 3y ago
I'm fairly convinced that trading venues should be substantially slowed such that HFT isn't a thing. I am unconvinced that the benefits of liquidity are worth the resources allocated. I'm also skeptical of the Religion of Liquidity -- I think maybe after a certain point increasing liquidity is not necessarily a net good.
- kmeisthax 3y ago
  There's actually an exchange that did this. In order to avoid software problems that would negate the slowing, they actually plugged their servers into a giant box full of fiber optic cable. If someone manages to get past that, then they don't have an HFT firm, they have a time machine. https://www.youtube.com/watch?v=d8BcCLLX4N4 https://www.youtube.com/watch?v=d8BcCLLX4N4
- nickysielicki 3y ago
  actively preventing price discovery only makes the market more unfair and brutal.
  - ke88y 3y ago
    I've spent a loooong time going down this particular rabbit hole. I just don't buy it. We can spill a lot of ink going back and forth, but honestly? At the end of the day, "this seems like useless bullshit" is actually a really good instinct. I know that sounds a tad anti-intellectual, but I think HFT for market making is one of those cases where very smart people delude themselves out of an obviously correct instinctive understanding of the truth. But we can dive down the rabbit hole if you want.